Linking microarchitectural deterioration to altered tissue-level mechanical environments in ovariectomized mouse femora Authors: Murat Horasan Published: 2026 ABSTRACT Estrogen deficiency is a primary driver of osteoporotic bone fragility. While its morphometric characteristics are well-documented, the different tissue-level mechanical environments across femoral regions remain poorly characterized. This study employed an integrated experimental-computational approach, utilizing ex vivo femoral three-point bending data and high-resolution micro-computed tomography (microCT) images from a mouse model. We quantified morphometric parameters within distal metaphyseal cancellous, midshaft cortical, and proximal cortical volumes of interest. Specimen-specific, microCT-based finite element (FE) models were developed to replicate the experimental loading configuration and were validated against experimental displacement data. Local mechanical environments were analyzed using percentile-based metrics of maximum and minimum principal strains, and their relationships with structural parameters were assessed via linear regression. The estrogen-deficient group exhibited a lower quality of metaphyseal cancellous microarchitecture, marked by lower bone volume fraction, trabecular thickness, and trabecular number, alongside higher trabecular separation. In contrast, cortical bone displayed more modest but significant differences in area and moments of inertia. The validated FE models revealed that the ovariectomy (OVX) group had higher tissue-level strain magnitudes across all femoral regions. The most dramatic differences in strain levels were observed in the metaphyseal cancellous bone, where strains were primarily dictated by trabecular microarchitecture. Conversely, cortical bone strains remained governed by cross-sectional geometry. These findings demonstrate that estrogen deficiency is associated with a different distribution of mechanical loads at the tissue level, providing a mechanistic explanation for the higher bone fragility observed. Furthermore, this study highlights the utility of secondary FE analysis for extracting deep biomechanical insights from existing experimental datasets, offering a more nuanced understanding of how structural deficiency relates to mechanical failure in osteoporosis.Supplementary InformationThe online version contains supplementary material available at https://doi.org/10.1038/s41598-026-58765-8. Introduction Osteoporosis is a systemic skeletal disorder characterized by low bone mass, deterioration of bone microarchitecture, and increased fracture risk1,2. In the United States, osteoporosis affects over 10 million adults, and annual costs of osteoporotic fractures exceed $20 billion3,4. Healthcare costs are expected to continue rising as the population ages. Postmenopausal women are the most common sufferers of this disorder. Estrogen deficiency following menopause plays a central role in the pathogenesis of osteoporosis by accelerating bone turnover, promoting trabecular bone loss, and inducing cortical thinning, thereby compromising whole-bone mechanical competence5,6. To model postmenopausal bone loss, ovariectomized (OVX) rodents are widely used and represent one of the most well-established preclinical models of estrogen-deficiency–induced osteoporosis7,8. OVX rodents consistently exhibit reductions in trabecular bone volume, connectivity, and mechanical strength, closely resembling skeletal changes observed in postmenopausal women7–10. Bone strength is determined not only by bone mineral density but also by bone geometry, microarchitecture, and the spatial distribution of cortical and cancellous compartments11,12. High-resolution micro–computed tomography (microCT) has become the gold standard for three-dimensional assessment of rodent bone morphology, allowing quantitative evaluation of cortical thickness, cortical area, trabecular bone volume fraction, trabecular number, thickness, separation, and tissue mineral density within well-defined volumes of interest13. Standardized microCT guidelines have enabled reproducible assessment of age-related changes, OVX-induced bone loss, and responses to pharmacological or mechanical interventions in rodents14. Beyond morphometric analysis, microCT data can be integrated with computational modeling to assess bone mechanical behavior. MicroCT-based finite element analysis (microFE) has been extensively applied in small-animal studies to predict apparent stiffness and strength, estimate tissue-level strain distributions, and investigate how alterations in bone microarchitecture influence mechanical performance15–18. Validation studies have demonstrated that microCT-derived FE models can accurately reproduce experimental force–displacement behavior and local strain patterns when appropriate material properties and boundary conditions are employed19–21. Such models have also been instrumental in mechanobiology research, linking local mechanical stimuli to site-specific bone formation and resorption22–24. Three-point bending of the mouse femur is a well-established experimental technique for assessing whole-bone mechanical properties, including stiffness, yield behavior, ultimate load, and displacement12,25. However, classical beam-theory–based analyses provide only apparent mechanical properties and may be confounded by differences in bone size, cross-sectional geometry, and material distribution12,26. To overcome these limitations, several studies have combined three-point bending experiments with specimen-specific microCT-based FE models, enabling virtual mechanical testing and direct comparison between experimental and predicted force–displacement responses27–29. This integrated experimental–computational approach allows detailed evaluation of spatial strain distributions and identification of regions susceptible to failure under bending loads. Despite substantial progress, relatively few studies have examined OVX-induced differences in both cortical and cancellous compartments of the mouse femur using a fully integrated framework that combines experimental three-point bending, site-specific microCT morphometric analysis, validated microCT-based FEA, and correlation of local strain metrics with underlying bone morphology. Most OVX microCT studies focus primarily on trabecular-rich regions, such as the proximal tibia or distal femoral metaphysis, and report global morphometric outcomes without directly relating them to local mechanical strain environments30–32. Characterizing the local mechanical environment within both cortical and cancellous bone is particularly important because these compartments respond differently to estrogen deficiency and mechanical loading33. OVX induces pronounced trabecular bone loss, accompanied by more modest but mechanically significant reductions in cortical thickness and cross-sectional geometry8–10,30,34. MicroFE models derived from high-resolution microCT images are capable of resolving heterogeneous strain distributions within the cortical midshaft and metaphyseal cancellous bone under physiologically relevant loading conditions16–19. Linking these strain fields to morphological parameters such as bone volume fraction, cortical area, trabecular thickness, and separation may help identify structural determinants of local mechanical behavior and susceptibility to microdamage and fracture35. Rigorous validation of FE predictions against experimental measurements is essential if these models are to be used as reliable tools for hypothesis testing or virtual mechanical screening. Previous studies have emphasized the importance of matching experimental and computational boundary conditions and validating model predictions using measured displacements or strains at defined load levels16,27. However, cancellous and cortical strains in ovariectomized (OVX) mouse femurs under three-point bending, as well as the relationships between morphological parameters and these strains, have not yet been addressed. Therefore, the objective of the present study was to employ a cross-sectional design to investigate the mechanical consequences of estrogen deficiency in the mouse femur. By analyzing bone microarchitecture and the internal strain environment at a single time point (9 weeks post-surgery), we compared the biomechanical state of Sham-operated and OVX groups to evaluate the impacts of established bone loss. Specifically, we aimed to (1) analyze high-resolution femoral microCT images to quantify cortical and cancellous morphometric parameters within defined midshaft cortical, proximal cortical, and distal metaphyseal cancellous volumes of interest; (2) develop specimen-specific microCT-based finite element (FE) models of the femur subjected to three-point bending, validate FE-predicted displacements against experimentally measured displacements, and compute local principal strain distributions; and (3) examine relationships between microCT-derived morphological parameters and FE-predicted principal strains in both cortical and cancellous compartments. By integrating these cross-sectional morphometric analyses with validated FE modeling, this study seeks to elucidate how OVX-induced structural differences translate into altered local mechanical environments, thereby improving mechanistic understanding of osteoporotic bone fragility at both the whole-bone and tissue levels. Materials and methods This study presents homogeneous finite element models developed to calibrate and validate the previously published ex vivo 3-point bending tests36 for 17-week-old C57BL/6JOlaHsd mice, including seven sham-operated (sham) and six ovariectomized (OVX) mice. It also explores the relationships between morphological parameters and strains derived from the finite element models. Animals The experimental protocols and procedures were outlined earlier36. The protocols were reviewed and approved by the Local Ethical Committee for Animal Research at the University of the Basque Country (UPV/EHU, CEEA, ref M20/2019/176)36. All experiments were performed in accordance with relevant guidelines and regulations.36 In brief, 8-week-old ovariectomized (OVX) and sham-operated (sham) female mice were purchased from ENVIGO36. The animals were housed in controlled environmental conditions with a 12-hour light–dark cycle and free access to food and water36. At 17 weeks of age (9 weeks post-surgery), the animals were euthanized to ensure that the biomechanical analysis was performed on skeletally mature bone, where the effects of estrogen deficiency had fully manifested and stabilized36. Following euthanasia, femoral bones were harvested, stored frozen, and later analyzed using micro-computed tomography and biomechanical testing36. Fig. 1(A) Experimental setup for 3-point bending biomechanical tests on the mouse femur. The lower part is the fixed section of the apparatus. The upper moving part is the top roll. The force is measured on this part, as shown in the figure36. (B) Roll displacement and applied force on a representative femur36. (C) A finite element model of a mouse femur illustrating loading conditions and VOIs for replicating the three-point bending test. The midshaft cortical, proximal cortical, and distal metaphyseal cancellous volumes of interest (VOIs) are highlighted in green, blue, and red, respectively. (A) Experimental setup for 3-point bending biomechanical tests on the mouse femur. The lower part is the fixed section of the apparatus. The upper moving part is the top roll. The force is measured on this part, as shown in the figure36. (B) Roll displacement and applied force on a representative femur36. (C) A finite element model of a mouse femur illustrating loading conditions and VOIs for replicating the three-point bending test. The midshaft cortical, proximal cortical, and distal metaphyseal cancellous volumes of interest (VOIs) are highlighted in green, blue, and red, respectively. MicroCT imaging and analysis Details of the imaging procedure were also described previously36. Briefly, the femur samples were scanned with an isotropic voxel resolution of 13.7 microns (SkyScan1172, Bruker MicroCT, Kontich, Belgium)36. For the scanning procedure, femurs were placed vertically in a tube sample holder and fixed in foam to avoid slight movements during high-resolution scanning. While the bones were scanned in a dry environment, beam-hardening correction (60%) and ring artifact reduction (10) were applied during 3D volume reconstruction using NRecon software to mitigate scan artifacts and ensure accurate morphometric analysis36. An aluminum filter was used to reduce beam hardening effects36. For analysis, cortical and cancellous volumes of interest (VOI) were established (Fig. 1C). The first VOI, focusing on the distal metaphyseal cancellous volume, began approximately 1.5 mm above the distal femoral end, excluding the primary spongiosa and cortical shell, and represented 10% of the total femoral length16,18. The second VOI for midshaft cortical bone was located midway between the right lower roller and the upper loading roller, covering 2.5% of the total femoral length16,18. The third VOI, targeting the proximal cortical bone, was placed midway between the left lower roller and the upper loading roller, representing 2.5% of the total femoral length. A Gaussian filter with a 3 × 3 × 3 kernel, mean 0, and standard deviation 0.65 was applied to the microCT image stack to minimize high-frequency noise14. Histogram analysis guided the segmentation of the metaphyseal cancellous bone and the rest of the bone using global cancellous-based and cortical-based threshold values, respectively, across all femora. These thresholds were set using the metaphyseal cancellous and midshaft and proximal cortical VOIs14,16,18,37–39 following established methods37–39, and were verified through visual inspection. Parameters of midshaft and proximal cortical bone, such as cortical area (Ct.Ar), total area (Tt.Ar), marrow area (Ma.Ar), as well as maximum and minimum moments of inertia (Imax and Imin), were measured14. Additionally, metaphyseal cancellous bone parameters, including bone volume fraction (BV/TV), trabecular thickness (Tb.Th), trabecular separation (Tb.Sp), and trabecular number (Tb.N), were assessed14. Ex vivo femoral 3-point bending test The experimental procedure details were previously outlined36. In brief, on the testing day, femoral samples were thawed at room temperature while kept wrapped in saline-soaked gauze to maintain hydration36. Biomechanical properties were measured using a three-point bending test on a Zwick/Roell ZwickiLine Z1.0 universal testing machine with a 50 N load cell (Fig. 1A)36. The test employed a custom fixture with three stainless steel rollers (2.0 mm diameter): two fixed supports 8.0 mm apart at the bottom, and a central upper roller that moved vertically to apply load (Fig. 1A)36. Force–displacement data were captured using testXpert III software (Fig. 1B)36. For biomechanical testing, femoral specimens were placed on the two lower supports with the condyles facing upward36. Manual alignment was carefully performed to ensure the long axis of the bone was perpendicular to the support rollers and centered in the testing setup36. Before applying load, the force signal was zeroed, and the upper loading roller was lowered at 1 mm/min until reaching a preload of 0.01 N36. Then, the loading rate was increased to 3 mm/min, with force and displacement data recorded continuously (Fig. 1B)36. MicroCT-based finite element analysis For each specimen, segmented femoral micro-CT datasets were converted into three-dimensional tetrahedral meshes using a custom MATLAB-based meshing workflow40. The resulting meshes were subsequently converted to ANSYS-compatible solid geometries using an additional MATLAB script. Quadratic 10-node tetrahedral meshes were generated in ANSYS, resulting in approximately 2.63 ± 0.09 million elements for SHAM samples and 2.30 ± 0.37 million elements for OVX samples, with a maximum element size of 30 μm determined through a mesh convergence analysis (Table 1). To enable direct comparison between experimentally measured and numerically predicted displacements, the contact location of the upper loading roller was identified using voxel coordinates from the original micro-CT images. The contact locations of the lower rollers were also examined using voxel coordinates from the original micro-CT images. Distinct bone regions, including metaphyseal cancellous bone, midshaft cortical bone, and proximal cortical bone, were defined within the finite element model using corresponding region identifiers (Fig. 1C). Linearly elastic, homogeneous FE models were developed in ANSYS, with elastic moduli of 20 GPa for cortical bone and 1.2 GPa for cancellous bone41,42. Instead of modeling each trabecula individually, the bone was treated as a continuum, and microscale strains within the trabeculae were not analyzed. While the complex cancellous bone volume was explicitly modeled via micro-CT segmentation, the bone tissue was treated as a homogeneous isotropic continuum at the material level. Consequently, the analysis focused on the mechanical environment of the trabecular struts rather than sub-trabecular or matrix-level microscale strains. This approach primarily accounts for the significant difference in the elastic moduli of cortical and cancellous bone. The elastic moduli of cortical and trabecular bone were initially assumed to be 15 GPa and 1.2 GPa, respectively41,42. Prior studies performed sensitivity analysis using Cotter’s method reported that the elastic modulus of cortical bone is the significant FEA parameter18,43. Therefore, to validate the model against experimental three-point bending data, a calibration was performed by testing cortical moduli within a range of 13–20 GPa41. The FE models were ultimately validated using a cortical bone modulus of 20 GPa41 and a trabecular bone modulus of 1.2 GPa42. All elements representing the cortical bone and cancellous bone were given a Poisson’s ratio of 0.318,43,44. A vertical load was applied as a set of concentrated forces distributed over the finite element (FE) nodes located at the contact surface between the upper loading roller and the bone (Fig. 1C). To simulate the three-point bending test, the loading nodes were permitted to displace only along the vertical loading axis, while horizontal and out-of-plane translations were restricted to maintain structural alignment and prevent rigid body motion. All FE nodes on the contact surfaces between the bone and the two lower rollers were constrained against vertical displacement, while rotation about the longitudinal axis of the bone was left unconstrained. To prevent rigid-body motion, FE nodes at the contact surface with the left lower roller were constrained in the longitudinal direction of the bone, whereas those at the contact surface with the right lower roller were allowed to translate freely along the longitudinal axis. A peak compressive load level of 7 N was applied (Fig. 1C); this magnitude was selected because it lay within the linear region of the load–displacement curves for all sham-operated (SHAM) and ovariectomized (OVX) specimens (Fig. 1B). The finite element analysis was performed to replicate the ex vivo femoral three-point bending experiment. In total, 13 FE models were generated, including 7 models for SHAM specimens and 6 models for OVX specimens. Table 1Results of mesh convergence study for finite element modeling for a representative SHAM femoral specimen.Maximum element size (µm)Number of elementsMinimum principal strain (95th percentile) in the midshaft cortical bone (µε)Minimum principal strain (95th percentile) in the metaphyseal cancellous bone (µε)80574780−2050.1−1137.670771219−2073.1−1050.760117358−2082.3−1006.2501531935−2088.7−973.5401936610−2091.4−962.1302585150−2092.7−957.8 Table 1Results of mesh convergence study for finite element modeling for a representative SHAM femoral specimen.Maximum element size (µm)Number of elementsMinimum principal strain (95th percentile) in the midshaft cortical bone (µε)Minimum principal strain (95th percentile) in the metaphyseal cancellous bone (µε)80574780−2050.1−1137.670771219−2073.1−1050.760117358−2082.3−1006.2501531935−2088.7−973.5401936610−2091.4−962.1302585150−2092.7−957.8 Results of mesh convergence study for finite element modeling for a representative SHAM femoral specimen. Data analysis Following calibration of the finite element (FE) model using the ex vivo femoral three-point bending test, the model was validated by linear regression analysis comparing experimental and computational displacement measures. Experimental displacement corresponded to the vertical displacement of the upper loading roller recorded by the testing apparatus, whereas computational displacement was defined as the vertical displacement calculated at the FE nodes located at the contact surface between the upper loading roller and the bone. Maximum and minimum principal strains were calculated for each finite element within the distal metaphyseal cancellous and the midshaft and proximal cortical volumes of interest (VOIs). To evaluate differences in overall strain magnitudes between sham-operated (SHAM) and ovariectomized (OVX) femoral specimens, the principal strains were ranked and analyzed using percentile-based metrics (5th, 10th, 25th, 50th, 75th, 90th, and 95th percentiles) for cortical and cancellous VOIs, as previously described35,45. Each percentile value (e.g., 95th) represents the strain magnitude below which a specified proportion of bone elements resides (e.g., the upper 5% of elements exceed the 95th percentile)16,18,43. Differences in strain measures were evaluated using unpaired t-tests for SHAM and OVX samples. Relationships between strain magnitudes and cancellous or cortical microCT-derived parameters were assessed using linear regression analyses. All statistical analyses were performed in MATLAB, and statistical significance was defined as p < 0.05. Results Bone mass differences for SHAM and OVX femoral specimens Estrogen status (presence in SHAM and deficiency in OVX) was associated with distinct cancellous and cortical bone mass profiles, with more pronounced differences observed in distal metaphyseal cancellous bone than in midshaft or proximal cortical bone of the femur. Specifically, metaphyseal cancellous bone in the estrogen-deficient group exhibited a 51% lower bone volume fraction (BV/TV), an 18% lower trabecular thickness (Tb.Th), and a 20% lower trabecular number (Tb.N) compared to the Sham group, while trabecular separation (Tb.Sp) was 26% higher (Fig. 2, Table S1). In contrast, cortical bone showed smaller but significant differences; cortical area (Ct.Ar) and minimum moment of area (Imin) at the femoral midshaft were 12% and 13% lower, respectively, in the OVX group (Fig. 3, Table S2). Similarly, Ct.Ar and Imin of the femoral proximal cortical bone were 10% and 12% lower, respectively, under estrogen-deficient conditions (Fig. 4, Table S2). No significant difference in marrow area (Ma.Ar) was observed in the midshaft or proximal cortical bone (Figs. 3 and 4; Table S2). In line with this, measurements of periocortical dimensions, such as total tissue area (Tt.Ar), were approximately 6% lower in the estrogen-deficient group (Table S2), indicating that the structural differences associated with OVX were not characterized by periosteal expansion. Fig. 2Differences in microCT-measured parameters of the distal metaphyseal cancellous bone of the femora for SHAM and OVX samples. The distal metaphyseal cancellous volume of interest (VOI) began approximately 1.5 mm proximal to the distal femoral end, excluding the primary spongiosa and cortical shell, and extended 10% of the total femoral length. * p < 0.05 (unpaired t-test for SHAM vs. OVX limbs). Bars: mean + SD. Differences in microCT-measured parameters of the distal metaphyseal cancellous bone of the femora for SHAM and OVX samples. The distal metaphyseal cancellous volume of interest (VOI) began approximately 1.5 mm proximal to the distal femoral end, excluding the primary spongiosa and cortical shell, and extended 10% of the total femoral length. * p < 0.05 (unpaired t-test for SHAM vs. OVX limbs). Bars: mean + SD. Fig. 3Differences in microCT-measured parameters of the midshaft cortical bone of the femora for SHAM and OVX samples. The midshaft cortical VOI was centered midway between the right lower roller and the upper loading roller and spanned 2.5% of the total femoral length. * p < 0.05 (unpaired t-test for SHAM vs. OVX limbs). Bars: mean + SD. Differences in microCT-measured parameters of the midshaft cortical bone of the femora for SHAM and OVX samples. The midshaft cortical VOI was centered midway between the right lower roller and the upper loading roller and spanned 2.5% of the total femoral length. * p < 0.05 (unpaired t-test for SHAM vs. OVX limbs). Bars: mean + SD. Fig. 4Differences in microCT-measured parameters of the proximal cortical bone of the femora for SHAM and OVX samples. The proximal cortical VOI was centered midway between the left lower roller and the upper loading roller and spanned 2.5% of the total femoral length. * p < 0.05 (unpaired t-test for SHAM vs. OVX limbs). Bars: mean + SD. Differences in microCT-measured parameters of the proximal cortical bone of the femora for SHAM and OVX samples. The proximal cortical VOI was centered midway between the left lower roller and the upper loading roller and spanned 2.5% of the total femoral length. * p < 0.05 (unpaired t-test for SHAM vs. OVX limbs). Bars: mean + SD. Differences in femoral tissue-level mechanical environments associated with estrogen status For both SHAM and OVX samples, the measured displacement values matched closely with the predicted displacements at the contact surface between the upper loading roller and the bone, showing coefficients of determination (R²) of 0.95 and 0.99 for the SHAM and OVX groups, respectively (Fig. 5). Fig. 5Correlation between experimental and computational displacements at the upper roller–bone contact surface for (A) SHAM (n = 7) and (B) OVX (n = 6) femoral specimens. Correlation between experimental and computational displacements at the upper roller–bone contact surface for (A) SHAM (n = 7) and (B) OVX (n = 6) femoral specimens. The tissue-level mechanical environment exhibited distinct profiles throughout the femur under estrogen-deficient conditions, with pronounced differences in both cancellous and cortical bone regions (Table 2; Figs. 6 and 7). In all examined regions, OVX femora displayed consistently higher maximum and minimum principal strain magnitudes compared to SHAM femora, indicating an overall higher local strain environment in the OVX group. In the distal metaphyseal cancellous bone, the most substantial differences in strain distributions were observed. While numerical differences in the lowest strain percentiles were considerable (e.g., over 250% at the 5th percentile), these did not reach statistical significance (Table 2; Fig. 6A). However, statistically significant higher strain magnitudes were consistently observed in the 50th through 95th percentiles for maximum principal strain (ranging from 10 to 13% difference) and the 50th through 95th percentiles for minimum principal strain (10–13% difference). These differences in strain reflect the lower quality of cancellous bone microarchitecture associated with estrogen deficiency. In contrast, strain differences in cortical bone were more moderate but consistent across the femoral midshaft and proximal regions. In the midshaft cortical bone, OVX femora exhibited significantly higher maximum principal strains across all strain percentiles (9–17%) and minimum principal strains (12–26%) (Table 2; Figs. 6B and 7B). Similar patterns were observed in the proximal cortical bone, where the estrogen-deficient group exhibited significantly higher strain magnitudes, particularly at the 5th and 50th–75th percentiles for maximum principal strain (up to 18%) and across most percentiles for minimum principal strain (10–35%) (Table 2; Figs. 6C and 7C). Overall, these findings show that tissue-level strain magnitudes are higher throughout the femur in the OVX group, especially in the distal metaphyseal cancellous bone. This region-specific higher strain state underscores the higher mechanical vulnerability of cancellous bone in the estrogen-deficient model. Finite element–derived strain maps revealed distinct spatial distributions of maximum and minimum principal strains across the femur under three-point bending in both SHAM and OVX specimens (Figs. 8 and 9). Maximum principal strains were concentrated primarily along the anterior and lateral surfaces of the midshaft and proximal cortex, whereas compressive strains were localized predominantly to the posterior and medial surfaces, consistent with bending-induced tension–compression patterns. Strain magnitudes were lower and more uniformly distributed along the cortical shell, with localized strain concentrations in the metaphyseal cancellous region. Bone mass–strain relationships across femoral regions Distinctive relationships were observed between tissue-level principal strain percentiles and microCT-derived measures of bone mass and geometry, with different patterns between SHAM and OVX femora (Tables S3, S4, S5). In the metaphyseal cancellous bone, maximum and minimum principal strain magnitudes showed strong correlations with trabecular microarchitectural parameters (Table S3). For SHAM specimens, trabecular thickness (Tb.Th) consistently correlated well with both tensile and compressive strain percentiles across the distribution. Meanwhile, bone volume fraction (BV/TV) exhibited increasingly strong links at higher strain percentiles. In the OVX group, strain–microarchitecture relationships were different, characterized by lower correlations between strain magnitudes and BV/TV at higher percentiles. In contrast, Tb.Th and Tb.Sp remained moderately to strongly associated with both maximum and minimum principal strains, indicating different structural determinants of strain under estrogen-deficient conditions. Table 2Differences in microCT-based finite element analysis-derived strains of the metaphyseal cancellous bone and the midshaft and proximal cortical bone of the femora with estrogen status (presence in SHAM and deficiency in OVX).Maximum principal strain (µε)Minimum principal strain (µε)Sham (n = 7)OVX (n = 6)diff (%)Sham (n = 7)OVX (n = 6)diff (%)Metaphyseal cancellous bone5th percentile0.09 ± 0.020.32 ± 0.10255−0.10 ± 0.02−0.37 ± 0.0727010th percentile3.24 ± 0.683.27 ± 0.881−3.15 ± 0.7−3.34 ± 0.8625th percentile24.31 ± 5.5126.09 ± 6.417-24.03 ± 4.92−27.93 ± 6.691650th percentile71.41 ± 17.1878.91 ± 23.01 10 −76.19 ± 19.16−84.09 ± 19.121075th percentile174.40 ± 46.18196.34 ± 52.13 13 −188.61 ± 45.70−209.35 ± 51.13 11 90th percentile431.02 ± 143.01445.98 ± 140.37 3 −454.62 ± 119.40−511.89 ± 119.24 13 95th percentile871.91 ± 193.96894.71 ± 248.51 3 −885.40 ± 181.73−971.01 ± 211.49 10 Midshaft Cortical bone5th percentile476.34 ± 107.02520.82 ± 99.349−345.62 ± 99.17−385.83 ± 80.54 12 10th percentile524.58 ± 125.81571.94 ± 108.079−403.19 ± 118.58−470.41 ± 87.961725th percentile622.41 ± 155.01701.74 ± 137.11 13 −651.47 ± 153.46−821.15 ± 167.96 26 50th percentile806.05 ± 210.33944.31 ± 207.10 17 −1119.21 ± 255.58−1378.72 ± 262.91 23 75th percentile1125.98 ± 345.891313.08 ± 266.86 17 −1516.46 ± 357.64−1793.80 ± 383.10 18 90th percentile1550.03 ± 501.661771.42 ± 348.51 14 −1731.24 ± 406.82−2043.23 ± 443.81 18 95th percentile1764.49 ± 590.802015.50 ± 397.08 14 −1860.09 ± 446.43−2187.18 ± 459.73 18 Proximal Cortical bone5th percentile131.67 ± 31.72154.82 ± 28.98 18 −151.08 ± 45.29−203.36 ± 66.333510th percentile281.08 ± 83.19284.48 ± 78.331−245.61 ± 66.83−325.38 ± 41.163225th percentile495.79 ± 140.52532.01 ± 34.657−459.42 ± 125.92−567.65 ± 29.952450th percentile694.95 ± 172.75787.16 ± 26.96 13 −866.25 ± 230.89−982.39 ± 48.94 13 75th percentile967.34 ± 240.131086.35 ± 35.13 12 −1311.25 ± 301.21−1445.67 ± 113.24 10 90th percentile1382.46 ± 386.971420.97 ± 96.49 3 −1601.88 ± 374.25−1814.57 ± 198.78 13 95th percentile1611.70 ± 402.831646.07 ± 86.72 2 −1758.49 ± 447.95−1993.22 ± 217.02 13 Data are presented as mean ± SD.Bold indicates p < 0.05 by unpaired t-tests. Table 2Differences in microCT-based finite element analysis-derived strains of the metaphyseal cancellous bone and the midshaft and proximal cortical bone of the femora with estrogen status (presence in SHAM and deficiency in OVX).Maximum principal strain (µε)Minimum principal strain (µε)Sham (n = 7)OVX (n = 6)diff (%)Sham (n = 7)OVX (n = 6)diff (%)Metaphyseal cancellous bone5th percentile0.09 ± 0.020.32 ± 0.10255−0.10 ± 0.02−0.37 ± 0.0727010th percentile3.24 ± 0.683.27 ± 0.881−3.15 ± 0.7−3.34 ± 0.8625th percentile24.31 ± 5.5126.09 ± 6.417-24.03 ± 4.92−27.93 ± 6.691650th percentile71.41 ± 17.1878.91 ± 23.01 10 −76.19 ± 19.16−84.09 ± 19.121075th percentile174.40 ± 46.18196.34 ± 52.13 13 −188.61 ± 45.70−209.35 ± 51.13 11 90th percentile431.02 ± 143.01445.98 ± 140.37 3 −454.62 ± 119.40−511.89 ± 119.24 13 95th percentile871.91 ± 193.96894.71 ± 248.51 3 −885.40 ± 181.73−971.01 ± 211.49 10 Midshaft Cortical bone5th percentile476.34 ± 107.02520.82 ± 99.349−345.62 ± 99.17−385.83 ± 80.54 12 10th percentile524.58 ± 125.81571.94 ± 108.079−403.19 ± 118.58−470.41 ± 87.961725th percentile622.41 ± 155.01701.74 ± 137.11 13 −651.47 ± 153.46−821.15 ± 167.96 26 50th percentile806.05 ± 210.33944.31 ± 207.10 17 −1119.21 ± 255.58−1378.72 ± 262.91 23 75th percentile1125.98 ± 345.891313.08 ± 266.86 17 −1516.46 ± 357.64−1793.80 ± 383.10 18 90th percentile1550.03 ± 501.661771.42 ± 348.51 14 −1731.24 ± 406.82−2043.23 ± 443.81 18 95th percentile1764.49 ± 590.802015.50 ± 397.08 14 −1860.09 ± 446.43−2187.18 ± 459.73 18 Proximal Cortical bone5th percentile131.67 ± 31.72154.82 ± 28.98 18 −151.08 ± 45.29−203.36 ± 66.333510th percentile281.08 ± 83.19284.48 ± 78.331−245.61 ± 66.83−325.38 ± 41.163225th percentile495.79 ± 140.52532.01 ± 34.657−459.42 ± 125.92−567.65 ± 29.952450th percentile694.95 ± 172.75787.16 ± 26.96 13 −866.25 ± 230.89−982.39 ± 48.94 13 75th percentile967.34 ± 240.131086.35 ± 35.13 12 −1311.25 ± 301.21−1445.67 ± 113.24 10 90th percentile1382.46 ± 386.971420.97 ± 96.49 3 −1601.88 ± 374.25−1814.57 ± 198.78 13 95th percentile1611.70 ± 402.831646.07 ± 86.72 2 −1758.49 ± 447.95−1993.22 ± 217.02 13 Data are presented as mean ± SD.Bold indicates p < 0.05 by unpaired t-tests. Differences in microCT-based finite element analysis-derived strains of the metaphyseal cancellous bone and the midshaft and proximal cortical bone of the femora with estrogen status (presence in SHAM and deficiency in OVX). Data are presented as mean ± SD. Bold indicates p < 0.05 by unpaired t-tests. Fig. 6Effects of estrogen deficiency on maximum principal strain magnitudes in distal metaphyseal cancellous (A), midshaft cortical (B), and proximal cortical (C) bone tissues. * p < 0.05 by unpaired t-tests. Effects of estrogen deficiency on maximum principal strain magnitudes in distal metaphyseal cancellous (A), midshaft cortical (B), and proximal cortical (C) bone tissues. * p < 0.05 by unpaired t-tests. Fig. 7Effects of estrogen deficiency on minimum principal strain magnitudes in distal metaphyseal cancellous (A), midshaft cortical (B), and proximal cortical (C) bone tissues. * p < 0.05 by unpaired t-tests. Effects of estrogen deficiency on minimum principal strain magnitudes in distal metaphyseal cancellous (A), midshaft cortical (B), and proximal cortical (C) bone tissues. * p < 0.05 by unpaired t-tests. Fig. 8Finite element–derived maximum principal strain (µε) distribution in the mouse femur during a 7-N three-point bending test for a representative sham sample and a representative OVX sample. Whole-bone strain maps are shown for the anterior, medial, posterior, and lateral surfaces (top to bottom). Strain distributions in the proximal cortical, midshaft cortical, and metaphyseal cancellous VOIs are also shown. Finite element–derived maximum principal strain (µε) distribution in the mouse femur during a 7-N three-point bending test for a representative sham sample and a representative OVX sample. Whole-bone strain maps are shown for the anterior, medial, posterior, and lateral surfaces (top to bottom). Strain distributions in the proximal cortical, midshaft cortical, and metaphyseal cancellous VOIs are also shown. Fig. 9Finite element–derived mimimum principal strain (µε) distribution in the mouse femur during a 7-N three-point bending test for a representative sham sample and a representative OVX sample. Whole-bone strain maps are shown for the anterior, medial, posterior, and lateral surfaces (top to bottom). Strain distributions in the proximal cortical, midshaft cortical, and metaphyseal cancellous VOIs are also shown. Finite element–derived mimimum principal strain (µε) distribution in the mouse femur during a 7-N three-point bending test for a representative sham sample and a representative OVX sample. Whole-bone strain maps are shown for the anterior, medial, posterior, and lateral surfaces (top to bottom). Strain distributions in the proximal cortical, midshaft cortical, and metaphyseal cancellous VOIs are also shown. In the midshaft cortical bone, tissue-level strain magnitudes were mainly linked to cortical geometric features rather than marrow area (Table S4). In SHAM femora, maximum principal strain percentiles displayed moderate to strong correlations with cortical area (Ct.Ar) and the minimum second moment of area (Imin) across most percentiles, while correlations with marrow area (Ma.Ar) were weak or negligible. Minimum principal strains showed similar but generally weaker links with cortical geometry. In OVX femora, strain–geometry relationships varied more, with weaker correlations at lower percentiles and stronger ones at higher percentiles, especially between maximum principal strains and Imax, indicating increased sensitivity of the midshaft cortical strain environment to geometric differences after estrogen deficiency (Table S4). Similar patterns were seen in the proximal cortical bone (Table S5). In SHAM samples, both maximum and minimum principal strain magnitudes showed moderate correlations with cortical area and the minimum second moment of area across various percentiles, while their association with marrow area was weak. In OVX samples, the correlations between strain magnitudes and cortical geometric parameters continued but showed more variability across percentiles. Overall, these relationships were stronger at higher percentiles, especially for maximum principal strains (Table S5). Fig. 10Comparison of correlation coefficients (R2) between bone morphological parameters and principal strain percentiles in Sham and OVX groups: (A) correlation between bone volume fraction (BV/TV) and maximum principal strains in the metaphyseal cancellous bone, (B) correlation between trabecular thickness (Tb.Th) and minimum principal strains in the metaphyseal cancellous bone, (C) correlation between cortical area (Ct.Ar) and maximum principal strains in the midshaft cortical bone, and (D) correlation between minimum moment of inertia (Imin) and maximum principal strains in the proximal cortical bone. Comparison of correlation coefficients (R2) between bone morphological parameters and principal strain percentiles in Sham and OVX groups: (A) correlation between bone volume fraction (BV/TV) and maximum principal strains in the metaphyseal cancellous bone, (B) correlation between trabecular thickness (Tb.Th) and minimum principal strains in the metaphyseal cancellous bone, (C) correlation between cortical area (Ct.Ar) and maximum principal strains in the midshaft cortical bone, and (D) correlation between minimum moment of inertia (Imin) and maximum principal strains in the proximal cortical bone. To better visualize the impact of estrogen deficiency on the relationship between bone microarchitecture and mechanical environment, we analyzed the correlation coefficients (R2) between key morphological parameters and strain percentiles (Fig. 10). In the Sham group, strong correlations were observed between architectural integrity and strain reduction; specifically, BV/TV and Tb.Th were highly predictive of the 95th maximum strain (R2 = 0.83) and 5th minimum strain (R2 = 0.88), respectively (Fig. 10A,B). However, in the OVX group, these relationships were substantially weakened or abolished. For instance, the correlation between BV/TV and the 95th maximum strain dropped to nearly zero (R2 = 0.02), suggesting that in the osteoporotic state, bulk bone volume is no longer a reliable predictor of peak tensile strain. Similar trends were observed in the cortical bone (Fig. 10C,D), where Sham animals showed moderate to strong correlations between Ct.Ar or Imin and strain, whereas OVX animals exhibited a marked decoupling of geometry and mechanical performance. Overall, these findings show that tissue-level strain environments in the femur are strongly connected to regional bone mass and structural features. Cancellous bone strains mainly depend on trabecular microarchitecture (Table S3), while cortical bone strains are affected primarily by cortical geometry at the midshaft and proximal areas (Tables S4 and S5). Estrogen-deficient conditions are associated with differences in the strength and reliability of the relationship between bone mass and strain, especially at higher strain levels, underscoring the distinct mechanobiological states of femoral load-bearing capacity between the two groups. Discussion The present study integrated ex vivo three-point bending experiments, high-resolution microCT morphometry, and validated microCT-based finite element (FE) modeling to investigate the differences in femoral bone structure and tissue-level mechanical environments associated with estrogen deficiency in a mouse model of osteoporosis. By combining experimental measurements with specimen-specific computational analyses, this work directly links the lower bone mass and microarchitecture in the OVX group to different local strain distributions across cancellous and cortical compartments of the femur. Our findings confirm that femora from the estrogen-deficient group exhibit not only lower bone quantity but also a lower mechanical quality of the bone tissue, characterized by higher local strains, particularly in regions most susceptible to fracture. Consistent with prior studies, femora from the estrogen-deficient group exhibited lower metaphyseal cancellous bone microarchitecture quality. The observed 51% difference in bone volume fraction (BV/TV) and the significantly lower trabecular thickness (Tb.Th) and trabecular number (Tb.N) align with the landmark study by Bouxsein et al.30, which established that while the severity of OVX-induced bone loss can be strain-dependent in mice, the distal femur remains a primary site of rapid resorption. Furthermore, the aggressive nature of this trabecular decay compared to cortical bone suggests that initial microarchitecture is a critical predictor of the rate of decline, a concept recently emphasized by Li et al.32. In the cortical regions, the OVX group exhibited more modest but significant differences in cortical area and moments of inertia at both the midshaft and proximal regions compared to the Sham group, echoing findings that the cortical integrity associated with estrogen deficiency is lower due to higher microporosity and different mineralization states31. The minimal differences observed in the marrow area (Ma.Ar) suggest that estrogen deficiency in this model predominantly promotes pericortical erosion rather than compensatory periosteal expansion. This is an important distinction, as it suggests a failure of the bone to undergo the geometric modeling often regarded as a compensatory mechanism to maintain bending stiffness, a phenomenon observed in various rodent models of postmenopausal osteoporosis and even in studies with mechanical loading33. Importantly, these structural differences were associated with pronounced differences in tissue-level mechanical environments. Our FE models, which showed excellent agreement with experimental displacements (R2 > 0.94), demonstrated that under three-point bending, the estrogen-deficient group exhibited higher magnitudes of both tensile and compressive strains throughout the femur. The metaphyseal cancellous bone exhibited a pronounced shift across the entire strain spectrum. While the relative percentage difference was mathematically highest in the lower percentiles, the absolute strain magnitudes in those regions remained negligible. Instead, the potential for accelerated accumulation of microdamage is critically driven by the significant absolute strain increases observed within the higher percentiles (75th–95th percentiles). As Li et al.32 suggested, the initial peak trabecular microstructure is a strong predictor of the bone state; our data extends this by demonstrating that in the compromised architecture of the OVX group, the remaining tissue at these high-strain zones operates under elevated mechanical demands that could accelerate tissue-level microdamage. Regional differences in strain levels highlight the higher mechanical vulnerability of cancellous bone in the estrogen-deficient group. Lower trabecular connectivity is associated with lower load-sharing within the metaphysis and higher localized strain concentrations even under modest external loads. These findings support the concept that although cortical bone provides most of whole-bone stiffness, the cancellous compartment is critical in shaping the local mechanical environment. Our results align with the work of Warden et al.33, who observed that although mechanical loading can confer benefits independent of estrogen status, the baseline fragility associated with OVX, characterized here by spatial strain distributions shifted toward more extreme values, creates a significantly higher risk profile for the skeleton. Analysis of bone mass–strain relationships provided additional insight into the structural determinants of local mechanical behavior. In metaphyseal cancellous bone, strain magnitudes were most strongly associated with trabecular thickness (Tb.Th). In OVX specimens, however, the correlation between strain and BV/TV was lower at higher strain percentiles, suggesting that under conditions of significant bone deficiency, volume alone is a less reliable predictor of fracture risk than the actual architectural connectivity. In cortical bone, the strain magnitudes remained primarily related to geometric properties such as cortical area and moments of inertia, consistent with classical beam theory. However, the higher sensitivity of cortical strain to geometric differences in OVX femora suggests that estrogen deficiency is also associated with a lower safety factor of the cortical shell. This is likely compounded by differences in the material quality itself, such as the higher microporosity and different mineralization states reported by Sharma et al.31, which were not explicitly modeled here but likely contribute to the overall higher fragility observed in the OVX group. In cortical bone, strain magnitudes exhibited a primary relationship with geometric properties such as cortical area and second moments of inertia, rather than marrow area, at both the midshaft and proximal regions. This observation is consistent with classical beam-theory principles and previous experimental–computational studies of long bones under bending loads27,28. Notably, in OVX femora, strain–geometry relationships were characterized by higher variability and were stronger at higher strain percentiles compared to the Sham group. This suggests that the estrogen-deficient state is associated with a higher sensitivity of cortical strain environments to local geometric differences. These findings demonstrate the added value of microCT-based FE modeling in capturing spatial heterogeneity in mechanical behavior that cannot be resolved using apparent mechanical properties alone. Collectively, these results demonstrate that estrogen-deficient femora exhibit not only lower bone mass but also different distributions of mechanical loads at the tissue level across femoral regions. The higher local strain levels, particularly in cancellous bone, provide a mechanistic link between the lower microarchitectural quality in the OVX group and the higher fracture risk observed. Several limitations of this study should be acknowledged. The FE models assumed homogeneous, linearly elastic material properties. They did not explicitly resolve individual trabeculae, an approach commonly adopted in comparative microCT-based FE analyses, but one that may underestimate microscale strain heterogeneity. Additionally, loading conditions were limited to three-point bending at a single load level within the linear range, which may not capture strain behavior at higher loads or under alternative physiological loading conditions. Finally, the modest sample size inherent to detailed experimental–computational studies may limit generalizability and warrants confirmation in larger cohorts. Additionally, pinning the longitudinal displacement at the contact surface of the left lower roller to prevent rigid-body motion may introduce a slight over constrained to the FE models. While this boundary condition could minorly alter the exact spatial distribution of tissue-level strains, its identical application across both experimental groups ensures that the validity of our direct statistical comparisons remains unaffected. Despite these limitations, the integrated framework employed here provides a robust approach for linking bone morphology to mechanical function in osteoporotic bone. By validating microCT-based FE predictions against experimental measurements and systematically relating tissue-level strain distributions to region-specific microCT parameters, this study provides critical insights that extend beyond what is typically captured by traditional density-based measures, such as Bone Mineral Density (BMD) or Bone Volume Fraction (BV/TV). While such measures are effective at documenting global bone loss, they fail to resolve the spatial heterogeneity of the mechanical environment at the tissue level. Our specimen-specific FE models demonstrate that estrogen deficiency does not merely increase strain magnitudes as a uniform consequence of mass loss; rather, it accentuates local mechanical vulnerabilities and alters the standard structure–function relationships across the femoral compartments, particularly where the loss of trabecular connectivity disrupts normal load-sharing mechanisms. By identifying the distal metaphysis as a biomechanical weak link where the loss of trabecular connectivity, rather than volume alone, becomes the primary driver of redirected mechanical loads, this study advances the mechanistic understanding of bone fragility. Ultimately, these insights may prove valuable for improving fracture risk assessment and for evaluating therapeutic strategies aimed at preserving or restoring both bone structure and mechanical competence. Conclusions In this study, ex vivo three-point bending tests of the mouse femur were combined with high-resolution microCT imaging and validated, specimen-specific finite element (FE) modeling to establish a robust mechanistic link between region-specific structural differences and altered tissue-level mechanical environments in the estrogen-deficient femur. The new knowledge gained through this work highlights that skeletal fragility is driven by complex spatial shifts in strain distributions that are not fully predictable through apparent morphometric properties alone. While the estrogen-deficient group exhibited lower metaphyseal cancellous bone quality and smaller but significant differences in cortical geometry, our integrated approach reveals that the resulting mechanical vulnerability is highly region-dependent. Specifically, the most dramatic redistributions of tensile and compressive strains were observed in the metaphyseal cancellous bone, highlighting the extreme mechanical vulnerability of trabecular-rich regions in the estrogen-deficient state. In contrast, cortical regions showed more moderate but consistent differences, emphasizing the mechanical significance of even subtle geometric alterations under bending loads. Correlation analyses further clarified these relationships, indicating that cancellous bone strains were mainly influenced by trabecular microarchitecture, particularly trabecular thickness, whereas cortical bone strains were primarily determined by geometric properties such as cortical area and moments of inertia. Overall, these results underscore the utility of validated computational modeling as a powerful tool for extracting nuanced biomechanical insights and identifying localized mechanical hotspots that traditional measurements might overlook. By establishing region-specific bone structure–function relationships, this study offers a more comprehensive framework for evaluating fracture risk and for testing targeted interventions aimed at preserving both bone structure and mechanical competence in osteoporotic conditions. Supplementary Information Below is the link to the electronic supplementary material. Supplementary Material 1 Supplementary Material 1