Can body fat be both protective and harmful to bone? This study finds a U-shaped relationship in postmenopausal women, but a linear pattern in men over fifty. Osteoporosis is characterized by reduced bone mineral density and microarchitectural deterioration, and it poses a significant health risk, particularly in older individuals.

Fragility fractures, particularly of the hips or vertebrae, are common comorbidities of osteoporosis. The projected scale is substantial: an estimated 2.6 million hip fractures are anticipated globally by 2025, with projections rising to 7.3 to 21.3 million.

The obesity–health picture is not always simply harmful. Although obesity is widely accepted to be associated with increased cardiovascular events and all-cause mortality, paradoxical findings have emerged. Early studies indicated that overweight or obese patients on dialysis exhibited a lower mortality risk in a dialysis patient cohort, suggesting potential benefits of being slightly overweight in certain circumstances.

Nonlinear U-shaped associations have also been observed between obesity and all-cause mortality in patients with type two diabetes. That possibility makes the nonlinear relationship between adiposity and osteoporosis important to study further. The association between obesity and osteoporosis remains controversial because of intricate interactions, including direct mechanical loads and cellular effects, between adipocytes and osteoblasts.

Traditionally, body weight has been considered beneficial for bone formation because mechanical loading enhances metabolism, increases growth factor production, and synthesizes the bone matrix. But body weight and BMI do not adequately reflect obesity, especially because they provide limited information about body composition.

That is why indices targeting adiposity and abdominal obesity include WHR, WHtR, AVI, and VAI. Because studies suggest U-shaped or J-shaped relationships between BMI and health outcomes, the nonlinear nature of adiposity and its association with osteoporosis requires further investigation.

The interaction includes biologically active molecules such as adiponectin, estrogen, and interleukin-six secreted from adipocytes. Adipocytes and bone cells also share mesenchymal stem cells, which participate in the homeostasis of both cell types. The study therefore investigated various adiposity indices and osteoporosis diagnosed using DXA, employing nonlinear models.

In 2019, a prospective study enrolled males over fifty years and postmenopausal women. The study collected anthropometric measurements, blood biochemistry, and osteoporosis measurements made by DXA. Associations between adiposity indices and osteoporosis were analyzed using a generalized additive model and a segmented regression model.

Participants with missing anthropometric or biomedical measurements, and participants undergoing medical treatment for osteoporosis, were excluded. The exclusions left 2,193 participants for the final analysis, including 872 women and 1,321 men.

Bone mineral density was measured using a DXA scanner at the lumbar vertebrae from L1 through L4, the femoral neck, and the total hip. T-scores of at least negative one were classified as normal, scores from negative one to negative 2.5 indicated low bone mass, and scores at or below negative 2.5 were diagnosed as osteoporosis.

Participants with the lowest T-score at or below negative 2.5 among the lumbar spine, femoral neck, or total hip were diagnosed with osteoporosis. Basic information and anthropometric data, including sex, age, tobacco use, alcohol consumption, betel nut habits, and menopausal status, came from questionnaires and interviews conducted by trained nurses.

Waist and hip circumference were measured following WHO protocol, with participants standing with their arms relaxed at their sides, feet together, and weight evenly balanced on both feet. Traditional adiposity indices, including WC, BMI, WHR, and WHtR, were examined alongside novel indices such as AVI and VAI.

AVI estimates abdominal volume by conceptualizing the body as a cylinder or vertical cone, while VAI assesses visceral fat using WC, BMI, HDL, and TG, with sex-specific risk estimation. Generalized additive models avoid assuming specific functional forms and use smoothing techniques such as splines to fit curves while minimizing noise.

They were used to identify nonlinear relationships between the variables and the binary outcome of osteoporosis. An effective degree of freedom equal to one indicates a linear association, while larger values indicate a nonlinear relationship. Segmented regression, also called broken-line or piecewise regression, assumes different linear relationships joined at breakpoints that can be interpreted as thresholds.

Table one summarizes demographic and body-composition characteristics for two thousand one hundred ninety-three participants, divided into osteoporosis and non-osteoporosis groups overall and by sex. Osteoporosis was defined by a lowest T-score of less than or equal to negative two point five.

Across the total sample, the table reports significant differences in age, bone mineral density, waist and hip measurements, body mass index, and abdominal volume index, while visceral adiposity index and smoking were not statistically significant. The multivariable logistic regression incorporated significant variables from simple logistic regression and excluded variables showing collinearity based on Pearson’s correlation.

Most indices showed protective effects, with odds ratios below one, and remained statistically significant in both crude and adjusted models. WHR in women and VAI in men were not significant in either model, while WHtR in postmenopausal women became significant after adjustment.

The AUC in the female group exceeded 0.75 for all models, indicating good diagnostic power and superiority to the male group. Table four reports crude and adjusted logistic regression models linking anthropometric indices with osteoporosis, separately for females and males.

In females, the adjusted odds ratios for waist circumference, hip circumference, abdominal volume index, body mass index, visceral adiposity index, waist-to-hip ratio, and waist-to-height ratio are shown with their confidence intervals and p values; the corresponding AUC values range from zero point seventy-five to zero point seventy-seven.

The male results are presented in the same format, with AUC values from zero point sixty-five to zero point seventy-two. Figure one shows adjusted odds ratios for osteoporosis across seven adiposity indices, separately for females and males. The dashed line at one marks the reference point: estimates positioned to its left represent decreased risk, while those to the right represent increased risk, with horizontal bars showing uncertainty.

The authors report that most indices had statistically significant protective effects in both crude and adjusted models, supporting the importance of examining body shape and adiposity in relation to osteoporosis. Figure 2 depicts smooth relationships between adiposity indices and the outcome for osteoporosis.

Significant nonlinear associations were found for all indices in the women’s group, including WC, HC, AVI, BMI, VAI, WHR, and WHtR. An EDF value equal to one indicates a linear relationship, while an EDF greater than one suggests a nonlinear relation and a more complex curve than a quadratic.

The results revealed nonlinear protective effects of WC, AVI, BMI, VAI, and WHtR, whereas HC had a nearly linear association because its EDF was low. Figure two shows generalized additive model curves for abdominal volume index and body mass index, with separate female and male fits, confidence bands, and osteoporosis log-odds on the vertical axis.

Both panels visibly show curved, non-linear relationships rather than simple straight-line effects, while uncertainty widens toward the extremes. This matters because the authors report significant nonlinearity for both indices in women, so these adiposity measures require more nuanced interpretation than a single linear association.

Figure two continued shows smooth, sex-specific relationships between visceral adiposity index, waist-to-hip ratio, and waist-to-height ratio and the log-odds of osteoporosis. The red female and blue male curves are surrounded by shaded uncertainty bands, and their changing shapes indicate nonlinear associations rather than simple straight-line effects.

The authors report significant nonlinear associations for all indices in the women’s group, making these plots important for showing how osteoporosis risk may vary across different adiposity measures. Because U-shaped and sinusoidal patterns with potential inflection points were observed in postmenopausal women, Figure 3 evaluated BMD and adiposity indices using both GAM and SRM.

Abrupt changes in BMD were observed at WC of 94 centimeters, AVI of 17.67 square centimeters, VAI of 4.29, and WHtR of 0.61. Changes in BMD were less significant beyond an HC of 99 centimeters or a WHR of 0.72. The positive association between BMI and BMD became less significant after 25.74 kilograms per square meter.

Figure three compares generalized additive and segmented regression models for bone mineral density against waist and hip circumference in postmenopausal women. The curves suggest a change around a waist circumference of ninety-four centimeters, while the hip-circumference threshold is ninety-nine centimeters; shaded regions show the ninety-five percent confidence intervals.

The authors extend this analysis to other adiposity indices, identifying abrupt BMD changes for AVI, VAI, and WHtR as well, making these thresholds relevant for assessing skeletal-health risk and possible early intervention. Figure three continued shows generalized additive and segmented regression models relating bone mineral density to the adiposity indices AVI and BMI.

The authors use these models to identify possible inflection points, supporting a more detailed assessment of how adiposity relates to bone mineral density in postmenopausal women. Figure three continued shows generalized additive and segmented regression models relating bone mineral density to visceral adiposity index and waist-to-hip ratio.

Dashed lines mark the reported thresholds of four point two nine for visceral adiposity index and zero point seven two for waist-to-hip ratio, while shaded regions show uncertainty around the fitted relationships. The visual matters because it illustrates the potential inflection points that motivated the authors’ analysis of nonlinear adiposity–bone-density associations.

This panel examines the relationship between waist-to-height ratio, or WHtR, and bone mineral density in postmenopausal women using a generalized additive model and a segmented regression model. In both plots, a dashed vertical line marks the reported threshold of zero point six one, where the fitted relationship changes pattern, while the gray bands show uncertainty around the red fitted curve.

The figure matters because it visualizes the nonlinear association and the potential inflection point highlighted by the authors. In women, WC, HC, AVI, BMI, VAI, WHR, and WHtR had nonlinear protective effects against osteoporosis, although HC was nearly linear with an EDF of 1.001.

Nonlinear patterns in both GAM and SRM indicated an optimal body shape for postmenopausal women that avoids unlimited weight gain or loss. The observed thresholds for protection against BMD loss were WC of 94 centimeters, AVI of 17.67 square centimeters, VAI of 4.29, and WHtR of 0.61.

In men, no apparent inflection point was found in the GAM, which served as a linear indicator of bone mineral density deterioration. In men, VAI showed an inverse linear correlation with osteoporosis in the GAM model, but statistical significance was not achieved in the adjusted logistic regression.

The discrepancy might be due to fluctuations in TG levels, including variations depending on the day of the week when TG was measured. The results suggest distinct impact patterns of optimal adiposity on osteoporosis in women and men, with nonlinear effects particularly in women.

The study lacked information on exercise habits, and querying exercise habits could potentially introduce recall bias. The cross-sectional design prevented causal inferences. The osteoporosis diagnosis was made by DXA based on the WHO gold standard, and the analysis included almost all possible confounders without collinearity.

Further longitudinal studies should include hormonal changes in both genders and strictly control the timing of TG measurements to provide more evidence regarding gender differences. To maintain BMD and reduce the risk of osteoporosis, the study recommends thresholds of WC 94 centimeters, AVI 17.67 square centimeters, BMI 25.74 kilograms per square meter, VAI 4.29, or WHtR 0.61, along with weight-bearing exercise and dietary calcium supplement.

In men, AVI, BMI, WC, HC, and WHtR were identified as linear protective factors, while WHR showed a nonlinear sinusoidal relationship with osteoporosis. These findings highlight the importance of using specific adiposity indices to assess osteoporosis risk differently in men and women, particularly the continuous U-shaped interaction in postmenopausal women.

Future research should confirm these results and explore biological and hormonal changes, TG measurement timing, and the interaction with exercise using longitudinal approaches. The study links adiposity with lower osteoporosis risk overall, but the pattern differs by sex: women show a nonlinear relationship with thresholds, while men show a linear inverse association.