The Saturation Effect of Body Mass Index on Bone Mineral Density for People Over 50 Years Old: A Cross-Sectional Study of the US Population

Background: Previous studies had revealed that Body Mass Index (BMI) positively affected Bone Mineral Density (BMD). However, an excessively high BMI was detrimental to health, especially for the elderly. Moreover, it was elusive how much BMI was most beneficial for BMD in older adults to maintain. Objective: To investigate whether there was a BMI saturation effect value that existed to maintain optimal BMD. Methods: A cross-sectional study was conducted using the datasets of the National Health and Nutrition Examination Survey (NHANES) 2005–2006, 2007–2008, 2009–2010, 2013–2014, and 2017–2018. After adjusting for covariates, an analysis of the association between BMI and BMD in different femoral regions (Total femur, Femoral neck, Trochanter, Intertrochanter, and Ward's triangle) and lumbar spine regions (Total spine, L1, L2, L3, and L4) in the whole population was performed using the multivariate linear regression models, smoothing curve fitting, and saturation effects analysis models. Then, subgroup analyses were performed according to gender, age, and race. Results: A total of 10,910 participants (5,654 males and 5,256 females) over 50 years were enrolled in this population-based study. Multivariate linear regression analyses in the population older than 50 years showed that BMI was positively associated with femoral BMD and lumbar spine BMD (P < 0.001, respectively). Smoothing curve fitting showed that the relationship between BMI and BMD was not simply linear and that a saturation value existed. The saturation effect analysis showed that the BMI saturation value was 26.13 (kg/m2) in the total femur, 26.82 (kg/m2) in the total spine, and showed site-specificity in L1 (31.90 kg/m2) and L2 (30.89 kg/m2). The saturation values were consistent with the whole participants in males, while there was high variability in the females. BMI saturation values remained present in subgroup analyses by age and race, showing specificity in some age (60–70 years old) groups and in some races. Conclusions: Our study showed a saturation value association between BMI and BMD for people over 50 years old. Keeping the BMI in the slightly overweight value (around 26 kg/m2) might reduce other adverse effects while obtaining optimal BMD.


INTRODUCTION
Osteoporosis was a common disease of the skeletal system that imposed a substantial economic burden on society (1,2) and posed a risk of fracture in the elderly (3,4). As the most common clinical examination index, BMD was the gold standard for assessing osteoporosis (5). Obesity was one of the major public health issues to face in today's society, especially in the United States, with the highest adult obesity prevalence globally (6). Similarly, BMI was used in clinical situations to assess overweight and obesity (7).
Most of the available studies showed a positive association between BMI and BMD. A study in a 70-year-old population showed that patients with high BMI had a higher BMD and a lower risk of osteoporosis than those with a normal BMI (8). In older men and postmenopausal women, an increase in BMI was accompanied by BMD (9)(10)(11). However, an excessive BMI caused various other systemic diseases and complications such as hypertension (12), coronary heart disease (13), and type 2 diabetes (14). Possible mechanisms involved in the above diseases were the activation of the sympathetic nervous system (15), damage to the vascular endothelium (16), or insulin resistance (17).
Therefore, it was essential to strike a balance between BMI and BMD. However, existing researches were unclear exactly how much BMI was most beneficial for BMD while reducing the occurrence of other obesity-related complications. Therefore, we investigated the association between BMI and BMD in this work using a cross-sectional population survey sample from the NHANES (18) database for participants aged above 50 years, representing all regions and major ethnicities in the United States. We hypothesize that there was a saturation value of BMI and that keeping BMI at this value would result in an optimal balance between BMI and BMD.

Data Sources
The present study used the data sets from the NHANES database to carry out a cross-sectional study. NHANES is a national health information source to collect the health and nutrition of adults and children in the USA. The NHANES project had approved by the National Center of Health Statistics (NCHS) Research Ethics Review Board (ERB), and every willing participant signed a consent document before starting. (NCHS IRB/ERB Protocol Number: Protocol #2005-06, Continuation of Protocol #2005-06, Continuation of Protocol #2011-17, Protocol #2018-01, https:// www.cdc.gov/nchs/nhanes/irba98.htm).

Participants Selected
Data sets were used from NHANES 2005, NHANES 2007-2008, NHANES 2009, NHANES 2013-2014, and NHANES 2017-2018 because femur BMD data were only collected the above period. Before the study began, the following people were excluded from BMD testing: (1) Participants refused to measure BMD. (2) Pregnancy. (3) History of radiographic contrast agents in the past 7 days. (4) Test weight over 450 pounds (Beyond the measurement range of DXA equipment). (5) History of bilateral hip fractures, bilateral hip replacements, and pins or steel in both hips. (6) Participants with degenerative spine disease include severe scoliosis, stiffening of the spine, previous spinal fusion, laminectomy, and spine fracture. (7) Insufficient scans of the participant's vertebrae and hips to complete the whole scan. (8) Overlapping of body parts during BMD measurement, e.g., overlapping of hands and legs. (9) Other causes were affecting BMD measurements, such as cardiac stenting, morbid obesity producing excessive x-ray noise. A total of 50,463 participants were included in this study initially. After excluded 36,297 participants under 50 years old, 869 participants without BMI data, and 2,387 participants without BMD data, leaving 10,910 participants included in this study finally (Figure 1).

Definition of Lead Exposure BMI
BMI was calculated as weight (kg) divided by height (m) squared. Participants would not be excluded during the body measurements protocol for medical, safety, or other reasons. Values above the 99th percentile or below the 1st percentile were reviewed for data reasonableness. Data were reviewed for reasonableness based on a combination of height, weight, age, and gender of the participants. Data that were determined to be unrealistic after review will be artificially removed. All data during the BMI measurement would be the original data without any modifications.

Definition of Outcomes Femoral BMD and Lumbar Spine BMD
The primary outcomes of this study were femoral BMD and lumbar spine BMD. Dual-energy X-ray absorptiometry (DXA) was used to measure the BMD of the participants (19). BMD measurement regions include femoral regions (Total femur, Femoral neck, Trochanter, Intertrochanter, and Ward's triangle) and lumbar spine regions (Total spine, L1, L2, L3, and L4). All measurement data would be subject to data review.

Characteristics of the Study Participants
The characteristics of study participants were shown in   also had a higher BMD in BMI categorical analysis (β-Value, 95% CI, and P-value were showed in Table 2). When a smoothing curve fitting was conducted in the adjusted model (Figure 2), we found a saturation effect value between BMI and BMD. We further used the saturation effect analysis model to investigate the BMI turning point and found that the saturation effect value was 26

Subgroup Analysis
The subgroup analysis conducted by gender was shown in Supplementary Table 2 (Male) and Supplementary Table 3 (Female). After adjusting the covariates, the smoothing curve fitting showed a saturation effect value between BMI and BMD in male participants (Supplementary Figure 1). After adjusting for covariates and using a saturated effects model, we found that the BMI saturation value was 26.25 (kg/m 2 ) in total femur BMD and 26.84 (kg/m 2 ) in the total spine in males participants. This was consistent with the results obtained for the whole participants. However, saturation effects were not evident in the female participants, and saturation values for each femur and lumbar spine region varied widely (Supplementary Figure 2).   The subgroup analysis conducted by age was shown in Supplementary Table 4. After adjusting the covariates, the smoothing curve fitting also showed a saturation effect value. The saturation effect model showed that in the participants aged 60-70 years, the saturation values of BMI in the Femoral neck, Ward triangle, Total spine, and L1-L4 reached 32 (kg/m 2 ), and the results in other age groups were similar in the whole participants (Supplementary Figure 3).
The subgroup analysis conducted by race was shown in Supplementary Table 5. The results showed that the non-Hispanic population had a higher BMI saturation value at the lumbar spine region than at around 30 (kg/m 2 ), while others had a BMI saturation value at around 24 (kg/m 2 ). Mexican Americans and other Hispanics had higher BMI saturation values than other races at several sites in the femoral region, reaching around 35 (kg/m 2 ) in some areas (Supplementary Figure 4).

DISCUSSION
In the present study, a BMI saturation value (around 26 kg/m 2 ) was founded in the femur BMD and spine BMD in all participants over 50 years old. A positive association was found between BMI and BMD at BMI level <26 (kg/m 2 ) and a minimal increase in BMI level above 26 (kg/m 2 ), this being important for maintaining optimal BMD.
A positive association was found between BMI and BMD. When BMI was <26 (kg/m 2 ), for each unit increased in BMI, the BMD of the femur and spine increased by 0.023 (g/cm 2 ) and 0.018 (g/cm 2 ), which was consistent with the previous studies (20,21). A cross-sectional study in a Polish population also showed a positive association between BMI and BMD (22). Another study by Morin et al. indicated that low BMI predicted the future occurrence of osteoporosis and increased fracture risk (23). In the middle-aged and elderly female population, maintaining obesity could prevent the onset of postmenopausal bone loss (24). Similarly, a 10.5-year prospective cohort study suggested that obesity may be associated with the delayed bone loss (25). There were several possible mechanisms for a positive association between BMI and BMD as follows: (1) The increased static mechanical compliance produced by excessive fat accumulation was one of the possible mechanisms. Excessive fat accumulation and high body weight could impose greater static mechanical loads on the bones, and bone tissue produced a series of changes when it felt the mechanical forces exerted by the body (26,27). (2) More body fat in patients with high BMI was accompanied by an increase in various hormones, such as estrogen (28), insulin (29), and leptin (30), which had a beneficial effect on BMD by inhibiting bone resorption (31) and bone remodeling (32,33). (3) Androgens in adipose tissue were converted to estrogen, which would increase bone mass (34). (4) Studies also reported the effect of some genes (35) on BMD, such as the mutation of the Pro10 allele in tumor necrosis factor-β1 (TGF-β1), which is more frequent in obese patients (36).
Nevertheless, when BMI exceeded a specific value of 26 (kg/m 2 ), the BMD of the femur and spine increased with each unit of BMI by only 0.007 (g/cm 2 ) and 0.006 (g/cm 2 ). The mechanisms of keeping the BMI around 26 to have the most optimal BMD had not been fully explained. A review of the literature allowed us to draw some possible clues. (1) Genetic determinism. Bone growth trajectories (37) and peak bone mass (38) were determined early in life, which might be one of the possible explanations that BMD no longer increased after a limited increase in value in adults. Pocock et al. (39) explored genetic effects on BMD in 38 identical and 27 dizygotic twins and found that genetics determines adult bone mass. Genetic influences explained 75% of BMD variance, regardless of whether the twin was male (40) or female (41). Meanwhile, other acquired environmental factors, such as increased BMI (21), calcium intake (42), estrogen intake (43), and physical exercise (44), have a limit to the increase in BMD in adults. (2) Multi-factors coleading. Researches had shown a specific bone-adipose axis (45) between adipose and bone tissues within the body, connected by a variety of bioactive molecules and maintained bone homeostasis, which might be another possible mechanism for the presence of BMI saturation effects. Available studies clarified that bone and adipocytes originated from a common stem cell precursor and were competitive, with excess fat gain leading to bone loss (46,47). Several experiments in animal models induced by highfat diets had confirmed that BMD in obese animals decreased with increasing obesity (48)(49)(50). The PPAR-γ (Peroxisome Proliferator-Activated Receptor-γ) pathway (46) was involved in adipose and bone differentiation in obesity animal models in vivo. Activation of PPAR-γ (51) stimulates adipogenesis and bone loss, while inhibition of PPAR-γ prevents (52) bone loss. Other studies had shown that adipose was an endocrine organ that secreted inflammatory cells such as interleukin-1 (IL-1) and tumor necrosis factor-α (TNF-α) (53). The above cytokines inhibit BMD via the OPG/RANKL/RANK pathway (54). We hypothesized that the multiple factors mentioned above combined to cause an increase in BMI to a certain range without a significant increase in BMD, but direct experimental evidence was insufficient, so more advanced studies were still needed.
Our results showed that although the BMI saturation values were similar in the total femur (BMI saturation value, 26.13 kg/m 2 ) and total spine (BMI saturation value, 26.82 kg/m 2 ), they were significantly higher in the L1 (BMI saturation value, 31.09 kg/m 2 ) and L2 (BMI saturation value, 30.89 kg/m 2 ) than in the other sites, showed site-specificity. This might be related to the increased spondylarthrosis in the elderly, which led to reactive changes and increasing BMD in the lumbar (55). Unfortunately, we could not extract osteoarthritis-related data for a related study because of limitations in the database itself. Subgroup analysis by gender showed that the BMI saturation values for male participants' femur and spine were concentrated around 26 kg/m 2 . However, the BMI saturation values for female participants varied greatly by each site, which might be linked to the different levels of sex hormones in males and females (56). Age-subgrouped analysis showed that people aged 60-70 years had higher BMI saturation values for the spine than other age groups. A 10-year survey of bone loss rates in the elderly population from Japan suggested that this was related to the prevalence of spine osteophytes in this age group (57). In addition, race subgroup analysis revealed significant differences in BMI saturation values across races, suggesting that different genetic backgrounds and ancestry might be associated with this phenomenon (58).
Indeed, excessive BMI was detrimental to the elderly population. For one, high BMI brought a range of bone-related diseases, such as increased bone fragility (59) and increased fracture risk (60). On the other hand, obesity could lead to a variety of chronic diseases and complications, such as cardiovascular disease (61), type 2 diabetes (62), gallbladder disease (63), and fatty liver (64). More seriously, excessive obesity might be associated with increased cancer risk (65) and cancerrelated mortality (66). Therefore, we believed that keeping BMI at a reasonable value (around 26 kg/m 2 ) would maintain optimal BMD and reduce the risk of other obesity-related diseases and complications.
There were several limitations in this study. Firstly, this was a cross-sectional study, and we could only conclude whether there was an association between BMI and BMD, not a direct cause-and-effect correlation. Secondly, due to limitations in the database itself, we could not identify the medication-taking participants, menstrual and menopausal (female participants), and osteoarthritis participants, so our conclusions need to be interpreted with caution. Thirdly, we performed covariates detection to control confounders, but there might still be unpredictable covariates, such as medicine use, menstrual status (female), body fat percentage, etc. Last but not least, to the best of our knowledge, the present study was conducted in a US population, and the findings should be cautiously extended to other populations.

CONCLUSION
In the present study, we used the multivariate linear regression models, smoothing curve fitting, and saturation effects analysis models to investigate the association between BMI and BMD of people over 50 years old in the US. We found not only a simple linear positive association between BMI and BMD, but a saturation value existed, and this saturation value persisted during site-specific analysis, sex, age, and race subgroup analysis. This work indicated that keeping BMI at a reasonable value (around 26 kg/m 2 ) will provide the most significant benefit to older adults in maintaining optimal BMD and reducing other obesity-related diseases.

DATA AVAILABILITY STATEMENT
The original contributions presented in the study are included in the article/Supplementary Material, further inquiries can be directed to the corresponding author/s.

ETHICS STATEMENT
The studies involving human participants were reviewed and approved by NCHS IRB/ERB Protocol Number: Protocol #2005-06, Continuation of Protocol #2005-06, Continuation of Protocol #2011-17, Protocol #2018-01. The patients/participants provided their written informed consent to participate in this study.

AUTHOR CONTRIBUTIONS
MM and ZF conceived the study, data curation, data analysis, and draft writing. XL and GJ completed images and tables preparation. BG conceived the study and writing Instruction. YX conceived the study, funding acquisition, and writing-review editing. All