The Hidden Dangers of Plant-Based Diets Affecting Bone Health: A Cross-Sectional Study with U.S. National Health and Nutrition Examination Survey (NHANES) Data from 2005–2018

The plant-based dietary pattern has been recommended for its potential health and environmental benefits, but its association with bone loss needs to be further explored. This study aimed to investigate the association between three plant-based diet indexes and bone loss in 16,085 adults, using data from the National Health and Nutrition Examination Survey. Three plant-based diet indexes (PDI, hPDI, and uPDI) were calculated from two NHANES 24-h dietary recall interviews, to characterize a plant-based diet. A multinomial logistic regression model was used to estimate the odds ratios (OR) and 95% confidence intervals (95% CI). Higher hPDI and PDI were associated with increased risk of bone loss (ORQ5 vs. Q1 = 1.50; 95% CI: 1.24–1.81 for hPDI; ORQ5 vs. Q1 = 1.22; 95% CI: 1.03–1.45 for PDI), while higher uPDI was associated with increased risk of osteoporosis (ORQ5 vs. Q1 = 1.48; 95% CI: 1.04–2.11). A harmful association between plant-based diet indexes (hPDI and PDI) and osteopenia was observed at the lumbar spine rather than the femoral neck. We conducted several sensitivity analyses to ensure the robustness of results, including subgroup analysis, exclusion of people taking anti-osteoporotic and estrogenic drugs, further adjustment for menopausal status, corticosteroid usage, and dietary supplements, and calculation of E-value. Our study demonstrates the deleterious effects of a plant-based diet on bone health and emphasizes the importance of a balanced diet.


Introduction
Osteoporosis is a commonly occurring metabolic bone disease characterized by a reduction in bone mineral density (BMD) and the deterioration of bone microarchitecture, frequently leading to increased risk of bone pain and fragility fractures [1,2]. This disease affects over 200 million individuals globally, and its incidence continues to rise annually, particularly among middle-aged and elderly populations [3,4]. The annual medical cost of osteoporosis-related fractures in the United States alone is estimated to be around $17.9 billion annually, imposing a heavy economic burden [5]. Consequently, preventing osteoporosis has become a significant public health concern. In addition, although medication is an effective treatment for osteoporosis, the percentage of patients receiving medication treatment remains low due to the fragmented nature of the healthcare system

Study Population
The present study utilized publicly available data from the NHANES, affiliated with the National Center for Health Statistics (NCHS) of the Centers for Disease Control and Prevention (CDC). The NHANES was designed to evaluate nutrition status and the prevalence of disease in the US population. Due to the unavailability of BMD data for the femoral neck in 2011-2012 and 2015-2016, subject information was collected from five 2-year NHANES cycles (2005-2006, 2007-2008, 2009-2010, 2013-2014, and 2017-2018). Inclusion criteria were as follows: (i) participants aged ≥20 years; (ii) participants with complete BMD and dietary interview data; (iii) participants with reported energy intake levels within predefined limits (≥600 and ≤3500 kcal/d for women and ≥800 and ≤4200 kcal/d for men). Participants with energy intakes outside these limits were excluded from the analysis, as such extremes in energy intake may not represent the general population and could introduce bias into the results. We identified 50,463 potential participants from the five NHANES cycles (NHANES 2005(NHANES -2006(NHANES , 2007(NHANES -2008(NHANES , 2009(NHANES -2010(NHANES , 2013(NHANES -2014(NHANES , and 2017(NHANES -2018. After excluding the participants who did not meet the inclusion criteria, we ultimately recruited 16,085 participants. The detailed depiction of the inclusion and exclusion process is illustrated in Figure 1. All study participants provided informed consent, and the Ethics Review Board of the NCHS approved all study procedures.
Nutrients 2023, 15, x FOR PEER REVIEW 3 of 18 findings of this study may provide a valuable theoretical foundation for the development of strategies to prevent osteoporosis.

Study Population
The present study utilized publicly available data from the NHANES, affiliated with the National Center for Health Statistics (NCHS) of the Centers for Disease Control and Prevention (CDC). The NHANES was designed to evaluate nutrition status and the prevalence of disease in the US population. Due to the unavailability of BMD data for the femoral neck in 2011-2012 and 2015-2016, subject information was collected from five 2-year NHANES cycles (2005-2006, 2007-2008, 2009-2010, 2013-2014, and 2017-2018). Inclusion criteria were as follows: (i) participants aged ≥20 years; (ii) participants with complete BMD and dietary interview data; (iii) participants with reported energy intake levels within predefined limits (≥600 and ≤3500 kcal/d for women and ≥800 and ≤4200 kcal/d for men). Participants with energy intakes outside these limits were excluded from the analysis, as such extremes in energy intake may not represent the general population and could introduce bias into the results. We identified 50,463 potential participants from the five NHANES cycles (NHANES 2005(NHANES -2006(NHANES , 2007(NHANES -2008(NHANES , 2009(NHANES -2010(NHANES , 2013(NHANES -2014(NHANES , and 2017(NHANES -2018. After excluding the participants who did not meet the inclusion criteria, we ultimately recruited 16,085 participants. The detailed depiction of the inclusion and exclusion process is illustrated in Figure 1. All study participants provided informed consent, and the Ethics Review Board of the NCHS approved all study procedures.

Bone Mineral Density Assessment
All subjects had BMD (g/cm 2 ) measured at the lumbar spine (L1-L4) and the femoral neck using a dual-energy X-ray absorptiometry densitometer (Hologic QDR-4500A; Bedford, MA, USA). Participants were excluded from the DXA examination if they satisfied any of the following criteria: (i) participants who were pregnant; (ii) participants who weighed more than 450 pounds; (iii) participants who had a selfreported history of radiographic contrast material in the past 7 days; (iv) participants with bilateral hip fractures, replacements, or pinning. As recommended by the World Health Organization, the mean BMD of non-Hispanic white females aged 20-29 years from NHANES III was used as the reference group for the femoral neck, while the mean BMD of non-Hispanic white females aged 30-39 years from NHANES was used as the

Bone Mineral Density Assessment
All subjects had BMD (g/cm 2 ) measured at the lumbar spine (L1-L4) and the femoral neck using a dual-energy X-ray absorptiometry densitometer (Hologic QDR-4500A; Bedford, MA, USA). Participants were excluded from the DXA examination if they satisfied any of the following criteria: (i) participants who were pregnant; (ii) participants who weighed more than 450 pounds; (iii) participants who had a self-reported history of radiographic contrast material in the past 7 days; (iv) participants with bilateral hip fractures, replacements, or pinning. As recommended by the World Health Organization, the mean BMD of non-Hispanic white females aged 20-29 years from NHANES III was used as the reference group for the femoral neck, while the mean BMD of non-Hispanic white females aged 30-39 years from NHANES was used as the reference group for the lumbar spine [28,29]. In addition, the 16,085 participants were classified into three categories (normal, osteopenia, and osteoporosis) based on the minimum BMD T-score of the two measuring sites. Osteopenia was diagnosed according to a BMD T-score between −1.0 and −2.5, while osteoporosis was diagnosed according to a BMD T-score ≤ −2.5.

Plant-Based Diet Index
Dietary intake data were collected from two NHANES 24-h recall interviews and extracted after conversion to the respective food equivalents in the food-pattern-equivalence database. Additionally, dietary intake was estimated using the average of two 24-h recall data. To calculate the three plant-based diet indexes, we assessed the intake of 15 food groups, which were divided into three categories: healthy plant-based foods, unhealthy plant-based foods, and animal-based foods. We assigned positive or reverse scores to each food item based on the quintile of intake. Details of the 15 food groups and scoring rules are shown in Supplementary Table S1. Each subject's scores were summed to obtain a score for each index, with a theoretical range of 15 to 75. Finally, these index variables were treated as continuous (per 10-unit increment) and categorical (in quintiles), respectively.

Covariates
Participants' demographical characteristics (age, sex, ethnicity, educational level, poverty income ratio (PIR), body mass index (BMI), and marital status), lifestyle (smoking status and physical exercise), and history of disease (T2DM, hypertension, chronic kidney disease (CKD), cancer, and history of fracture) were considered as covariates in the present study. PIR was calculated as the ratio of the midpoint of the household's self-reported income to the corresponding poverty threshold for the household [30]. PIR values below 1 indicate poverty, while PIR values of 1 to 3 and above 3 reflect relatively higher socioeconomic status [31]. Based on serum cotinine levels, we defined three categories of (i) non-smoker (<1.0 ng/mL); (ii) environmental tobacco smoke (ETS) exposure (1.0-9.9 ng/mL); (iii) current smoker (≥10 ng/mL) [32]. Following the WHO guidelines, we defined four physical activity categories as: (i) inactive (participants with no regular physical activity); (ii) insufficient (<8.33 MET-hours/week); (iii) moderate (8.33-16.67 METhours/week); (iv) high (>16.67 MET-hours/week) [33]. Participants were diagnosed with T2DM if they satisfied any of the following criteria: (i) self-reported doctor diagnosis of diabetes or treatment with hypoglycemic drugs; (ii) fasting plasma glucose (FPG) of ≥7.0 mmol/L; (iii) 2-h blood glucose after oral glucose tolerance test (OGTT) of ≥11.1 mmol/L; (iv) hemoglobin A1c (HbA1c) of ≥6.5%; (v) any one of three random blood glucose test results ≥11.1 mmol/L [34]. CKD was diagnosed if participants met any of the following criteria: (i) estimated glomerular filtration rate (eGFR) of <60 mL/min/1.73m 2 ; (ii) albumin-to-creatinine ratio (ACR) of >30 mg/g [35]. Hypertension was considered present if participants met either of the following criteria: (i) self-reported physiciandiagnosed hypertension or treatment with anti-hypertensive medication; (ii) average of three systolic blood pressures (SBP) of ≥140 mmHg or average of three diastolic blood pressures (DBP) of ≥90 mmHg. Covariate data for cancer and history of fracture were obtained from the respective questionnaires administered to the study participants.

Statistical Analysis
All analyses used sampling weights recommended by the NCHS to account for the complex NHANES survey design. Initially, categorical variables were described by the frequency (percentage) of participants, and the differences between groups were compared using the chi-square test. Secondly, we measured the associations between the three plantbased diet indexes and the BMD T score, using Spearman's correlation coefficients, and induced the corresponding 95% confidence intervals. Correlation coefficients were classified into five categories: very strong (0.90-1.00), strong (0.70-0.89), moderate (0.40-0.69), weak (0.10-0.39), and negligible (0-0.10) [36]. A significance test was necessary to control for the possibility that an observed difference between two correlations may be due to chance alone. Overlapping correlations in dependent groups were compared using Hittner, May, and Silver's modification of Dunn and Clark's z test and Zou's confidence interval test [37]. In addition, multinomial logistic regression analysis was applied to examine the relationship between the three plant-based diet indexes and different BMD statuses. We developed two separate models for the association between each plant-based diet index and different BMD status: (1) Model 1: adjusted for age, sex, and ethnicity; (2) Model 2: Model 1 plus education, marital status, PIR, BMI, smoking status, physical exercise, hypertension, T2DM, CKD, cancer, and history of fracture. In the fully adjusted model (Model 2), we also explored the independent relationships of 15 individual food items with different BMD statuses. We further conducted sensitivity analyses to evaluate the robustness of our findings. First, subgroup analyses were performed for variables associated with different BMD statuses, with stratification factors including age (20-50, 50-65, ≥65), sex (male, female), ethnicity (non-Hispanic black, non-Hispanic white, Mexican American, other), T2DM (yes, no), CKD (yes, no), history of fracture (yes, no), and smoking status (non-smoker, current smoker). Second, we excluded participants who had previously taken anti-osteoporotic and estrogenic drugs, as the use of these medications may affect the accuracy of the results. Then, we additionally adjusted the models for more potential confounders, including menopausal status, corticosteroid usage, and dietary supplements (vitamin D and calcium). Next, by calculating E values we evaluated the possibility of unmeasured confounding between the three plant-based dietary indexes and bone loss [38,39]. The E value estimates the required magnitude of an unmeasured confounding that could negate the observed association between the three plant-based dietary indexes and different BMD statuses.
All statistical tests were two-tailed, and a statistically significant difference was defined as p < 0.05. All analyses were performed using R software (4.1.0, R core team).

Characteristics of Participants
A total of 16,085 participants were included in this study, among whom 8238 (51.22%) were female, and 4631 (28.79%) were over 65 years of age, regardless of gender. The characteristics of the study population are shown in Table 1, categorized for the groups with normal BMD, osteopenia, and osteoporosis. The three groups significantly differed in age, sex, ethnicity, education, marital status, PIR, BMI, smoking status, physical exercise, hypertension, T2DM, CKD, cancer, history of fracture, hPDI, PDI, and uPDI (p < 0.05).

Correlation between Plant-Based Diet Indexes and BMD T-Score
In the current study, we employed correlation analysis to assess the strength of the association between plant-based dietary indexes and BMD T scores. The Spearman correlation coefficients of the three plant-based dietary indexes and BMD T scores are illustrated in Figure 2 and Supplementary Table S2. hPDI was negatively and weakly correlated with BMD T score (r = −0.17; p < 0.001), while PDI (r = −0.09) and uPDI (r = 0.03) each had a negligible correlation with BMD T score. Furthermore, we conducted a comparative analysis of these associations and observed a significantly stronger correlation between hPDI and BMD T score in comparison with the other two associations (r (T score, hPDI) vs. r (T score, PDI): p-value < 0.001 (Hittner2003); 95% CI: −0.10; −0.06 (Zou2007); r (T score, hPDI) vs. r (T score, uPDI): p-value < 0.011 (Hittner2003); 95% CI: −0.23; −0.17 (Zou2007)). See Supplementary Table S3

Correlation between Plant-Based Diet Indexes and BMD T-Score
In the current study, we employed correlation analysis to assess the strength of the association between plant-based dietary indexes and BMD T scores. The Spearman correlation coefficients of the three plant-based dietary indexes and BMD T scores are illustrated in Figure 2 and Supplementary Table S2. hPDI was negatively and weakly correlated with BMD T score (r = −0.17; p < 0.001), while PDI (r = −0.09) and uPDI (r = 0.03) each had a negligible correlation with BMD T score. Furthermore, we conducted a comparative analysis of these associations and observed a significantly stronger correlation between hPDI and BMD T score in comparison with the other two associations (r (T score, hPDI) vs. r (T score, PDI): p-value < 0.001 (Hittner2003); 95% CI: −0.10; −0.06 (Zou2007); r (T score, hPDI) vs. r (T score, uPDI): p-value < 0.011 (Hittner2003); 95% CI: −0.23; −0.17 (Zou2007)). See Supplementary Table S3 for the detailed results of the correlation comparisons.

Associations between Plant-Based Diet Indexes and Different BMD Status Groups
We considered the association between three plant-based diet indexes and different BMD status groups, using multinomial logistic regression. Table 2 presents the logistic regression results (OR with 95% CI) and reports the results of a linear trend test (P for trend) to examine whether there was a linear trend in the association between bone loss and variations in three plant-based diet indexes. In the fully adjusted model, participants in the highest quintile for both hPDI (OR Q5 vs. Q1 = 1.50; 95% CI: 1.24-1.81) and PDI (OR Q5 vs. Q1 = 1.22; 95% CI: 1.03-1.45) had a positive association with osteopenia compared with participants in the lowest quintile, while the highest uPDI (OR Q5 vs. Q1 = 1.48; 95% CI: 1.04-2.11) was positively associated with osteoporosis. Furthermore, we found a positive association between a 10-unit increment in hPDI (OR per 10-unit increment = 1.17; 95% CI: 1.08-1.27) and osteopenia, while a 10-unit increment in uPDI (OR per 10-unit increment = 1.29; 95% CI: 1.08-1.54) was positively associated with osteoporosis. We further compared the differences in the association between the three plant-based dietary indexes and different BMD statuses in the femoral neck and lumbar spine (

Associations between 15 Individual Food Items and Different BMD Status Groups
We analyzed the independent associations of 15 individual food items with different BMD statuses.

Sensitivity Analysis
Multiple sensitivity analyses were conducted to assess the robustness of our results. Firstly, subgroup analysis suggested a consistent association between plant-based dietary indexes and osteopenia/osteoporosis across most subgroups. The positive association between osteopenia and a 10-unit increment in hPDI was observed in strata defined by age (20-50 and 50-65), sex (male and female), ethnicity (non-Hispanic white), T2DM (no), CKD (no), fracture history (no), and smoking status (non-smoker) (Figure 3a). Also, we found a positive association between a 10-unit increment in uPDI and osteoporosis manifested in participants with the following characteristics: age from 50 to 65 years, female, non-Hispanic white, non-diabetes, fracture history, and current smoker (Figure 3b). Further detailed results are listed in Figure 3a defined by age (20-50 and 50-65), sex (male and female), ethnicity (non-Hispanic white), T2DM (no), CKD (no), fracture history (no), and smoking status (non-smoker) ( Figure  3a). Also, we found a positive association between a 10-unit increment in uPDI and osteoporosis manifested in participants with the following characteristics: age from 50 to 65 years, female, non-Hispanic white, non-diabetes, fracture history, and current smoker (Figure 3b). Further detailed results are listed in Figure 3a,b. Figure 3. (a) ORs and 95% CI for osteopenia per 10-unit increment in adherence to plant-based diet indexes, stratified by selected characteristics; (b) ORs and 95% CI for osteoporosis per 10-unit increment in adherence to plant-based diet indexes, stratified by selected characteristics. All models were multivariable adjusted for age, sex, and ethnicity, education, marital status, PIR, BMI, smoking status, physical exercise, hypertension, T2DM, CKD, cancer, and history of fracture. In each stratified analysis, the stratification variable was excluded in the adjustments. The red dots presented hPDI; the blue dots presented PDI; the green dots presented uPDI. Figure 3. (a) ORs and 95% CI for osteopenia per 10-unit increment in adherence to plant-based diet indexes, stratified by selected characteristics; (b) ORs and 95% CI for osteoporosis per 10-unit increment in adherence to plant-based diet indexes, stratified by selected characteristics. All models were multivariable adjusted for age, sex, and ethnicity, education, marital status, PIR, BMI, smoking status, physical exercise, hypertension, T2DM, CKD, cancer, and history of fracture. In each stratified analysis, the stratification variable was excluded in the adjustments. The red dots presented hPDI; the blue dots presented PDI; the green dots presented uPDI.
Secondly, the results remained robust and significant after excluding participants with a history of taking anti-osteoporosis and estrogenic drugs. Further detailed results are listed in Supplementary Table S4.
Moreover, after further adjustment for three potential confounders (menopausal status, corticosteroid usage, and dietary supplements), respectively, the results remained robust and significant (Supplementary Tables S5-S7).
Finally, we calculated E values to evaluate the sensitivity of our findings to potential unmeasured confounding factors. As shown in Supplementary Table S8, the E values indicated that our results were robust unless there was an unmeasured confounding factor with a relative risk greater than the E values that were associated with the three plant-based dietary indexes.

Discussion
This study assessed the association between three plant-based dietary indexes and osteopenia/osteoporosis, based on a cross-sectional study of the adult population in the United States. According to the correlation coefficient analysis, a close association was observed between hPDI and BMD T score. Moreover, we observed a harmful association between PDI and hPDI with osteopenia, and a harmful association between uPDI and osteoporosis. Meanwhile, higher hPDI and PDI were significantly correlated with higher risk of osteopenia at the lumbar spine but not at the femoral neck. Among individual food items, vegetables, refined grains, animal fats, eggs, and meat were the main protective contributors, while nuts were associated with higher odds of osteopenia.
Several previous studies have indicated that a plant-based diet may have a negative impact on bone health, which is consistent with our findings. For instance, a Chinese cohort study reported significant differences in hPDI between three groups (normal, osteopenia, and osteoporosis), with the normal group having a significantly lower hPDI than the osteopenia and osteoporosis groups [27]. In an Iranian case-control study of postmenopausal women with osteoporosis, a higher uPDI was associated with an increased risk of bone loss in the femoral neck (OR: 2.63; 95% CI: 1.37-5.06) and lumbar spine (OR: 4.23; 95% CI: 2. 19-8.19) [26]. Furthermore, several meta-analyses have compared the impact of plant-based and omnivorous diets on bone health [40,41]. Li et al. concluded that individuals following a plant-based diet had significantly lower BMD at the lumbar spine, femoral neck, and whole body compared to those following an omnivorous diet, with a reduction of approximately 3-4% [40]. Another meta-analysis found that vegans had 6% lower BMD compared with omnivores [41]. Taken together, our findings provide further evidence supporting the potential harmful association between bone health and a plant-based diet.
Several potential mechanisms have been proposed to explain the association between a plant-based diet and an increased risk of bone loss. A clinical trial reported that a plantbased diet reduced calcium and vitamin D intake and increased N-tetrapeptide biomarkers, indicating increased bone resorption [42]. Furthermore, a cross-sectional study of Iranian older adults showed that higher uPDI was significantly associated with decreased levels of osteocalcin, a protein that plays an critical role in regulating bone formation and remodeling by modulating bone mineralization and the activity of osteoblasts and osteoclasts [8,43]. Moreover, individuals following a plant-based diet typically have lower dietary calcium and protein intake than omnivores [44,45]. Adequate protein intake is crucial for the formation and maintenance of the bone matrix and influences the secretion of insulin-like growth factor I (IGF-I), an orthotropic hormone that promotes calcium and phosphorus absorption in the gut and synthesizes calcitriol [16]. However, the relationship between protein and dietary calcium is complex and interdependent [41]. While protein has positive effects on bone health, adequate dietary calcium intake is also necessary. Inadequate calcium levels may lead to adverse affects of protein on bone density [16]. Therefore, individuals following a plant-based diet require additional sources of protein and dietary calcium. For individual foods, we found beneficial associations between bone health and vegetables, eggs, and meat, which are high-quality dietary calcium and protein sources. We recommend that individuals following a plant-based diet carefully select appropriate foods or supplements to enhance their calcium and protein intake, ensuring adequate intake to avoid potential nutrient shortfalls and to help maintain a healthy bone balance.
It is worth noting that our study revealed that an overall plant-based diet and a healthy plant-based diet affect osteopenia rather than osteoporosis. Bone metabolism and BMD are sensitive to subtle changes in nutrient intake and acid-base balance [46]. While several studies and potential mechanisms described above suggest that a plant-based diet can negatively affect bone health, it is crucial to note that plant-based diets also possess potential benefits for bone health. Higher PDI and hPDI, compared with uPDI, may reflect higher intake of whole grains, vegetables, fruits, nuts, legumes, tea, and coffee. These healthy plant foods typically contain higher levels of minerals (magnesium, potassium) and vitamins, as well as antioxidant and anti-inflammatory phytochemicals. Consequently, these foods result in a lower acid load on the body, which is beneficial for bone health [47]. For instance, fresh fruits and vegetables are rich in potassium, which can reduce the acid load to maintain calcium levels in bones [26]. The positive effects of a plant-based diet on bone health may counteract some of the negative effects and thus attenuate the negative impact on BMD, which may potentially explain our findings. Therefore, future public health recommendations should focus on both the quantity and quality of plant-based foods in order to maintain optimal bone health.
The prevalence of osteoporosis varies widely within the body, with the lumbar spine and femoral neck being the most common sites for diagnosing osteoporosis [48]. The findings of the present study indicate a significant association between plant-based dietary indexes (hPDI and PDI) and osteopenia at the lumbar spine, whereas no significant association was observed at the femoral neck. There are several potential explanations for these results. Firstly, the lumbar spine and femoral neck have different rates of bone loss, with trabecular bone (in the lumbar spine) having a more rapid rate of bone loss than cortical bone (in the femur) [49,50]. Secondly, most etiologies of secondary osteoporosis, such as malabsorption, liver disease, and rheumatoid arthritis, affect the spine rather than the femur [2,51]. Moreover, weight bearing, particularly in the hip and femur, is known to increase BMD and is a cause of physiological variation [2].
The subgroup analysis findings suggest that certain groups may experience more pronounced effects under plant-based dietary patterns, such as non-Hispanic white populations and relatively young individuals. To provide a comprehensive explanation for the observed variations in bone health across ethnicities, it is imperative to consider the influence of genetic factors on BMD. Genomic studies indicate that disparities in BMD across ethnicities may primarily be attributed to genetic dissimilarities [52]. Earlier investigations established that the incidence of osteoporosis significantly differs between ethnic groups, with more than 60.0% of the variance in BMD being attributed to genetic factors [53]. Additionally, genetic factors affect differences in body composition, such as greater cortical thickness and higher trabecular BMD in blacks, suggest inherent differences in bone strength according to ethnicity [54]. Several potential explanations could be considered regarding the observed differences in the effects of plant-based diets on bone health among distinct demographic strata defined by age. For example, younger individuals may have higher rates of bone turnover and may require more nutrients, such as calcium and vitamin D, to maintain bone health [55]. Additionally, older individuals may be at higher risk of osteoporosis and may require more targeted interventions to maintain bone health [2]. However, it is important to note that the complex interplay of these factors with ethnicity and age is not clear, and further studies are needed to fully elucidate their underlying impact on bone health.
To the best of our knowledge, there remains relatively scarce research assessing the association between plant-based dietary indexes and bone loss, especially in the US population. Our study offers timely and unique evidence with important implications. Given the increasing recommendation of plant-based diets for environmental protection and prevention of chronic disease, it is necessary to investigate their relationship with bone health. Our findings may contribute to future nutritional policy development and promote systemic and synergistic shifts toward a healthier food system. The present study has several notable strengths. First, the relatively large sample size in NHANES and the use of population-based random cluster selection ensured the representativeness of the sample and its applicability to the entire U.S. population. Second, the design of the subgroup and sensitivity analyses enabled a more in-depth exploration of the association between plant-based dietary indexes and bone health, strengthening the reliability of the study results. Taken together, these strengths contribute to the validity and generalizability of our findings. However, there are several limitations that need to be cautiously considered when interpreting our findings. First, the present study employed a cross-sectional design, precluding us from establishing a causal relationship between plant-based dietary indexes and osteopenia/osteoporosis. Second, relying on 24-h dietary recalls to collect the dietary data may not accurately reflect participants' usual dietary intake. Finally, since the environment surveyed by NHANES is restricted to the USA, the generalizability of the findings to other dietary cultures may be limited, and further research remains necessary.

Conclusions
In conclusion, our findings provide evidence that adherence to a plant-based dietary pattern is associated with decreased BMD in a nationally representative population of US adults, highlighting the importance of a balanced diet for maintaining bone health, especially including foods rich in dietary calcium and protein such as vegetables, eggs, and meat. Meanwhile, a negative association was revealed between two plant-based dietary indexes (hPDI and PDI) and osteopenia, which was more significantly at the lumbar spine rather than the femoral neck. Among 15 individual food items, vegetables, refined grains, animal fat, eggs, and meat were the main protective contributors, whereas nuts were associated with increased odds of osteopenia. From a clinical perspective, dietary interventions rather than medications may be more effective in improving bone health and preventing fractures. Individuals following a plant-based diet should carefully plan their nutritional intake and monitor their bone health regularly. Moreover, further research is needed to explore the causality and generalizability of our findings, as well as to investigate the potential benefits and risks of specific types of plant-based diets in terms of their effects on bone health.  Table S8: E values for the effects of plant-based diets on bone loss (lower limit of 95% CI) in the fully adjusted model.
Author Contributions: K.X. and Y.Z. contributed to the conception of the study; J.W. and Y.W. contributed significantly to analysis and manuscript preparation; Y.Z. analyzed and interpreted the data; Y.Z. prepared the initial manuscript draft with input from K.X. and X.C; K.X. and X.C. supervised the revision of the manuscript. K.X. and X.C. are corresponding authors. All authors have read and agreed to the published version of the manuscript. Institutional Review Board Statement: Ethical approval and review were not sought or obtained for this study given that it involved secondary analysis of publicly available, de-identified data.

Informed Consent Statement:
The studies involving human participants were reviewed and approved by the ethics review board of the National Center for Health Statistics. The patients/participants provided their written informed consent to participate in this study.

Data Availability Statement:
Publicly available datasets were analyzed in this study. This data can be found here: https://www.cdc.gov/nchs/nhanes/. It was accessed on 6 March 2023.