Data from NHANES 2017-2018 were used for this study. NHANES is an annual cross-sectional survey conducted by the National Center for Health Statistics (NCHS) of the Centers for Disease Control and Prevention (CDC) (23). The data were collected using a complex multi-stage probabilistic sampling design and a complex weighting scheme that over-sampled certain ethnic and age groups and ultimately provided a representative sample of the non-institutionalized civilian population of the United States. 3,069 adults (over 50 years) participated in the NHANES 2017-2018 survey, and we excluded those who did not have laboratory testing at a mobile examination center (N=171). In addition, we also excluded those with missing data on Dual-Energy X-ray Absorptiometry (DXA) (N=612) or Vibration controlled transient elastography (VCTE) N=255). Therefore, the final study sample enrolled 2031 adults with complete data ( Figure 1 ). Because all participants had signed written informed consent and their personal information was fully de-identified, our study was granted an exemption from the Institutional Review Board.

We collected demographic, anthropometric, related laboratory, and self-reported questionnaire data. Demographic variables, including age, gender (Male, Female), and race/ethnicity (Mexican American, Other Hispanic, Non-Hispanic White, Non-Hispanic Black, and Other Race), educational level (Below high school, High school and above), marital status (Married/Living with partner, Widowed/Divorced/Separated/Never married) were ascertained by questionnaire. Anthropometric indicators include body mass index (BMI) and waist circumference (WC). The indexes measured included glycated hemoglobin A1c (HbA1c), glucose, insulin, triglyceride, total cholesterol, high-density lipoprotein cholesterol (HDL-C), alanine aminotransferase (ALT), aspartate aminotransferase (AST), alkaline phosphatase (ALP), gamma glutamyl transferase (GGT), total bilirubin, platelet count, creatinine, total serum calcium, serum phosphorus, albumin, uric acid, high sensitivity c-reactive protein (HS-CRP), as previously described (24). Blood pressure is measured by a blood pressure inspector who has passed an accredited training program. Participants were measured three times after sitting for 5 minutes and to determine the maximum dilation level (MIL). We took the average of the three measurements as the representative blood pressure of the participant. Hypertension was defined by the presence of one of two criteria (25): systolic blood pressure ≥140 mmHg and/or diastolic blood pressure ≥90 mmHg or currently taking antihypertensive medications. Diabetes mellitus was defined by the presence of one of the following conditions: a) self-reported diabetes; b) fasting blood glucose ≥126mg/dL; c) HBA1c level ≥6.5%; d) use of anti-diabetic drug, including insulin (26). Homoeostasis model assessment of insulin resistance (HOMA-IR) was conducted to estimate insulin resistance using the equation by Matthews et al. (27).

We categorized drinking status as drinker or never drinker, according to the following questions “Ever had a drink of any kind of alcohol?”. Besides, we categorized smoking status as smokers or never smokers, according to the following questions “Smoked at least 100 cigarettes in life?”.

Although liver biopsy is recommended as the gold standard for diagnosing fatty liver disease, it may not be appropriate in large population studies due to its invasive nature. VCTE is a noninvasive, convenient, and widely available technique that uses controlled attenuation parameter (CAP) and liver stiffness measurement (LSM) to define fatty liver and fibrosis (28). In the NHANES 2017-2018 cycle, health technicians completed a 2-day training program with survey staff and an expert FibroScan^® Technician (reference examiner). Training will be provided by NCHS personnel, Westat, and an Echosens expert trainer. The training included an overview of the component, demonstrations conducted by the reference examiner with volunteer subjects. The reference examiner reviewed and demonstrated the proper technique of the FibroScan^® examination. The elastography measurements were obtained in the NHANES Mobile Examination Center (MEC), using the FibroScan^® model 502 V2 Touch equipped with a medium (M) or extra-large (XL) wand (probe). Supervised practice exercises followed, conducted with several volunteer subjects. The reference examiner would certify the health technician after observing 3 satisfactory exams. The inter-rater reliability (health technician compared with reference examiners, n=32) was 0.86 for stiffness, and 0.94 for CAP-steatosis. CAP ≥274 dB/m indicated liver steatosis (S ≥1), as this cut-off point has 90% sensitivity in identifying any degree of hepatic steatosis (29). The cut-off for liver fibrosis (LF) were obtained from previous study (29). Significant LF, severe LF and cirrhosis were defined as LSM ≥ 8.2 KPa (fibrosis grade ≥ F2), LSM ≥ 9.7 KPa (fibrosis grade ≥ F3) and LSM ≥ 13.6 KPa (fibrosis grade ≥ F4), respectively.

According to the 2020 international expert consensus statement (7), a positive diagnosis of MAFLD is based on VCTE evidence of fat accumulation in the liver (hepatic steatosis) in addition to one of the following three criteria, namely overweight/obesity, presence of T2DM mellitus (T2DM), or evidence of metabolic dysregulation. Metabolic dysregulation exhibits the following characteristics: 1) waist circumference ≥102/88 cm in Caucasian men and women; 2) blood pressure ≥130/85 mmHg or specific drug treatment; 3) plasma triglycerides ≥150 mg/dl or specific drug treatment; 4) plasma HDL-cholesterol <40 mg/dl for men and <50 mg/dl for women or specific drug treatment; 5) prediabetes (fasting glucose levels 100 to 125 mg/dl, or HbA1c 5.7% to 6.4%; 6) HOMA-IR score ≥2.5; 7) plasma high-sensitivity C-reactive protein level >2 mg/L.

BMD was measured in the left hip bone using a Dual-Energy x-ray absorptiometry (DXA) scan with a fan-beam densitometer (QDR4500A; HOGICF). Pregnant women and the participants less than 50 years old or with self-reported history of radiographic contrast material in the past week or weighted over 450 pounds were ineligible for the DXA scan. Besides, participants with hip fracture, replacements or pins in both hips were also excluded. Further details of the DXA examination protocol were documented in the Body Composition Procedures Manual located on the NHANES website. According to the guidelines of the World Health Organization (WHO), the diagnostic criteria of osteoporosis or osteopenia is based on T-score results. T-scores were calculated as (BMD measured − mean BMD reference)/SD reference. Osteoporosis was defined as a T-score of BMD ≤ − 2.5, and osteopenia was defined as − 2.5 < T-score ≤ − 1. We used non-Hispanic white women between the ages of 20-29 in NHANES III as a reference group for calculating femoral BMD scores (30). The lumbar BMD T -score references for calculating were extracted from previous study (31). A T-score ≤ − 2.5 in any part of the femur, femoral neck, trochanter, intertrochanter and lumbar was diagnosed as osteoporosis.

Considering the complex sample design adopted by NHANES, we utilized weights in the statistical analysis to ensure that the data were representative of the entire United States non-institutionalized population. The weights were applied in accordance with the guidelines for using raw data from NHANES provided by the NCHS. Continuous variable was presented as weighted mean ± SE and classified variable as weighted percentage. The Chi-square test was performed for categorical variables and Kruskal-Wallis test for continuous variables to describe the clinical characteristics of patients with or without MAFLD. Weighted multiple linear regression analyses were conducted to analyze the association between MAFLD and BMD after adjusting for known or selected confounders. Confounders were selected for their significant association with the BMD, or a changed in effect estimate of more than 10%. To avoid multicollinearity, variance inflation factors were assessed before adjustment. The subgroup analyses were stratified by sex (male and female), age (<60, 60 to 70, and ≥70 years), abdominal obesity (yes or no), hyper-hs CRP (yes or no) and Low HDL (yes or no).

Given that the influence of fatty liver or liver fibrosis on femoral or lumbar BMD may be inconsistent. Hence, in sensitivity analysis, we excluded those without lumbar bone mineral density data and explored the relationship between MAFLD and liver fibrosis and whole osteoporosis. All analyses were performed in R software (version 3.6.1). All tests were two-tailed and P values <0.05 were considered statistically significant.

A total of 2031 participants (1070 males and 961 females) were included in the study. Table 1 summarizes the baseline characteristics of participants with and without MAFLD. Based on diagnostic criteria for MAFLD, 446 men and 400 women were diagnosed, with a weighted prevalence of 50.9% (95% CI, 45.3%-56.3%) and 40.7% (95% CI, 36.4%-45.1%), respectively. Compared with participants without MAFLD, those with MAFLD were more likely to be man, IR, hypertensive, diabetic, pre-diabetic, overweight and abdominal obesity. In addition, these MAFLD participants were noted to have a higher body mass index, alanine aminotransferase, alkaline phosphatase, gamma glutamyl transferase, fasting plasma glucose, fasting insulin, uric acid, HbA1c, hs-CRP, triglyceride, larger waist circumference, waist circumference and lower total cholesterol, HDL-C. As we predicted, participants diagnosed with MAFLD had a lower weighted prevalence of osteoporosis than those without MAFLD, although this outcome may be affected by BMI. A lower rate of osteoporosis observed in women with MAFLD did not reach significance.

We performed a multiple linear regression analysis to examine the association between MAFLD and femoral BMD. The regression coefficients (β) and 95% confidence intervals (CIs) were estimated in three models. Higher femoral BMD, including the BMD of the femur, femoral neck, intertrochanter, and trochanter, were found in the patients with MAFLD, even after fully adjusting for confounders ( Table 2 ). Nonetheless, when BMI was adjusted in the multiple liner regression analyses, the association between MAFLD femoral BMD also became insignificant. To further examine the association between MAFLD and femoral osteoporosis, we carried out the multiple logistic regression analysis to explore the association between MAFLD related variables and femoral osteoporosis. In the adjusted II model, apart from diabetes, prediabetes, hypertension, hypertriglyceridemia and insulin resistance, all association was observed between the MAFLD related variables and femoral osteoporosis ( Table 3 ). As shown in Table 4 , MAFLD was negatively associated with femoral osteoporosis when BMD was adjusted (OR=0.89, 95% CI: 0.52 to 1.53), although the relationship became insignificant.

Subsequently, subgroup analysis was performed to further understand the association between MAFLD and femoral osteoporosis ( Table 4 ). In subgroup analyses stratified by sex, a significantly negative association between MAFLD and femoral osteoporosis was observed in both men and women. When stratified by age (<60, 60 to 70, and ≥70 years), abdominal obesity (yes or no), hyper-hs CRP (yes or no) and Low HDL (yes or no), we could still observe a negative association between MAFLD and femoral osteoporosis, although the correlation becomes tenuous. Furthermore, in subgroup analyses stratified by age, we found that the protective effect of the higher age group on osteoporosis was decreased compared with those younger than 60 years, which is consistent with a previous research suggesting that aging is a risk factor for osteoporosis. Nonetheless, when BMI was adjusted in the subgroup analyses, the association between MAFLD and femoral osteoporosis became insignificant.

We explored the relationship between different grades of liver stiffness and femoral BMD in the multiple linear regression. We found that in the crude, adjust I and adjust II model, patients with significant liver stiffness (fibrosis grade ≥ F2) had higher femoral BMD levels. Besides, this result was still observed in patients with severe liver stiffness (fibrosis grade ≥ F3) or cirrhosis (fibrosis grade ≥ F4). Nonetheless, when BMI was further adjusted in the adjust III model, the relationship between significant liver stiffness, severe liver stiffness, cirrhosis and femoral osteoporosis became insignificant (showed in the Supplementary Table 1 ).

In the sensitivity analysis, we did the following things. First, we explored the relationship between MAFLD and whole osteoporosis (include femoral and lumbar osteoporosis). As showed in the Supplementary Table 2 , those patients with MAFLD had a higher lumbar BMD,when adjusted the confounders but not included BMI, compare with their counterparts. Nonetheless, the lumbar BMD of patients with MAFLD was parallel with those without MAFLD, when adjusted BMI. Besides, similar to the relationship between MAFLD and femoral osteoporosis, the relationship between MAFLD and whole osteoporosis was showed in adjust II model that patients with MAFLD had a lower risk ratio for whole osteoporosis compared to those without MAFLD (OR=0.46, P value=0.0022), and then once BMI was adjusted, the relationship became insignificant (OR=0.81, P value=0.4539). (showed in the Supplementary Table 3 ).

Second, the relationship between liver stiffness and whole osteoporosis was showed in the Supplementary Table 4 . There was no association between liver fibrosis and whole osteoporosis, regardless of adjusting BMI. These results may suggest that patients with MAFLD and liver fibrosis have a higher BMD and a lower risk ratio for osteoporosis, but may be confounded by BMI.