Liver fibrosis is associated with impaired bone mineralization and microstructure in obese individuals with non-alcoholic fatty liver disease

Chronic liver diseases are associated with increased bone fracture risk, mostly in end-stage disease and cirrhosis; besides, data in non-alcoholic fatty liver disease (NAFLD) are limited. Aim of this study was to investigate bone mineralization and microstructure in obese individuals with NAFLD in relation to the estimated liver fibrosis. For this cross-sectional investigation, we analyzed data from 1872 obese individuals (44.6 ± 14.1 years, M/F: 389/1483; BMI: 38.3 ± 5.3 kg/m2) referring to the Endocrinology outpatient clinics of Sapienza University, Rome, Italy. Participants underwent clinical work-up, Dual-Energy X-ray Absorptiometry for assessing bone mineral density (BMD) and microarchitecture (trabecular bone score, TBS). Liver fibrosis was estimated by Fibrosis Score 4 (FIB-4). Serum parathyroid hormone (PTH), 25(OH) vitamin D, osteocalcin and IGF-1 levels were measured. Obese individuals with osteopenia/osteoporosis had greater FIB-4 than those with normal BMD (p < 0.001). FIB-4 progressively increased in presence of degraded bone microarchitecture (p < 0.001) and negatively correlated with the serum osteocalcin (p < 0.001) and IGF-1 (p < 0.001), which were both reduced in presence of osteopenia/osteoporosis. FIB-4 predicted IGF-1 reduction in multivariable regression models adjusted for confounders (β: − 0.18, p < 0.001). Higher FIB-4 predicted bone fragility with OR 3.8 (95%C.I:1.5–9.3); this association persisted significant after adjustment for sex, age, BMI, diabetes, smoking status and PTH at the multivariable logistic regression analysis (OR 1.91 (95%C.I:1.15–3.17), p < 0.01), with AUROC = 0.842 (95%C.I:0.795–0.890; p < 0.001). Our data indicate the presence of a tight relation between NAFLD-related liver fibrosis, lower bone mineral density and degraded microarchitecture in obese individuals, suggesting potential common pathways underlying liver and bone involvement in obesity and insulin resistance-associated disorders.


Introduction
Non-alcoholic fatty liver disease (NAFLD) is the most common chronic liver disorder worldwide and is characterized by abnormal intrahepatic fat accumulation, it can be accompanied by different degrees of inflammation and fibrosis and put its basis on a substratum of insulin resistance [1]. For this reason, recent reclassification and alternative definition for NAFLD have been proposed, introducing the expression of "metabolic (dysfunction) associated fatty liver disease" (MAFLD), to underlie the tight relationship between abnormal fat liver content and presence of metabolic disease [2].
Once hepatosteatosis occurs, the degree of insulin resistance worsens. The existence of a bidirectional association between the presence of fatty liver disease and abnormal glucose metabolism (AGM) is well established, and so the higher incidence and severity of diabetes' complications in the course of NAFLD [3]. Nonetheless, NAFLD is considered as an independent determinant of all-cause and cardiovascular mortality independent of traditional risk factors [4].
Among histological features of NAFLD, the presence of fibrosis represents the major driver of increased mortality in affected individuals, regardless of the entity of lipid infiltration into the hepatocytes [4].
Besides influencing metabolic outcomes, the presence of NAFLD may negatively impact on other organs and systems, as the bone. Increased bone fragility and fracture risk have been reported in chronic liver diseases, i.e. viral-induced chronic hepatopathies, mostly in presence of advanced disease's stage and cirrhosis, when reduced synthesis of specific liver-derived growth factors, such as the insulin-like growth factor 1 (IGF-1), may negatively impact on bone metabolism [5]. Some evidence also reported reduced IGF-1 levels in both adults [6] and children [7] with NAFLD.
A certain increased of fracture risk has been also reported in NAFLD; however, studies on this regard have contrasting results [8,9], and this association is particularly debated in women [10].
Notably, NAFLD is associated with overweight and obesity, where the presence of elevated body mass index and fat content interferes with a correct estimation of bone mineralization at the dual-energy X-ray absorptiometry (DXA) scan, the gold standard technique for osteoporosis clinical assessment. Besides, increase bone fragility is described in both type 1 and type 2 diabetes mellitus (T2DM) [11]: in individuals with diabetes, qualitative impairment of bone structure and degraded microarchitecture lead to increased fracture risk in presence of normal to high bone mineral density (BMD) [12]. Thus, although in dysmetabolic individuals, the evaluation of bone structure quality more than mineral density may unravel increased fragility and fracture risk, data on this regard are scarce and not available in NAFLD populations. Similarly, evidence on the association between liver fibrosis and bone mineralization in obese individuals is limited. Therefore, this study was designed to investigate the relationship between qualitative and quantitative impairment of bone mineralization and NAFLD in individuals with obesity and different degrees of glucose tolerance and to explore potential correlates of altered bone metabolism in these cohorts.
Eligible for this study were individuals who met the following inclusion criteria: male or female subjects aged between 18 and 65 years; presence of obesity, as defined by having a body mass index (BMI) > 29.9 kg/m 2 , no history of excessive alcohol drinking (considered as an average daily consumption of alcohol > 30 gr/day in men and > 20 gr/day in women); negative tests for the presence of hepatitis B surface antigen and antibody to hepatitis C virus; no cirrhosis and other causes of liver diseases (hemochromatosis, autoimmune hepatitis, Wilson's disease); no treatment with drugs potentially inducing hepatic fat accumulation (e.g., corticosteroids, estrogens, methotrexate, tetracycline, amiodarone and/or calcium channel blockers).
All study participants underwent physical examination at the study site; weight and height were measured light clothes on and shoes off, then the BMI was calculated. Waist circumference (cm) was measured midway between the 12th rib and the iliac crest. Systemic systolic (SBP, mmHg) and diastolic (DBP, mmHg) blood pressure were measured after 5 min resting; the average of the second and third measurements was calculated and recorded to be used in the analyses.
Individuals without an established diagnosis of T2DM underwent standard oral glucose tolerance test as for the American Diabetes Association criteria [13], blood glucose and insulin area under the curves (gAUC, iAUC) were calculated. Participants with impaired fasting glucose and/or impaired glucose tolerance or overt T2DM were indicated as patients with abnormal glucose metabolism (AGM) [13].
Study population underwent medical history collection and clinical work-up. Fasting blood samples were drawn for clinical biochemistry and experimental purposes.
As indicators of insulin resistance and sensitivity, the Homeostasis model assessment for insulin resistance

NAFLD and liver fibrosis assessment
In the entire study population, the presence and degree of NAFLD were estimated by clinical history and calculating the NAFLD Liver Fat Score  [15], a proton magnetic resonance spectroscopy-validated score which has been demonstrated to best predict NAFLD in high risk populations (estimate the liver fat contents > 5.56%, with an AUROC of 0.88 [15] [16]. Finally, as a noninvasive estimator of liver fibrosis, we calculated the FIB-4 score [17] and a value above 1.3 was used as cut-off for predicting liver fibrosis [17,18].

Bone evaluations
Study participants underwent BMD assessment via Dual-Energy X-ray Absorptiometry (DXA, Hologic Inc, Bedford, MA, USA, QDR 4500 W, S/N 47,168) at the lumbar spine (L1-L4 anteroposterior) and hip (left femoral neck), performed by trained technicians using standardized procedures and calibration schedule recommended by the producer. Specific delimiters for regional analysis were determined by standard software (Hologic Inc, QDR 4500 W S/N 47,168 VER. 11.2). BMD was expressed in g/cm 2 . At each site, patients with a T score ≥ − 1 were included in the normal group, patients with a T score between < − 1 and − 2.5 were included in the osteopenia group, and patients with a T score < − 2.5 were included in the osteoporosis group, as defined by the World Health Organization.
Bone structure and the quality of microarchitecture were estimated by the Trabecular Bone Score using the integrated software TBS iNsight, version 2.1.2.0, starting from the site-matched spine DXA scans. A TBS greater than 1.350 was considered normal, TBS between 1.200 and 1.350 was indicative of partially degraded microarchitecture and TBS < 1.200 was considered as an expression of degraded bone microarchitecture, according to previous reports [19].

Statistics
SPSS software version 27.0 was used to perform all the statistical analyses. Continuous variables are reported as the mean ± standard deviation (SD) and categorical variables as percentages. Skewed variables (i.e. FIB-4, triglycerides, FBI, HOMA-IR) underwent natural logarithmic transformations before performing the analyses. Student's T test for continuous normally distributed variables, Mann-Whitney for non-normally distributed variables or χ 2 test for categorical variables were used to compare mean values between two independent groups; comparisons between more than two groups were performed by ANOVA test and Bonferroni-adjustment was applied when indicated. Correlations between parameters were explored by Pearson's and Spearman's coefficients, as appropriate. Multivariable regression models were built forcing for sex and age and entering variables significantly associated at the bivariate correlation analyses. Multivariate logistic models were used to investigate the association between FIB-4 and the presence of osteopenia/osteoporosis using the T score as a dependent dichotomous variable (categorized as above and below-1 SD) and FIB-4 (continuous value) as an independent variable, adjusting for potential confounders. Similarly, we tested the correlation between FIB-4 and the presence/entity of bone microarchitecture disarrangement by multivariable regression analysis using the TBS as a dependent continuous variable. A sensitivity analysis using the two-tailed Monte Carlo method was performed to confirm the association between the age-adjusted FIB-4 as a categorical variable (low vs. intermediate-high fibrosis risk) and the presence of osteopenia/osteoporosis vs. normal BMD at the DXA examination. Two-sided p-value < 0.05 was considered statistically significant, with a confidence interval of 95%.
This study was approved by the local Ethics Committee before study initiation and conducted in conformance with the Helsinki Declaration. Informed written consent was obtained from participants before any study procedure.

NAFLD and liver fibrosis risk in obese individuals
In our obese population, the prevalence of NAFLD ranged between 92 and 96% when estimated by FLI and LFS, respectively. 52% of study participants had a clinical diagnosis of MS, as identified by the presence of at least two components in addition to increased waist circumference, 36% had abnormal glucose metabolism and 13% T2DM. At the time of study recruitment, 46% study participants were treated with one or more classes of anti-hypertensive agents (renin-angiotensin-aldosterone system inhibitors: 37.4%, beta-blockers: 15.2%, calcium channel blockers: 9.8%, diuretics: 23.5%), 23% with metformin, 11% with statins and 10.4% with acetylsalicylic acid. No independent association was found between ongoing treatment and serum transaminases in multiple age-and sex-adjusted multivariate models (not shown).
The mean ± SD FIB-4 value in the study cohort was 0.77 ± 0.46 [median (range): 0.68 (0.20-4.9)]; using a FIB-4 cut-off above 1.3 for intermediate-high risk of liver fibrosis, consisting in estimated F score ≥ 3, we obtained a prevalence of 10.3% individuals (n = 192) at increased risk of hepatic fibrosis.
Since a different FIB-4 cut-off (FIB-4 > 2) has been proposed for individuals aged above 65 years to reduce the entity of false positive results [20], we repeated the analysis considering different cut-offs according to the age, and obtained an overall prevalence of 8% (n = 148) individuals at intermediate to high risk of advanced liver fibrosis (F ≥ 3) in the overall population. Characteristics of study participants belonging to the subgroup with high or low FIB-4 are presented in Table 1.
FIB-4 was significantly higher in men (mean ± SD FIB-4: 0.88 ± 0.56 vs 0.74 ± 0.43 in female participants, p < 0.001) and in individuals with T2DM compared to non-diabetic subjects (mean ± SD FIB-4: 1.02 ± 0.62 vs. 0.73 ± 0.42, p < 0.001). The association between greater FIB-4 and male sex was confirmed after adjustment for age and presence of impaired glucose regulation or T2DM at the partial correlation analysis (r = 0.14, p < 0.001). Similarly, the relationship between FIB-4 and T2DM was independent from sex and age in the partial correlation model, although the strength of this association was slightly reduced in this adjusted model (r = 0.07, p = 0.004). Among obese individuals, greater FIB-4, as considered as a continuous variable, associated with the presence of metabolic impairment, such as greater total and LDL cholesterol, triglycerides, FBG and HbA1c, and with greater gAUC during OGTT. Conversely, FIB-4 negatively correlated with fasting blood insulin and with iAUC at the OGTT. Significant associations between FIB-4, more elevated systolic and diastolic blood pressure and reduced kidney function were also found. Correlations between FIB-4 and clinical parameters are reported in Table 2.
An additional model adjusted for the same covariates was built to test the interaction between FIB-4 and IGF-1 in the prediction of osteopenia/osteoporosis and showed that changes of IGF-1 by FIB-4 value was an independent predictor of osteopenia/osteoporosis (R 2 = 0.18; p = 0.009).

FIB-4 and bone architecture disarrangement
In this study, the trabecular bone score (TBS) was used to assess the bone microarchitecture quality and was available for 654 study participants, representative of the entire study population ( Patients with altered bone structure had progressively higher FIB-4 score (normal bone, mean ± SD TBS: 0.60 ± 0.33; partially degraded, TBS: 0.77 ± 0.44, degraded, TBS: 0.95 ± 0.56; p < 0.001; Fig. 2).
Greater exposure to blood glucose during the OGTT, as assessed by gAUC, was associated with lower TBS (r = − 0.36, p < 0.001); no correlation was found, instead, between TBS and iAUC (r = − 0.07, p = 0.19). Similarly, we did not find any significant relationship between glucose or insulin AUC and parameters of bone mineralization, as the BMD and the T score, in any of the sites explored.

Discussion
In this study, we demonstrated that the estimated liver fibrosis is associated with osteopenia and osteoporosis in a large cohort of over 1800 obese individuals with NAFLD recruited in the same clinical setting. We showed that both the mineral bone density and microstructure progressively get impaired with the severity of liver damage in NAFLD, and this association is independent from classical risk factors of bone fragility and other metabolic disorders, such as impaired glucose tolerance and T2DM. In our population, the FIB-4 was higher in obese individuals with T2DM, regardless of age and sex; this result corroborates the existence of a reciprocal relationship between NAFLD and diabetes, where NAFLD-related liver damage is more severe in diabetic than non-diabetic individuals and once NAFLD occurs, it worsens metabolic control and diabetes' clinical outcomes [3]. Indeed, in patients belonging to the high FIB-4 group, T2DM prevalence rose from 11% to almost 29% in comparison to the low FIB-4. Although NAFLD is associated to hyperinsulinemia and insulin resistance, we observed significantly greater blood glucose and lower insulin AUC during the OGTT in individuals with higher estimated fibrosis score, as an additional evidence that the more pronounced liver damage in NAFLD correlates with relative insulin deficiency and, in turn, impaired glucose tolerance in obesity.
A very recent investigation conducted on a large sample from the U.S. general population demonstrated that android fat deposition pattern is associated with increased prevalence of NAFLD in both sexes, and correlated with the presence of fibrosis in women [21]. In line with our study, NAFLD individuals also displayed greater prevalence of T2DM and metabolic disease. Since visceral fat is an established risk factor for dysmetabolic conditions, android fat distribution may explain an excess of NAFLD risk and greater fibrosis, mostly in female sex. Further studies are warranted to evaluate this aspect also in the obese population.  In our study, the association between liver fibrosis and impaired bone metabolism was not mediated by the presence of impaired glucose tolerance and T2DM. This aspect suggests that specific hepatic factors, other than insulin resistance, could potentially affect bone health in dysmetabolic individuals with NAFLD.
In this context, a role may be played by the IGF-1, a hepato-derived growth hormone which displays anabolic effects on bone by inhibiting osteoblast apoptosis and enhancing osteoclast genesis [6]. In patients with advanced liver disease, the reduction of growth hormone receptors associated with the hepatocyte dysfunction was shown to contribute to the development of osteoporosis [22].
The finding of lower IGF-1 levels in our study participants with osteopenia and osteoporosis, prompted us to explore serum IGF-1 modifications in relation to the estimated liver fibrosis, as a potential link between NAFLD and bone fragility, finding linear inverse correlation between IGF-1 and FIB-4. Finally, we demonstrated the existence of an independent association between FIB-4 and lower IGF-1 after considering age, sex, BMI and the presence of MS' components in multivariable regression models. Indeed, IGF-1 reduction may represent the nexus between altered bone metabolism and NAFLD, mostly in presence of relatively advanced liver damage.
The relationship between impaired bone metabolism and NAFLD could also be bidirectional, as described for dysmetabolic conditions as T2DM, and mediated by circulating factors, other than insulin, that may influence both bone and liver metabolism. Among them, osteocalcin may play a role. Osteocalcin is a bone-derived non-collagen protein produced by the osteoblasts, its carboxylated form acts exclusively in the bone facilitating mineralization processes; the uncarboxylated osteocalcin is detectable in the blood circulation and displays several extra-skeletal metabolic functions [23]. Among those, osteocalcin protects from hepatosteatosis and progression toward NASH in experimental models, by binding its specific hepatic receptor GPRC6A [24,25]. Our study showed the existence of a linear negative correlation between osteocalcin levels and the FIB-4 in obese individuals with NAFLD; longitudinal studies ad hoc are warranted to investigate potential consequences of alter bone metabolism on liver pathophysiology in the clinical setting.
Here, we demonstrated that the bone mineral density linearly decreases with more elevated FIB-4 levels in obese subjects, despite an overall low prevalence of osteopenia and osteoporosis, even in presence of almost 80% female participants. However, one should consider the relative young age of study subjects, around 45 years, and that the DXA scan traditionally over-estimates the mineral bone density in overweight and obese individuals, likely in relation to interferences with the excessive fat mass [26].
The TBS was available in the subgroup of obese individuals undergoing DXA after the implementation of our system with the specific software for TBS assessment and no selection criteria was applied. In these participants, representative of the entire study population, the qualitative assessment of bone microarchitecture did unravel a prevalence of almost 75% individuals with degraded bone structure, as evaluated by the trabecular bone score assessment at lumbar spine level, once again associated with higher FIB-4 score. Indeed, FIB-4 may represent a predictor of impaired bone metabolism in obesity considering both mineral density and qualitative aspects of bone structure.
The estimated hepatic fibrosis, and not the amount of fat into the liver, associated with bone alterations in our study population. This result is in line with what observed in longitudinal investigations on NAFLD and BMD, where the presence of hepatosteatosis did associate with higher BMD at the baseline, but the BMD decline later in life was not different between NAFLD and non-NAFLD subjects. Nonetheless, greater basal FIB-4 predicted greater bone mineralization loss at the follow-up in the same cohort [10].
Very few studies so far investigated the association between NAFLD, liver fibrosis and bone metabolism, obtaining conflicting results that ranged from NAFLD being a risk factor [8,10], to have no [9] or even protective effects [27] on bone metabolism. However, studies finding favorable effects of NAFLD on bone density hypothesize potential interferences exerted by the increased body fat percentage on the BMD estimation in obese individuals [27]. More than half of the study participants had a diagnosis of MS at the time of study recruitment and a number of them were taking chronic therapies for the underlying cardiometabolic conditions. Although we considered this potential cofounder in the statistical procedures, finding no independent association between ongoing treatments and liver enzymes, the study design does not allow to definitely rule out an additional impact of treatments on blood transaminases in this population.
Our study has some strengths. First, in this investigation, we evaluated qualitative changes of bone architecture rather than measuring BMD-derived indexes alone, and to our knowledge, this is the first study exploring the association between TBS and FIB-4 in NAFLD. Moreover, this study was performed in a large cohort of obese individuals recruited in the same center: lab measurements and DXA assessments were, therefore, centralized for all participants.
Finally, we explored potential endocrine and metabolic pathways behind our study findings. Static and dynamic indexes of insulin resistance were calculated and accurate metabolic profiling of study participants was obtained; the circulating levels of biomarkers and hormones influencing bone metabolism were measured and correlated to parameters of liver involvement in NAFLD.
The cross-sectional design of this investigation is a limitation and does not allow to demonstrate a mechanistic link behind our study findings. However, the overall data point to potential independent effects of liver damage in NAFLD, i.e. lower IGF-1, that may contribute, along with insulin resistance and related metabolic impairment, to accelerated BMD loss and/or altered bone architecture in NAFLD patients. In this study, data on liver histology were not available and the presence of liver fibrosis was estimated calculating the FIB-4, a clinical index widely used in bariatric and non-bariatric populations, with predictive values for advanced fibrosis ranging from 0.80 and 0.86 in validation studies [17,18]. Data from meta-analysis recently concluded that FIB-4 provides comparable performance to liver biopsy in the stratification of patients for liver-related morbidity and mortality [27] and is currently mentioned as a first step evaluation in NAFLD risk stratification algorithms from International scientific societies [28]. Moreover, along with transaminases, among clinical parameters included in the FIB-4 index, the blood platelet count is proven to negatively correlate with histological liver damage, i.e. hepatocyte ballooning, steatosis and liver fibrosis, and is therefore considered a reliable marker of hepatic impairment in chronic liver diseases [29]. The presence of the age as a FIB-4 parameter could lead, instead, to falsely worse score in the elderly population, increasing the false-positive rate in population studies [20]. To mitigate this risk, in the present study, we considered a higher FIB-4 cut-off for individuals above 65 years old, as proposed in previous investigations [20], obtaining an overall rate of 8% individuals at moderate to high risk of liver fibrosis within our study participants, percentage that is in line with other reports [30]. Indeed, the FIB-4 could represent an easy, costless and repeatable index to improve the risk stratification strategies not only for liver disease progression but also for extra-hepatic metabolic alterations, i.e. bone fragility, in obese individuals with NAFLD. Another limitation is that this study has been conducted on a population of sole obese individuals; therefore, the general applicability of these findings should be demonstrated in additional studies conducted on different cohorts.

Conclusion
Our study demonstrates that increased FIB-4 score is associated with the presence of bone fragility in obese individuals with NAFLD, regardless of classical risk factors and other metabolic disturbances. Moreover, we reported for the first time the existence of relationship between estimated liver fibrosis and alterations of bone architecture. Studies with longitudinal design are warranted for exploring the possible involvement of liver-related factors in the development of increased bone fragility and fracture risk in dysmetabolic individuals.
Author contributions IB and MGC conceived the study. IB, MGC, FAC, GS, FL, AL and CL: coordinated the study; GP, SD, AD, ADB, FL, GS and CL: oversaw patient recruitment; FAC, SD, ADB, CL: finalized the data set. FAC, SD, ADB and CL: oversaw collection and analysis of biological samples. IB and MCG: performed statistical analyses. IB and MGC: drafted the paper, which was reviewed by all authors. All authors read and approved the final manuscript.
Funding This research was supported by Sapienza Università di Roma.
Data availability statement All data generated or analysed during this study are included in this published article.
Ethics approval and consent to participate All procedures followed were in accordance with the ethical standards of the responsible committee on human experimentation (Sapienza University, Rome, Italy) and with the Helsinki Declaration of 1975, as revised in 2008. Informed consent was obtained from all patients for being included in the study.