A predictive model for low bone mass in pediatric and adolescent patients with transfusion-dependent beta-thalassemia

Introduction

Beta-thalassemia (β-thalassemia) is an inherited disorder of impaired hemoglobin (Hb) synthesis caused by mutations in the β-protein gene. The global annual incidence is approximately 1/100,000 and is particularly high in the Mediterranean coast, Southeast Asia, and southern China (1). Patients with transfusion-dependent beta-thalassemia (TDT) often suffer from complications, such as iron overload, endocrine disruption, and multi-organ damage due to their long-term dependence on red blood cell transfusions for life support (2). Among them, low bone mass (LBM) due to skeletal system involvement has become a prominent issue that affects the quality of life of TDT patients. The risk of skeletal health is particularly prominent in the young population because they are in the critical period of bone mass accumulation (3,4).

Recent studies have stated that LBM prevalence in TDT patients is as high as 40–74%, greatly higher than that in the healthy population of the same age (5,6). LBM represents a precursor stage of osteoporosis in which the bone mineral density (BMD) is below the normal level but has not yet reached the diagnostic criteria for osteoporosis. In TDT patients, the pathophysiological mechanisms of LBM involve both hereditary and acquired factors. Hereditary factors include polymorphisms in the collagen type I alpha 1 gene (COL1A1) and the vitamin D receptor (VDR) gene. Acquired factors encompass primary conditions (e.g., bone marrow expansion and iron overload) and secondary factors, such as endocrine disorders [e.g., hypogonadism, growth hormone (GH) deficiency, diabetes, thyroid and parathyroid dysfunction], malnutrition (deficiencies in vitamin D), lack of physical activity, chronic liver and kidney disease, and the use of iron chelators. These factors disrupt bone metabolism by inhibiting osteoblast activity and increasing osteoclast function, leading to reduced bone formation and/or increased bone resorption, ultimately resulting in decreased BMD and impaired bone microstructure (7-12). Although international guidelines recommend regular monitoring for BMD in TDT patients, early warning tools for the risk of LBM in young patients remain scarce in clinical practice. Traditional dual-energy X-ray absorptiometry (DXA) can assess BMD but is limited by the high cost of testing instruments, the inadvisability of multiple measurements in the short term, and the lack of universal access (13,14). Predictive models based on accessible clinical indicators are urgently needed for risk stratification and individualized interventions.

Currently, domestic and international studies on skeletal complications in TDT patients focus on pathological mechanisms or risk factors, while multidimensional prediction models integrating clinical features, laboratory parameters, and bone metabolism markers are still lacking. In addition, dynamic changes in bone metabolism in young TDT patients have specific patterns due to growth demands and disease burden, and the conclusions of existing studies based on adults or non-transfusion-dependent thalassemia (NTDT) patients may not apply to this population (15,16). Therefore, constructing a prediction model for LBM risk in young TDT patients is crucial for optimizing clinical management strategies, reducing the risk of osteoporosis and fracture, and improving long-term prognosis. In recent years, with the continuous development of medical technology and the deepening of understanding of diseases, building predictive models has become an effective means to identify and assess disease risk in the early stage. In the field of TDT, although some studies have explored the pathogenesis of LBM and related risk factors, few studies focus on prediction models for LBM.

Given the high incidence of LBM in TDT patients and its significant impact on their quality of life, this study focuses on constructing a prediction model for LBM risk in young TDT patients. The Guangxi region of China provides a unique background for studying TDT-related LBM. The carrier rate of the β-thalassemia gene in this region is 6.66%, the highest in the country (17). As the National Hemopoietic Stem Cell Transplantation Center for Thalassemia, The First Affiliated Hospital of Guangxi Medical University has treated a large number of young TDT patients requiring hemopoietic stem cell transplantation (HSCT). Through a retrospective analysis of clinical data from this center, this study aims to develop a prediction model for LBM risk in young TDT patients, addressing the current gap in research. The findings are expected to provide a practical tool for clinical practice, enabling early identification of high-risk individuals and guiding targeted bone health management. We present this article in accordance with the STROBE reporting checklist (available at https://tp.amegroups.com/article/view/10.21037/tp-2025-385/rc).

Methods

Patient population

The data were obtained from the examination results of TDT patients in the Department of Hematology of The First Affiliated Hospital of Guangxi Medical University between January 2015 and December 2024. Inclusion criteria included: (I) patients diagnosed with TDT by Hb electrophoresis, DNA analysis, and clinical evidence of transfusion dependence, meeting the diagnostic criteria set in established guidelines (based on discharge or initial outpatient records) (18); (II) complete BMD test; (III) age less than 20 years old. Exclusion criteria were as follows: (I) with any disease known to affect bone mass; (II) long-term use of glucocorticoids or other drugs that affect bone metabolism; (III) missing data more than 30%. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study was approved by the Medical Ethics Committee of The First Affiliated Hospital of Guangxi Medical University (No. 2024-S1000-01) and individual consent for this analysis was waived due to the retrospective nature.

Data collection

The study collected patient demographics (age, sex, height, weight, and history of splenectomy), laboratory parameters [Hb, albumin (ALB), alkaline phosphatase (ALP), creatinine (Cr), total cholesterol (TC), triglycerides (TG), low-density lipoprotein cholesterol (LDLC), high-density lipoprotein cholesterol (HDLC), calcium (Ca), phosphorus (P), parathyroid hormone (PTH), 25-hydroxyvitamin D [25(OH)D], fasting plasma glucose (FPG), serum ferritin (SF), insulin-like growth factor 1 (IGF-1), testosterone (T), estradiol (E2), thyroid hormones [free triiodothyronine (FT3), free thyroxine (FT4), thyroid-stimulating hormone (TSH), osteocalcin (OC)], and lumbar spine bone mineral density values and their corresponding Z scores. Body mass index (BMI) was calculated as weight in kilograms divided by height in meters squared (kg/m²).

Definition

BMD was assessed by DXA and was described as the mean BMD of the patient’s lumbar spine (L1–L4) along with the corresponding Z-score. LBM was defined as a Z-score of lumbar spine BMD that was less than or equal to −2.0. Normal bone mass was defined as a Z-score of lumbar spine BMD that was greater than −2.0 (19).

Hypogonadism was diagnosed in females aged 13 years or older and males aged 14 years or older when their sex hormone levels remained at prepubertal or early pubertal levels, specifically with serum T levels less than 3.5 nmol/L in males and serum E2 levels less than 50 ng/L in females (20).

Abnormal thyroid function included hypothyroidism and subclinical hypothyroidism.

Hypothyroidism was characterized by a TSH level above the upper limit of the reference range and a T4 level below the lower limit of the reference range.

Subclinical hypothyroidism was defined as TSH levels greater than the upper limit of the reference range, while free T4 levels remained within the normal range. IGF-1 below −2SD referred to serum IGF-1 levels that were at least 2 standard deviations below the mean for the general population of the same race, age, and sex (21).

Statistical analyses

Normally distributed measurement data were described as mean ± standard deviation, and the t-test was applied for inter-group comparisons. Non-normally distributed measurement data were described as median (M) and interquartile range (P25, P75), and the Mann-Whitney U test was applied for inter-group comparisons. Count data were reported as the number of cases and percentages and were analyzed using the χ² test or Fisher’s exact probability method. The dataset was randomly split into a training set (to train the model) and a test set (to test the model) at a ratio of 8:2. The predictive modeling factors were screened, and the predictive factors with P<0.05 in univariate regression were selected for correlation analysis. Independent predictive factors were further included in the multivariate regression, and those that were still significant were used as the final predictive factors. The final predictive factors were incorporated into six machine learning (ML, which is a branch of artificial intelligence that enables computers to learn automatically from data and make predictions or decisions) methods [eXtreme Gradient Boosting Classifier, Logistic Regression, Light Gradient Boosting Machine Classifier, Gradient Boosting Classifier, Gaussian Naive Bayes (Gaussian NB), Support Vector Machine Classification] to construct a risk prediction model. First, tenfold resampling was employed to systematically compare the performance and stability of each model, thereby identifying the optimal model. Subsequently, the optimal model was trained and optimized using a five-fold cross-validation strategy, in which the data were divided into five parts, with four parts used for training and one part for validation to assess the model’s generalizability. Finally, the predictive performance of the model in the independent test set was reported. The model performance was evaluated by the area under the receiver operating characteristic (AUC) curve, and the calibration curve and decision curve analysis (DCA) were adopted to evaluate the consistency between the model-predicted probability and the actual probability and clinical practicality (Figure 1). Single interpolation (median for measurement data and mode for count data) was utilized to fill in missing data. For all statistical analyses, significance was set at P<0.05. Statistical analysis was performed in the Beckman Coulter DxAI platform (https://www.xsmartanalysis.com/beckman/login/).

Figure 1 Flowchart of model development. AUC, area under the curve; DCA, decision curve analysis; ROC, receiver operating characteristic.

Results

Patient characteristics

A total of 389 children and adolescents with TDT were included. According to the BMD Z score, the patients were assigned to the LBM group (N=123) and the normal bone mass group (N=266). There were differences in age, TG levels, lumbar spine BMD values, Z scores, hypogonadism, and IGF-1 below −2SD prevalence between the two groups (P<0.05). The incidence of LBM in children and adolescents with TDT was 31.6% (123/389). The baseline characteristics of patients are listed in Table 1. Baseline analysis between the training set and the test set showed that the groups were well balanced, without statistical differences (P>0.05) (Table 2).

Selection of predictors

In the univariate analysis, age, TG, IGF-1 below −2SD, and hypogonadism were greatly associated with LBM (P<0.05) (Table 3).

These four predictors were subjected to correlation analysis. The results showed no significant correlation between them (correlation coefficient <0.70) (Figure 2). Then, multivariate logistic regression analysis was performed on these four variables, and the results revealed that age, IGF-1 below −2SD, and hypogonadism were independent predictors of LBM (P<0.05) (Table 4).

Figure 2 Correlation analysis of predictors. IGF-1, insulin-like growth factor 1; SD, standard deviation; TG, triglycerides.

Model construction and selection

Age, IGF-1 below −2SD, and hypogonadism were included, and the model performance of six ML methods in the training set (Table 5) and validation set (Table 6) was evaluated. In the validation set, the Gaussian NB model had the largest AUC value (Figure 3A), good prediction consistency (Figure 3B), and the greatest clinical benefit (Figure 3C). Hence, it was selected as the optimal model for predicting the risk of LBM.

Figure 3 Performance of multiple machine learning models in the validation set: ROC curve (A), calibration curve (B), DCA curve (C). AUC, area under the curve; DCA, decision curve analysis; GBDT, Gradient Boosting Decision Tree; GNB, Gaussian Naive Bayes; LightGBM, Light Gradient Boosting Machine; Logistic, Logistic Regression; ROC, receiver operating characteristic; SD, standard deviation; SVM, Support Vector Machine; XGBoost, eXtreme Gradient Boosting.

Construction of a risk prediction model

The Gaussian NB model was the optimal model, with AUC =0.730, accuracy =0.667, sensitivity =0.567, and specificity =0.729 in the test set (Table 7). The ROC curve of the test set is displayed in Figure 4A. The calibration curve indicated that the Gaussian NB model had high accuracy and good fitness (Figure 4B). The DCA curve demonstrated that the model had clinical benefits (Figure 4C).

Figure 4 Gaussian NB model performance in the test set: ROC curve (A), calibration curve (B), DCA curve (C). AUC, area under the curve; CI, confidence interval; DCA, decision curve analysis; GNB, Gaussian Naive Bayes; ROC, receiver operating characteristic.

Construction of an online interface to easily access the model

Based on the optimal model, we developed an online tool to calculate the exact probability of LBM risk in children and adolescents with TDT (https://www.xsmartanalysis.com/model/list/predict/model/html?mid=26767&symbol=9eN1RZ754vUHn2795263). By inputting the patient’s relevant information, the tool can predict whether the patient is at high risk of LBM. For example, a 16-year-old TDT patient with IGF-1 below −2SD had an LBM risk probability of 77.2% (Figure 5).

Figure 5 An example of the predictive model to predict low bone mass in children and adolescents with TDT via a link. IGF-1, insulin-like growth factor 1; SD, standard deviation; TDT, transfusion-dependent beta-thalassemia.

Discussion

In this study, we constructed a prediction model for LBM risk, with age, IGF-1 below −2SD, and hypogonadism as predictors. The AUC value of the model was 0.730, accuracy was 0.667, sensitivity was 0.567, and specificity was 0.729.

Our study showed that LBM prevalence was 31.6% in children and adolescents with TDT. Previous studies have shown that LBM prevalence in TDT patients ranges from 40% to 74% (5,6). The lower prevalence observed in our research may be due to the relatively young age of the patient population.

Our study showed that aging, IGF-1 below −2SD, and hypogonadism were independent risk factors for LBM in children and adolescents with TDT. This finding is highly consistent with previous studies on the mechanism of bone diseases in TDT patients (3,22-24) and further clarifies the core role of endocrine dysfunction in bone metabolism imbalance in young patients with TDT.

The TDT patients included were aged between 2 and 19 years, with an average age of 8 years, and they were in the peak period of bone mass accumulation. The findings revealed that with increasing age, the BMD value showed an upward trend, but the BMD Z value showed a downward trend. This suggests that the increase in BMD in TDT patients is greatly lower than that in normal people of the same race, age, and sex, indicating that aging is a risk factor for LBM in children and adolescents with TDT. Older age is associated with a higher risk of LBM. This phenomenon may be due to the following reasons. First, long-term chronic anemia leads to the expansion of bone marrow cavities, and ineffective hematopoiesis impairs BMD. The continuous expansion of bone marrow cavities destroys the microstructure of trabeculae and aggravates bone resorption by activating the RANKL pathway (24). Second, the cumulative effect of iron overload gradually becomes apparent with age. Excess free iron inhibits osteoblast differentiation and promotes osteoclast activation by inducing oxidative stress (25). Finally, the risk of complications related to endocrine disorders and chronic hepatitis C virus (HCV) infection increases, which in turn exacerbate abnormal bone metabolism (26). Notably, half of an individual’s peak bone mass is accumulated during childhood and adolescence. By the age of 18 years, 90% of an individual’s peak bone mass has been reached (27,28), which is a critical period for peak bone mass accumulation. The physiological demand for rapid bone growth in adolescent TDT patients conflicts with their pathological state, which may prevent the achievement of peak bone mass.

This study found that disturbances in the GH-IGF-1 axis were considerably associated with LBM. Iron deposition due to repeated blood transfusions in TDT patients can directly damage the function of the anterior pituitary gland, resulting in insufficient GH secretion (29). IGF-1 below −2SD means that the level of IGF-1 is significantly lower than normal, which affects linear growth and weakens osteoblast activity. Studies have highlighted that GH can stimulate bone formation by activating the JAK2-STAT5 pathway (30). The IGF-1 level below −2SD indicates that hormone secretion is insufficient, which may block this pathway. Notably, even patients who receive standardized iron removal treatment still have abnormal GH secretion, suggesting that pituitary damage may be irreversible (31).

This study showed the protective effect of sex hormones on bone mass, consistent with previous research (32). Hypogonadism in TDT patients may be due to iron deposition-related gonadal toxicity. Estrogen and testosterone maintain bone homeostasis by inhibiting RANKL expression and promoting osteoblast proliferation (33). Importantly, bone loss in patients with delayed puberty is more significant than that in patients with onset after sexual maturity, suggesting the significant role of sex hormones in the critical period of bone mass accumulation. In addition, hypogonadism may indirectly exacerbate bone loss by increasing the secretion of proinflammatory cytokines (e.g., IL-6) in adipose tissue (34).

This study identified risk factors for LBM in children and adolescents with TDT and developed an online prediction tool to accurately predict the risk of LBM in this population. The advantages are mainly reflected in the following three aspects. First, we constructed an LBM risk prediction model for the special patient group, which is one of the few currently available models. The model variables are in accord with the characteristics of children and adolescents with TDT. Second, the variables used to construct the prediction model are simple and easy to obtain, increasing the feasibility of the model in clinical practice. Third, this study focuses on children and adolescents among TDT patients, with nearly 400 samples included, which is large among similar studies and has higher credibility. However, our study also has certain limitations. First, this is a retrospective study that provides weaker evidence than prospective studies. Therefore, these results should be interpreted with caution. Second, this study employed a cross-sectional design and failed to clarify the causality between various factors. Third, all cases enrolled were from a single center, which inevitably introduced bias and weakened statistical power. Finally, this study did not collect data regarding patients’ physical exercise, Ca, and vitamin D supplementation. Therefore, to address the above key issues, a longitudinal, prospective, multicenter study is urgently needed.

Conclusions

In this retrospective study of children and adolescents with TDT, we identified age, IGF-1 below −2SD, and hypogonadism as key factors affecting LBM. Therefore, the first management step is to improve these influencing factors to reduce the risk of LBM in such patients. In addition, our study suggests that a simple prediction model can be used as a screening tool to provide a clinical basis for early identification of high-risk patients.

Acknowledgments

None.

Footnote

Reporting Checklist: The authors have completed the STROBE reporting checklist. Available at https://tp.amegroups.com/article/view/10.21037/tp-2025-385/rc

Data Sharing Statement: Available at https://tp.amegroups.com/article/view/10.21037/tp-2025-385/dss

Peer Review File: Available at https://tp.amegroups.com/article/view/10.21037/tp-2025-385/prf

Funding: This research was supported by the Open Project of NHC Key Laboratory of Thalassemia Medicine (No. GJWJWDP202402).

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://tp.amegroups.com/article/view/10.21037/tp-2025-385/coif). The authors have no conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study was approved by the Medical Ethics Committee of The First Affiliated Hospital of Guangxi Medical University (No. 2024-S1000-01) and individual consent for this analysis was waived due to the retrospective nature.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. Ali S, Mumtaz S, Shakir HA, et al. Current status of beta-thalassemia and its treatment strategies. Mol Genet Genomic Med 2021;9:e1788. [Crossref] [PubMed]
2. Farmakis D, Porter J, Taher A, et al. 2021 Thalassaemia International Federation Guidelines for the Management of Transfusion-dependent Thalassemia. Hemasphere 2022;6:e732. [Crossref] [PubMed]
3. Ananvutisombat N, Tantiworawit A, Punnachet T, et al. Prevalence and risk factors predisposing low bone mineral density in patients with thalassemia. Front Endocrinol (Lausanne) 2024;15:1393865. [Crossref] [PubMed]
4. Fung EB, Sarsour I, Manzo R, et al. Bone quality is associated with fragility fracture in patients with hemoglobinopathies. J Clin Densitom 2025;28:101565. [Crossref] [PubMed]
5. Behzadifard S, Arianezhad A, Nazarinia D, et al. Bone Mineral Density, Osteoporosis Prevalence and Influential Factors in Osteogenesis in Patients with Beta Thalassemia Major: A Cross-Sectional Study. Iran J Public Health 2024;53:1883-9. [Crossref] [PubMed]
6. Algodayan S, Balachandar R, Papathanasiou N, et al. Bone mineral density in adult thalassaemias: a retrospective longitudinal study. Nucl Med Commun 2024;45:658-65. [Crossref] [PubMed]
7. Ekbote V, Padidela R, Khadilkar V, et al. Increased prevalence of fractures in inadequately transfused and chelated Indian children and young adults with beta thalassemia major. Bone 2021;143:115649. [Crossref] [PubMed]
8. De Sanctis V, Soliman AT, Elsefdy H, et al. Bone disease in β thalassemia patients: past, present and future perspectives. Metabolism 2018;80:66-79. [Crossref] [PubMed]
9. Manolopoulos PP, Lavranos G, Mamais I, et al. Vitamin D and bone health status in beta thalassemia patients-systematic review. Osteoporos Int 2021;32:1031-40. [Crossref] [PubMed]
10. Kothimira VK, Kumar A, Richhele LR, et al. An Evaluation of Bone Health Parameters in Regularly Transfused Beta-Thalassemia Major Patients. J Pediatr Hematol Oncol 2020;42:381-5. [Crossref] [PubMed]
11. Shamoon RP, Yassin AK, Omar N, et al. Magnitude of Bone Disease in Transfusion-Dependent and Non-Transfusion-Dependent β-Thalassemia Patients. Cureus 2024;16:e56012. [Crossref] [PubMed]
12. Taher AT, Farmakis D, Porter JB, et al. editors. Guidelines for the Management of Transfusion-Dependent β-Thalassaemia (TDT). Nicosia, Cyprus: Thalassaemia International Federation; 2025.
13. Zhou Y, Zhang D, Wu L, et al. Epidemiological survey of osteoporosis in Beijing over the past decade: a single-center analysis of dual-energy X-ray absorptiometry scans from 30 599 individuals. Journal of Southern Medical University 2025;45:443-52. [Crossref] [PubMed]
14. Wang Z, Tan Y, Zeng K, et al. Bone density measurement in patients with spinal metastatic tumors using chest quantitative CT deep learning model. J Bone Oncol 2024;49:100641. [Crossref] [PubMed]
15. Youssry I, Saad N, Madboly M, et al. Bone health in pediatric transfusion-dependent beta-thalassemia: Circulating osteoprotegerin and RANKL system. Pediatr Blood Cancer 2022;69:e29377. [Crossref] [PubMed]
16. Ismail UN, Azlan CA, Khairullah S, et al. Marrow Fat-Cortical Bone Relationship in β-Thalassemia: A Study Using MRI. J Magn Reson Imaging 2024;60:2447-56. [Crossref] [PubMed]
17. Writing Group for Practice Guidelines for Diagnosis and Treatment of Genetic Diseases, Medical Genetics Branch of Chinese Medical Association. Clinical practice guidelines for β-thalassemia. Chinese Journal of Medical Genetics 2020;37:243-51. [Crossref] [PubMed]
18. Red Blood Cell Diseases (Anemia) Group, Chinese Society of Hematology, Chinese Medical Association. Chinese guideline for diagnosis and treatment of transfusion dependentβ-thalassemia. Chinese Journal of Hematology 2022;43:889-96. [Crossref] [PubMed]
19. Bachrach LK, Gordon CM. Bone Densitometry in Children and Adolescents. Pediatrics 2016;138:e20162398. [Crossref] [PubMed]
20. Pediatric Endocrine Genetics and Metabolism Group of Chinese Medical Doctor Association, The Subspecialty Group of Endocrinological, Hereditary and Metabolic Diseases, the Society of Pediatrics, Chinese Medical Association, Adolescent Health and Medical Professional Committee of Chinese Medical Doctor Association. Expert consensus on diagnosis and treatment of hypogonadotropic hypogonadism in children. Chinese Journal of Pediatrics 2023;61:484-90. [Crossref] [PubMed]
21. Cao B, Peng Y, Song W, et al. Pediatric Continuous Reference Intervals of Serum Insulin-like Growth Factor 1 Levels in a Healthy Chinese Children Population - Based on PRINCE Study. Endocr Pract 2022;28:696-702. [Crossref] [PubMed]
22. Thavonlun S, Houngngam N, Kingpetch K, et al. Association of osteoporosis and sarcopenia with fracture risk in transfusion-dependent thalassemia. Sci Rep 2023;13:16413. [Crossref] [PubMed]
23. Carsote M, Vasiliu C, Trandafir AI, et al. New Entity-Thalassemic Endocrine Disease: Major Beta-Thalassemia and Endocrine Involvement. Diagnostics (Basel) 2022;12:1921. [Crossref] [PubMed]
24. Vogiatzi MG, Macklin EA, Fung EB, et al. Bone disease in thalassemia: a frequent and still unresolved problem. J Bone Miner Res 2009;24:543-57. [Crossref] [PubMed]
25. Lee SLK, Wong RSM, Li CK, et al. Prevalence and risk factors of fractures in transfusion dependent thalassemia - A Hong Kong Chinese population cohort. Endocrinol Diabetes Metab 2022;5:e340. [Crossref] [PubMed]
26. Wiromrat P, Rattanathongkom A, Laoaroon N, et al. Bone Mineral Density and Dickkopf-1 in Adolescents with Non-Deletional Hemoglobin H Disease. J Clin Densitom 2023;26:101379. [Crossref] [PubMed]
27. Rodrick E, Kindler JM. Bone mass accrual in children. Curr Opin Endocrinol Diabetes Obes 2024;31:53-9. [Crossref] [PubMed]
28. Bachrach LK. Hormonal Contraception and Bone Health in Adolescents. Front Endocrinol (Lausanne) 2020;11:603. [Crossref] [PubMed]
29. Soliman AT, Yassin MA, De Sanctis V. Final adult height and endocrine complications in young adults with β-thalassemia major (TM) who received oral iron chelation (OIC) in comparison with those who did not use OIC. Acta Biomed 2018;89:27-32. [Crossref] [PubMed]
30. Sanpaolo ER, Rotondo C, Cici D, et al. JAK/STAT pathway and molecular mechanism in bone remodeling. Mol Biol Rep 2020;47:9087-96. [Crossref] [PubMed]
31. Xu K, Fu Y, Cao B, et al. Association of Sex Hormones and Sex Hormone-Binding Globulin Levels With Bone Mineral Density in Adolescents Aged 12-19 Years. Front Endocrinol (Lausanne) 2022;13:891217. [Crossref] [PubMed]
32. Zhang YY, Xie N, Sun XD, et al. Insights and implications of sexual dimorphism in osteoporosis. Bone Res 2024;12:8. [Crossref] [PubMed]
33. Hsu SH, Chen LR, Chen KH. Primary Osteoporosis Induced by Androgen and Estrogen Deficiency: The Molecular and Cellular Perspective on Pathophysiological Mechanisms and Treatments. Int J Mol Sci 2024;25:12139. [Crossref] [PubMed]
34. Forte YS, Renovato-Martins M, Barja-Fidalgo C. Cellular and Molecular Mechanisms Associating Obesity to Bone Loss. Cells 2023;12:521. [Crossref] [PubMed]