Development and validation of a non-invasive nomogram for predicting bone metastasis in newly diagnosed breast cancer

Introduction

Breast cancer represents the primary malignant threat to women globally. According to GLOBOCAN 2022 data, over 2.3 million new breast cancer cases occur annually worldwide, accounting for nearly 25% of all female cancer cases (1). In China, the age-standardized incidence rate of breast cancer shows a persistent upward trend (approximately 3–4% increase per decade), with approximately 420,000 new cases per year. The disease also exhibits a younger age of onset, posing a significant public health challenge (2).

Early-stage breast cancer is often asymptomatic, leading to diagnosis at non-localized stages for many patients. Its harm extends beyond local invasion of the primary tumor to a high risk of distant metastasis. Tumor cells can disseminate via hematogenous or lymphatic routes, with bone being the most common site of distant metastasis. Studies indicate that approximately 65–75% of patients with advanced breast cancer develop bone metastasis (BM), leading to skeletal-related events such as pathological fractures, spinal cord compression, and intractable bone pain, which significantly reduce treatment efficacy and shorten survival (3). In contrast, the five-year overall survival rate for breast cancer patients without BM typically exceeds 80% (4). Therefore, during the critical window of initial breast cancer diagnosis, accurately identifying individuals at high risk of BM is of paramount clinical value for implementing early interventions and improving prognosis.

However, a significant gap exists in current clinical practice for precisely predicting individualized BM risk at the initial diagnosis stage. Diagnosis of BM relies primarily on imaging [e.g., X-ray, computed tomography (CT), magnetic resonance imaging (MRI), bone scintigraphy, positron emission tomography-computed tomography (PET-CT)] and pathological biopsy, which have notable limitations: (I) conventional imaging (e.g., X-ray) requires a certain degree of bone destruction (>5 mm) for detection, potentially missing the optimal window for early intervention (5); (II) while PET-CT has higher sensitivity, it is expensive and has limited availability in primary care settings. Furthermore, skeletal inflammation or trauma can cause false positives, and lytic lesions are prone to being missed (6); (III) bone lesion biopsy is invasive and difficult to repeat (7,8). To address these challenges, there is an urgent need to develop a noninvasive tool for predicting BM risk based on routine clinical laboratory indicators available at the time of initial breast cancer diagnosis.

Nomograms, owing to their intuitiveness, ease of use, and suitability for rapid bedside calculation, demonstrate strong value in clinical prediction models (9). Although predictive models for distant metastasis or metastasis to specific organs (e.g., liver, lung) in breast cancer exist (10,11), nomogram models specifically focused on BM risk in newly diagnosed patients, combining clinical practicality with rigorous validation, remain relatively scarce. Therefore, this study aimed to construct and rigorously validate a nomogram model for predicting BM risk based on a cohort of patients at their first breast cancer diagnosis, providing a reliable and practical tool for clinical decision support in early warning and individualized management. We present this article in accordance with the TRIPOD reporting checklist (available at https://tcr.amegroups.com/article/view/10.21037/tcr-2025-1-2776/rc).

Methods

Study design and population

This retrospective case-control study analyzed data from breast cancer patients pathologically confirmed at Luzhou People’s Hospital between January 2010 and January 2024. Based on the presence or absence of BM at initial diagnosis, patients were categorized into the breast cancer with BM group (BCBM group) and the breast cancer without BM group (non-BCBM group). This study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study was approved by the Ethics Committee of Luzhou People’s Hospital, and all patients provided written informed consent. Inclusion criteria: (I) female patients aged 18–75 years; (II) histopathological diagnosis of breast invasive ductal carcinoma from breast mass tissue; (III) diagnosis of BM met one of the following criteria: imaging (X-ray, CT, MRI, PET-CT) showing typical signs of BM, or pathological confirmation of breast cancer metastasis via bone lesion biopsy; (IV) complete pathological and clinical data. Exclusion criteria: (I) concurrent other primary malignancies (e.g., lung cancer, liver cancer); (II) concurrent metastasis of breast cancer to other sites (e.g., liver metastasis, brain metastasis); (III) severe cardiovascular/cerebrovascular diseases or autoimmune diseases.

Study variables

• Demographic and clinical variables: age, body mass index (BMI), menstrual, hypertension, type 2 diabetes mellitus (T2DM), smoking, drinking.
• Laboratory test indicators: white blood cell (WBC), neutrophil (NEUT), platelet (PLT), red blood cell (RBC), hemoglobin, total cholesterol (TC), triglyceride (TG), alkaline phosphatase (ALP), albumin (ALB), cancer antigen 15-3 (CA15-3), carcinoembryonic antigen (CEA), calcium.
• Imaging features: primary breast tumor location, tumor number, tumor size.

Statistical analysis

Data analysis was performed using R statistical software (version 4.3.0). Continuous data are presented as: mean ± standard deviation for normally distributed data, analyzed using the two-sided t-test; median [interquartile range (IQR)] for non-normally distributed data, analyzed using the Mann-Whitney U test. Categorical data were analyzed using Pearson’s chi-square test or Fisher’s exact test. A two-sided P<0.05 was considered statistically significant.

Prior to formal statistical analysis, we systematically evaluated the completeness of the dataset. A comprehensive review of all candidate variables revealed that the missing rate for each variable was <5%, indicating minimal data incompleteness. To optimize data utilization and reduce potential selection bias, we employed multiple imputation to handle missing values, which is recommended by statistical methodological guidelines for datasets with low missing rates (in contrast to complete-case analysis, which may exclude valuable observations and introduce selection bias). The imputation was performed using the mice package in R software with 5 imputed datasets, and predictive mean matching was selected as the imputation method (appropriate for mixed continuous and categorical variables).

All eligible subjects were randomly allocated to a training cohort (for model construction) and a validation cohort (for model validation) in a 7:3 ratio. First, univariate logistic regression analysis (with BM as the dependent variable) was performed on all potential predictor variables within the training cohort. Variables with P<0.05 in the univariate analysis were included in the multivariate logistic regression analysis (using backward stepwise selection). Variables retaining P<0.05 in the multivariate analysis were identified as independent predictors of breast cancer BM.

Based on the final independent predictors selected from the multivariate logistic regression analysis, a nomogram model was constructed to predict the probability of breast cancer BM occurrence. The model was evaluated in both the training and validation cohorts using the following metrics. For discrimination assessment, the concordance index (C-index) was calculated, and receiver operating characteristic (ROC) curves with area under the curve (AUC) and 95% confidence intervals (CIs) were generated for both training and validation cohorts using the pROC package in R (version 4.3.1), which assessed the model’s ability to distinguish between patients with and without BM; for calibration assessment, the Hosmer-Lemeshow goodness-of-fit test (predefined P>0.05 for good calibration) and decile-based calibration curves were used to visualize the agreement between predicted probabilities and observed outcomes-calibration curves group predicted probabilities into deciles to evaluate prediction accuracy by comparing predicted values against actual observed probabilities, with ideal alignment closely following the diagonal line; for clinical utility assessment, decision curve analysis (DCA) was conducted via the rmda package to quantify net benefit across threshold probabilities, where the DCA curve plots threshold probability on the x-axis and net benefit on the y-axis, evaluating the model’s clinical net benefit across different decision thresholds to quantify its practical application value and provide a basis for optimal clinical decision-making.

Results

Overall patient characteristics

A total of 376 breast cancer patients were enrolled, including 134 patients (35.64%) with BM at diagnosis (BCBM group) and 242 patients (64.36%) without BM (non-BCBM group). Comparison of baseline characteristics between the two groups (detailed in Table 1) revealed statistically significant differences in six key variables: ALP, primary tumor size, CA15-3 level, and RBC were significantly higher in the BCBM group (all P<0.001), while serum ALB level (P<0.001) and the prevalence of T2DM (P=0.02) were significantly lower in the BCBM group.

Characteristics of training and validation cohorts

The 376 patients were randomly assigned to a training cohort (n=263) and a validation cohort (n=113) at a 7:3 ratio. Baseline characteristics were well-balanced between the two cohorts, with no statistically significant differences observed in any of the assessed variables (all P>0.05; detailed in Table 2), indicating the reliability of subsequent model development and validation.

Predictor variable screening and nomogram construction

In the training cohort, univariate logistic regression analysis was first performed on the data, preliminarily analyzing 23 variables including basic patient characteristics and laboratory tests. Multivariate analysis identified the following statistically significant variables: ALB [odds ratio (OR) =0.89, 95% CI: 0.83–0.95], ALP (OR =1.01, 95% CI: 1.01–1.03, P=0.02), CA15-3 (OR =1.02, 95% CI: 1.01–1.02, P<0.001), Tumor size (OR =1.02, 95% CI: 1.01–1.05, P=0.005). Table 3 displays the results of the logistic regression analyses. To rule out multicollinearity among the retained predictors, we conducted variance inflation factor (VIF) and tolerance diagnostics (results summarized in Table S1). The VIF values for the final four variables were: ALB: 1.0026, ALP: 1.0094, CA15-3: 1.0041, and tumor size: 1.0099. All VIF values are far below 3, confirming that the independent effects of each predictor are not distorted by inter-variable correlations. This result further supports the validity of our variable selection.

Based on these four independent predictors, a nomogram model for predicting breast cancer BM risk was constructed (Figure 1). The model visually demonstrates that tumor size, CA15-3 level, and ALP level contribute positively to BM risk (higher scores indicate greater risk), while ALB level contributes negatively (higher scores indicate lower risk). Users can obtain points based on the patient’s specific indicator values on the corresponding axis lines, sum all points to get a total score, and then read the corresponding probability of BM occurrence on the risk probability axis at the bottom. Conclusion: High ALP, high CA15-3, large tumor size, and low ALB are key factors predicting BM in newly diagnosed breast cancer patients.

Figure 1 Nomogram for predicting the probability of bone metastasis in breast cancer. ALB, albumin; ALP, alkaline phosphatase; CA15-3, cancer antigen 15-3.

Nomogram performance

The model’s effectiveness was evaluated using AUC. The AUC values for the training and validation cohorts were 0.85 (95% CI: 0.80–0.90) and 0.79 (95% CI: 0.69–0.89), respectively, indicating good discriminative ability for the training set and acceptable discriminative ability for the validation set (Figure 2A,2B). The Hosmer-Lemeshow goodness-of-fit test yielded P>0.99 for the training cohort and P=0.83 for the validation cohort, demonstrating good model fit. Furthermore, for both cohorts, the predicted probability curves in the calibration plots were very close to the standard diagonal line, indicating promising agreement with actual observations (Figure 2C,2D). DCA was used to assess the model’s clinical value. In the DCA curve, the horizontal coordinate represents the threshold probability, while the vertical coordinate represents the net benefit probability. In the validation cohort of this study, the DCA curve showed greater net benefit (Figure 2E,2F). Overall, the constructed nomogram performed well, exhibiting good predictive capability.

Figure 2 Prediction model performance. The discrimination between the training cohort (A) and the validation cohort (B) was shown by the receiver operating characteristic curves. The calibration curves described the agreement between the predicted probabilities and the observed results based on the training (C) and the validation cohorts (D). The clinical benefits of the training cohort (E) and validation cohort (F) were proven using the decision curve analysis. AUC, area under the curve; CI, confidence interval.

In routine clinical practice, the nomogram can be implemented as follows, with risk thresholds derived from the models DCA to optimize net benefit. Clinicians first retrieve four readily available baseline indicators (ALB, ALP, CA15-3, and primary tumor size) from initial diagnostic datasets. Corresponding points are assigned to each indicator using the nomogram’s s scoring scales, and the total score is mapped to the predicted probability of BM. Risk-stratified clinical decisions are then made: high-risk (predicted probability ≥20%) warrants immediate confirmatory screening and prophylactic interventions if metastasis is confirmed; intermediate-risk (10–20%) requires regular surveillance (serum marker recheck every 3 months and bone scintigraphy every 6 months); low-risk (<10%) follows standard breast cancer management without additional BM screening. A representative case: a patient with ALB =32 g/L, ALP =130 U/L, CA15-3 =40 U/mL, and tumor size =50 mm yields a total score of 18, corresponding to a ~32% predicted risk, indicating the need for urgent screening.

Discussion

This study successfully developed and validated a nomogram model based on routine clinical indicators (ALP, CA15-3, primary tumor size, ALB) for the individualized prediction of BM risk in newly diagnosed breast cancer patients. The model demonstrated good discrimination, excellent calibration, and positive potential for clinical net benefit in both the training and validation sets, providing a simple, practical, and reliable tool for the rapid bedside assessment of early BM risk.

Breast cancer metastasis occurs primarily through lymphatic spread, hematogenous dissemination, and direct invasion, with hematogenous spread being the main route for BM (12,13). As the most common site of distant metastasis in breast cancer, BM significantly impacts patient quality of life and survival prognosis. In this study, the incidence of BM among newly diagnosed breast cancer patients was 35.64%, lower than the reported rate (65–75%) for patients with advanced disease (3). This precisely underscores the value of this study focusing on the “newly diagnosed” patient population-aiming to identify high-risk individuals earlier. Currently, routine screening for BM (e.g., bone scintigraphy) is primarily based on clinical symptoms (bone pain, fracture risk), lacking evidence for universal screening in asymptomatic high-risk populations. Therefore, developing effective clinical prediction models to assist physicians in identifying patients at high risk of BM at the initial diagnosis stage is crucial for optimizing screening strategies and enabling early diagnosis and intervention.

CA15-3 is one of the most widely used circulating tumor markers for breast cancer, primarily reflecting tumor burden and disease progression (14). During breast cancer development, proliferation, invasion, and metastasis of tumor cells lead to substantial release of CA15-3 into the bloodstream, elevating its serum levels. High levels reflect active proliferation and strong invasive potential of tumor cells; these highly invasive cells are more likely to breach the basement membrane, enter the bloodstream or lymphatic circulation, and subsequently metastasize to distant sites like bone (15). Simultaneously, CA15-3 may participate in interactions between tumor cells and the extracellular matrix, promoting tumor cell migration and adhesion, facilitating colonization and metastasis formation in bone (16,17). Additionally, as a bone biochemical marker, elevated ALP levels can serve as a predictor of BM progression and prognosis in breast cancer patients (18). This study found significantly higher ALP in the BCBM group compared to the non-BCBM group, with an OR >1, clearly indicating its sensitivity as a marker of BM burden or osteoblastic reaction. Breast cancer BM is predominantly osteolytic. When BM occurs, tumor cells disrupt the normal bone microenvironment, activating osteoclasts and leading to bone resorption and destruction (19,20). During the repair process following bone destruction, osteoblast activity increases, resulting in abundant secretion of ALP (21). Consequently, elevated serum ALP levels serve as an important indicator of abnormal bone metabolism, reflecting the dynamic changes of bone destruction and remodeling during metastasis.

Tumor size is a risk factor for BM in breast cancer patients. As tumor volume increases, the number of tumor cells rises, and angiogenesis within the tumor becomes more active. Newly formed vessels often lack a complete basement membrane, are prone to forming arteriovenous anastomoses or blind ends, and are tortuous (22); their capillary permeability is also higher than normal vessels, providing more opportunities for tumor cells to enter the bloodstream (23). Larger tumors are more likely to breach local tissue barriers, invade surrounding blood and lymphatic vessels, and metastasize to distant sites like bone via hematogenous or lymphatic routes (24). Furthermore, increased tumor size may lead to greater tumor cell heterogeneity, with subpopulations possessing high metastatic potential being more prone to metastasis (25-27). Moreover, tumor size is closely related to tumor stage and malignancy; larger tumors often represent later stages and higher malignancy grades, consequently having a greater likelihood of developing BM.

Notably, this study is the first to identify low serum ALB level as an independent predictor (protective factor, OR <1) of breast cancer BM. Potential mechanisms may involve: (I) nutritional status and cachexia: low ALB is a marker of cancer-related malnutrition and cachexia. Cachexia weakens overall anti-tumor immunity; high tumor metabolic activity consumes nutrients, reducing ALB synthesis and increasing its breakdown (28). (II) Immune function: reduced ALB levels may impair immune function, weaken immune surveillance, allow tumor cells to evade immune clearance, and promote metastasis (29). (III) Tissue barrier function: low ALB may affect tissue repair and the maintenance of normal physiological barriers, creating conditions for tumor cells to breach the local microenvironment and enter the circulation for dissemination (30). (IV) Microenvironment regulation: ALB may indirectly influence tumor cell adhesion, migration, and invasion capabilities by affecting the composition or function of the extracellular matrix (31).

It is noteworthy that this study did not find significant differences in age/menopausal status between the two groups, which contrasts with some previous studies. Prior research suggested that menopause affects sex hormone and blood calcium levels and leads to reduced bone density, making bone tissue more fragile and susceptible, thereby providing a favorable environment for breast cancer cell growth and spread in bone, increasing BM risk (3,32). Furthermore, known important predictors such as breast cancer molecular subtypes (hormone receptor status, HER2 status) and detailed clinical stage (33) were not included in this model. However, this was a deliberate focus of the study design—to explore the feasibility of early risk prediction based on routinely available clinical and laboratory indicators at initial diagnosis, without relying on complex molecular testing or complete staging information. While molecular subtypes and staging information hold significant value, obtaining them at initial diagnosis may require additional time or resources.

The value of this model lies in providing a rapid, convenient tool for preliminary risk assessment specifically for patients newly diagnosed with Invasive Ductal Carcinoma. The nomogram format adopted perfectly aligns with the clinical need for accessible tools. Physicians can quickly calculate a total score based on the patient’s tumor size (imaging report), serum CA15-3, ALP, and ALB (routine tests), and read the corresponding quantified probability of BM risk. This individualized, continuous prediction approach overcomes the relative broadness of traditional staging systems, aiding in more precise risk stratification. Rigorous performance validation of the model (good AUC, excellent calibration, positive DCA results) provides strong support for its clinical application.

However, there are also limitations in this study: (I) retrospective studies carry inherent risks of selection bias and information bias; (II) enrolled patients had only invasive ductal carcinoma, excluding other pathological types of breast cancer, limiting the nomogram’s applicability to a broader breast cancer patient population; (III) this study only implemented internal validation, where both the training and validation cohorts were derived from the same single-center retrospective dataset; no external validation was conducted using independent patient populations from different institutions, regions, or clinical settings. The absence of external validation may restrict the external validity and generalizability of the constructed nomogram, as the model’s predictive performance could potentially vary across diverse clinical scenarios and patient demographics. Future research should employ prospective, multi-center designs, include patients with different pathological types of breast cancer, conduct rigorous external validation to verify the model’s stability and reliability in heterogeneous populations, and consider integrating more dimensional information such as molecular subtypes and genomic features to further optimize and validate this nomogram model. Simultaneously, in-depth investigation into the specific biological mechanisms linking the predictors identified by this model (especially ALB) to the occurrence and development of breast cancer BM will provide a deeper theoretical basis for risk prediction and potential intervention targets.

Conclusions

In summary, our study identified ALB, ALP, CA15-3, tumor size as independent risk factors for bone metastasis in newly diagnosed breast cancer patients. The nomogram model constructed from these readily available clinical indicators shows favorable discrimination, calibration, and clinical utility, representing a convenient noninvasive tool for individualized risk assessment. This predictive model may assist clinicians in early identification of high‑risk patients and facilitate timely targeted monitoring and therapeutic strategies to improve patient prognosis.

Acknowledgments

None.

Footnote

Reporting Checklist: The authors have completed the TRIPOD reporting checklist. Available at https://tcr.amegroups.com/article/view/10.21037/tcr-2025-1-2776/rc

Data Sharing Statement: Available at https://tcr.amegroups.com/article/view/10.21037/tcr-2025-1-2776/dss

Peer Review File: Available at https://tcr.amegroups.com/article/view/10.21037/tcr-2025-1-2776/prf

Funding: None.

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://tcr.amegroups.com/article/view/10.21037/tcr-2025-1-2776/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. This study was approved by the Ethics Committee of the Luzhou People’s Hospital, and conducted in accordance the Declaration of Helsinki and its subsequent amendments. All subjects gave informed consent to be included in the study before their participation.

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. Bray F, Laversanne M, Sung H, et al. Global cancer statistics 2022: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries. CA Cancer J Clin 2024;74:229-63. [Crossref] [PubMed]
2. The global, regional, and national burden of cancer, 1990-2023, with forecasts to 2050: a systematic analysis for the Global Burden of Disease Study 2023. Lancet 2025;406:1565-86.
3. Anwar SL, Avanti WS, Dwianingsih EK, et al. Risk Factors, Patterns, and Distribution of Bone Metastases and Skeletal-Related Events in High-Risk Breast Cancer Patients. Asian Pac J Cancer Prev 2022;23:4109-17. [Crossref] [PubMed]
4. Miller KD, Nogueira L, Devasia T, et al. Cancer treatment and survivorship statistics, 2022. CA Cancer J Clin 2022;72:409-36. [Crossref] [PubMed]
5. Gondim Teixeira PA, Lombard C, Moustache-Espinola P, et al. Initial Characterization of Focal Bone Lesions with Conventional Radiographs or Computed Tomography: Diagnostic Performance and Interobserver Agreement Assessment. Can Assoc Radiol J 2023;74:404-14. [Crossref] [PubMed]
6. Ariga A, Matsumoto S, Tanizawa T, et al. Bone metastases with "false negative" findings on 18F-FDG PET/CT in patients with angiosarcoma: A case series with literature review. Medicine (Baltimore) 2023;102:e34196. [Crossref] [PubMed]
7. Abboud S, Kosaraju V, Bhatt A, et al. Multidisciplinary approach for repeat musculoskeletal lesion biopsy after nondiagnostic initial sampling: A 10-year single-center experience. J Surg Oncol 2021;123:342-51. [Crossref] [PubMed]
8. Tomasian A, Jennings JW. Bone marrow aspiration and biopsy: techniques and practice implications. Skeletal Radiol 2022;51:81-8. [Crossref] [PubMed]
9. Yu AF, Lin IH, Jorgensen J, et al. Nomogram for Predicting Risk of Cancer Therapy-Related Cardiac Dysfunction in Patients With Human Epidermal Growth Factor Receptor 2-Positive Breast Cancer. J Am Heart Assoc 2023;12:e029465. [Crossref] [PubMed]
10. Lin Z, Yan S, Zhang J, Pan Q. A Nomogram for Distinction and Potential Prediction of Liver Metastasis in Breast Cancer Patients. J Cancer 2018;9:2098-106. [Crossref] [PubMed]
11. Zhang L, Pan J, Wang Z, et al. Construction of a MicroRNA-Based Nomogram for Prediction of Lung Metastasis in Breast Cancer Patients. Front Genet 2020;11:580138. [Crossref] [PubMed]
12. Markiewicz A, Nagel A, Szade J, et al. Aggressive Phenotype of Cells Disseminated via Hematogenous and Lymphatic Route in Breast Cancer Patients. Transl Oncol 2018;11:722-31. [Crossref] [PubMed]
13. Scully OJ, Bay BH, Yip G, et al. Breast cancer metastasis. Cancer Genomics Proteomics 2012;9:311-20.
14. Tarighati E, Keivan H, Mahani H. A review of prognostic and predictive biomarkers in breast cancer. Clin Exp Med 2023;23:1-16. [Crossref] [PubMed]
15. Gao Y, Liu M, Shi S, et al. Diagnostic value of seven biomarkers for breast cancer: an overview with evidence mapping and indirect comparisons of diagnostic test accuracy. Clin Exp Med 2020;20:97-108. [Crossref] [PubMed]
16. Fan Y, Chen X, Li H. Clinical value of serum biomarkers CA153, CEA, and white blood cells in predicting sentinel lymph node metastasis of breast cancer. Int J Clin Exp Pathol 2020;13:2889-94.
17. Li X, Dai D, Chen B, et al. Clinicopathological and Prognostic Significance of Cancer Antigen 15-3 and Carcinoembryonic Antigen in Breast Cancer: A Meta-Analysis including 12,993 Patients. Dis Markers 2018;2018:9863092. [Crossref] [PubMed]
18. Jung JH, Hong CM, Jo I, et al. Reliability of Alkaline Phosphatase for Differentiating Flare Phenomenon from Disease Progression with Bone Scintigraphy. Cancers (Basel) 2022;14:254. [Crossref] [PubMed]
19. Jiang C, Hu F, Xia X, et al. Prognostic value of alkaline phosphatase and bone-specific alkaline phosphatase in breast cancer: A systematic review and meta-analysis. Int J Biol Markers 2023;38:25-36. [Crossref] [PubMed]
20. Xu D, Tang M. Advances in the study of biomarkers related to bone metastasis in breast cancer. Br J Radiol 2023;96:20230117. [Crossref] [PubMed]
21. Jiang T, Zeng Q, He J. Do alkaline phosphatases have great potential in the diagnosis, prognosis, and treatment of tumors? Transl Cancer Res 2023;12:2932-45. [Crossref] [PubMed]
22. Hachey SJ, Hatch CJ, Gaebler D, et al. Targeting tumor-stromal interactions in triple-negative breast cancer using a human vascularized micro-tumor model. Breast Cancer Res 2024;26:5. [Crossref] [PubMed]
23. Liu Z, Xin B, Zhang N, et al. LSD1 modulates the bone metastasis of breast cancer cells through hnRNPA2B1-mediated sorting of exosomal miRNAs. Cell Death Discov 2024;10:115. [Crossref] [PubMed]
24. Zhang D, Dihge L, Bendahl PO, et al. Deep learning on routine full-breast mammograms enhances lymph node metastasis prediction in early breast cancer. NPJ Digit Med 2025;8:425. [Crossref] [PubMed]
25. Marusyk A, Almendro V, Polyak K. Intra-tumour heterogeneity: a looking glass for cancer? Nat Rev Cancer 2012;12:323-34. [Crossref] [PubMed]
26. Sopik V, Narod SA. The relationship between tumour size, nodal status and distant metastases: on the origins of breast cancer. Breast Cancer Res Treat 2018;170:647-56. [Crossref] [PubMed]
27. Sørlie T, Perou CM, Tibshirani R, et al. Gene expression patterns of breast carcinomas distinguish tumor subclasses with clinical implications. Proc Natl Acad Sci U S A 2001;98:10869-74. [Crossref] [PubMed]
28. Arends J, Bachmann P, Baracos V, et al. ESPEN guidelines on nutrition in cancer patients. Clin Nutr 2017;36:11-48. [Crossref] [PubMed]
29. Kumar V, Stewart JH 4th. Editorial: Immunology of cachexia. Front Immunol 2023;14:1339263. [Crossref] [PubMed]
30. Fu X, Yang Y, Zhang D. Molecular mechanism of albumin in suppressing invasion and metastasis of hepatocellular carcinoma. Liver Int 2022;42:696-709. [Crossref] [PubMed]
31. Zhao Y, Zheng X, Zheng Y, et al. Extracellular Matrix: Emerging Roles and Potential Therapeutic Targets for Breast Cancer. Front Oncol 2021;11:650453. [Crossref] [PubMed]
32. Yazdani A, Dorri S, Atashi A, et al. Bone Metastasis Prognostic Factors in Breast Cancer. Breast Cancer (Auckl) 2019;13:1178223419830978. [Crossref] [PubMed]
33. Xiong Z, Deng G, Huang X, et al. Bone metastasis pattern in initial metastatic breast cancer: a population-based study. Cancer Manag Res 2018;10:287-95. [Crossref] [PubMed]