The cancer spectrum in China is transitioning from that of developing countries to developed countries, with the aging of the population, changes in social lifestyle, and dietary structure. Prostate cancer has emerged as one of the significant factors posing a threat to the life and health of men in China (Zhang et al. 2023; Nasir et al. 2020).
Currently, numerous studies have confirmed that tumor development involves a wide range of factors and is the end product of a multi-step process. And the development of tumor bone metastasis is also a multi-step and complex process whose mechanism is not yet exact, including colonization, dormancy, reactivation and development, and reconstruction (Zhang 2019).
During tumor development, genomic instability and mutations in tumor cells are considered to be fundamental driving features in tumor progression. Tumor cells interact and co-evolve with the microenvironment they reside in, which is referred to as the tumor microenvironment (TME). This microenvironment enhances tumor cell proliferation, migration, and immune escape (Cheng et al. 2023). TME consists primarily of tumor cells, non-tumor cells (such as immune cells, vascular endothelial cells, and fibroblasts), and extracellular components (including extracellular matrix, extracellular vesicles, growth factors, chemokines, and cytokines) (Dzobo 2020).
Among non-tumor cells, immune cells play a crucial role in the tumor microenvironment (TME). Depending on the microenvironment and tumor type, these immune cells can either have an anti-TME effect or contribute to immunosuppression. Specifically, macrophages recruited from circulating monocytes to the TME are known as tumor-associated macrophages (TAM). TAMs are considered a diverse and adaptable cell population within the TME (Cassetta and Pollard 2020). Tumor-associated macrophages (TAM) play a role in promoting disease progression and resistance to therapy. They provide nutritional support to malignant cells by secreting growth factors, chemokines, cytokines, and promoting intra-tumor angiogenesis. M1-like macrophages, induced by IFN-γ or GM-CSF, exhibit highly glycolytic metabolism along with lactate secretion, lipid biosynthesis, and nucleotide production. These macrophages also produce reactive oxygen species, which contribute to their cytocidal function (Vitale et al. 2019). On the other hand, M2-like macrophages, induced by IL-4 and IL-13, employ oxidative metabolism for bioenergetic purposes. This metabolic profile is associated with immune regulation, repair functions, wound healing, and angiogenesis (Pan et al. 2021). In TME, neutrophils show a strong plasticity and are able to continuously adapt their functions in different inflammatory environments (Giese et al. 2019). In the initial phases of tumor growth, neutrophils are recruited into the tumor microenvironment (TME) where they contribute to inflammation by releasing cytokines and reactive oxygen species, thereby promoting apoptosis. However, as the tumor develops further, neutrophils play a role in facilitating tumor growth. These do this by modifying the extracellular matrix, releasing vascular endothelial growth factor, and producing matrix metalloproteinase-9. These actions stimulate angiogenesis, leading to tumor progression and local infiltration (Anderson and Simon 2020).
Furthermore, with a comprehensive and deeper understanding of coagulation mechanisms, a close link to TME has been found. Due to the interaction and influence of the coagulation process and TME, the clotting cascade can have multiple direct and indirect effects on cancer cells and the composition of TME both inside and outside the vascular space, and may ultimately enable us to control tumor immune response or vascular integrity (Galmiche et al. 2022). It has been found that the coagulation process has a role in regulating cells in the TME (Vago et al. 2019). Macrophages, for example, help remodel fibrin polymers and eliminate dead cells during normal healing. Fibrinolytic enzymes also stimulate macrophage chemotaxis and phagocytosis. In addition, dysfunctional tumor vasculature and fibrin create a "physical barrier" to T-cell infiltration by maintaining a microenvironment characterized by hypoxia, acidosis, and high interstitial pressure between cancer cells and immune cells, and this environment contributes to a pro-tumorigenic and immunosuppressive TME, among other effects (Schaaf et al. 2018).
Therefore, based on the in-depth knowledge of tumors and TME, chronic inflammation, coagulation status, and nutritional status have been considered as one of the important factors affecting the development, metastasis, and prognosis of malignant tumors, including prostate cancer. Up to now, several studies have included peripheral blood inflammatory indicators, coagulation indicators, and nutritional status indicators in the analysis of prostate cancer diagnosis (Masuda et al. 2021; Wang et al. 2022), metastasis (Salciccia et al. 2022) and prognosis (Yin et al. 2016; Li et al. 2020). However, previous studies have mostly been limited to a few indicators, such as NLR, PLR, SII, PNI, while not delving into the role of these indicators in prostate cancer bone metastasis. Therefore, in this study, in order to better investigate the relationship between peripheral blood biomarkers and prostate cancer bone metastasis, some commonly used inflammatory and nutritional indicators such as NLR, PLR, LMR, SII, PNI and some recent studies that point to a close correlation with tumor development such as HRR (Chi et al. 2022), HALP (Xu et al. 2023), RDW (Huang et al. 2021), AFR (Barone et al. 2022) were included in the study. Spearman correlation analysis was performed on the features not included in the model modeling, with the aim of mining these data more comprehensively and building accurate predictive models.
The results of this study showed that Gleason score, T stage, N stage, PSA and ALP were independent risk factors for bone metastasis of prostate cancer, while other indicators such as inflammatory and nutritional indicators, such as fibrinogen, NLR, LMR, PNI, HRR, HALP, and AFP were only statistically significant in Univariate LR and not in the multivariate analysis. However, Spearman correlation analysis in this study showed that fibrinogen, NLR, PLR, and SII had a low positive correlation with prostate cancer bone metastasis (P < 0.05); ALB, LY, HGB, BMI, LMR, PNI, HRR, HALP, and AFR had a low negative correlation with prostate cancer bone metastasis (P < 0.05). This suggests that we need further understanding of these inflammatory and nutritional indicators to detect prostate cancer bone metastases earlier in the clinic.
The machine learning models in this study showed good predictive performance, with KNN (AUC = 0.9390, 95% CI: 0.8760 ~ 1), RF (AUC = 0.9290, 95% CI:0.8718 ~ 0.9861), NB (AUC = 0.9268, 95% CI: 0.8615 ~ 0.9920) and LR (AUC = 0.9212, 95% CI: 0.8506 ~ 0.9917) had the best discrimination, while KNN performed, in combination, significantly better than the other models, and therefore, it was used as the optimal prediction model in this study.
There are some limitations of this study. First, this study is a retrospective study, which is influenced by selection bias and still needs prospective validation; second, our data is a single-center source, and the findings need to be tested in a multi-center and larger regional context; third, the interpretability of the models, except for LR, is insufficient to reflect the how and why of decision making.
In summary, this study developed a prediction model with good accuracy for prostate cancer patients by machine learning algorithms and found a low correlation between indicators such as inflammation, nutrition and bone metastasis.