BM is a significant adverse prognostic factor in patients with lung adenocarcinoma. Many patients with BM lack specific clinical manifestations in the early stages, and metastasis is often incidentally detected during routine screening or investigation of elevated tumor markers. Unfortunately, by the time patients experience skeletal-related events, the optimal window for early treatment has passed [15]. To timely prevent skeletal-related events through bone-targeted therapy and reduce the toxicity of subsequent treatments, it is necessary to screen for metastasis in lung cancer patients at high risk of BM. Several studies have focused on the diagnosis and risk prediction factors of BM in NSCLC patients. Teng et al. utilized machine learning techniques to analyze serum markers closely associated with BM and constructed an early diagnostic model for bone metastasis in NSCLC patients [20]. Su et al. developed a clinical radiomics nomogram model to predict the occurrence of BM in LUAD patients during the course of the disease [21]. Compared to the study by Su et al, our study focuses on category T1 LUAD, which generally has a relatively better prognosis, and we selected baseline synchronous BM patients as the study population. A comprehensive nomogram model was constructed through univariate and multivariate logistic regression analyses of 10 clinical features and 13 CT image features to achieve early identification and diagnosis of bone metastasis in category T1 LUAD. Our study revealed that elevated ALP levels, elevated CEA levels, solid nodules, N2-3 in CT-reported N staging, and present of pleural effusion are risk factors for BM in category T1 LUAD.
CEA, initially identified in the bloodstream of colorectal cancer patients in 1965, serves as a widely utilized serum glycoprotein marker across various cancer types. In the context of NSCLC, CEA holds significance as a serum biomarker in clinical settings, facilitating risk stratification and potentially informing disease management strategies. Notably, several studies have indicated that CEA reflects the extent of disease progression and is an adverse prognostic factor in NSCLC patients [22, 23]. Prior investigation has delineated a close correlation between elevated levels of serum CEA or cerebrospinal fluid CEA and the occurrence of brain metastasis in lung cancer, attributed to CEA's ability to traverse the blood-brain barrier [24]. Additionally, serum CEA has demonstrated predictive utility for mediastinal lymph node metastasis in lung cancer patients [25], and analogous associations have been established between serum CEA expression and bone metastasis in NSCLC [26]. In our study, elevated CEA levels demonstrated a strong correlation in multivariate logistic regression, indicating its value in diagnosing BM in T1-stage LUAD. ALP is primarily derived from tissues such as the liver and bone and is an important indicator of bone metabolism. Elevations in serum ALP expression among lung cancer patients coincide with escalated osteoblast activity, accelerated bone formation, and heightened risk of BM due to aberrant bone metabolism mechanisms [27].
Subsolid nodules can be categorized into part-solid nodules (which include ground-glass opacity (GGO) and solid components) and pure GGO nodules (which lack solid components) [28]. GGO is a nonspecific radiological manifestation characterized by a hazy opacity that does not obscure underlying pulmonary vessels or bronchial structures. Lung cancers predominantly presenting as GGO are typically non-invasive or minimally invasive low-grade adenocarcinomas [29]. Compared to solid lesions, lung adenocarcinomas with GGO components exhibit higher survival rates [30]. In this study, among 111 patients with lung adenocarcinoma and bone metastasis, only 2 exhibited pulmonary lesions showing GGO. Pleural effusion is a common clinical issue associated with over 50 causes [31]. In cancer patients, malignant pleural effusion is often due to lymphatic obstruction and increased vascular permeability caused by pleural cancer or metastases [32, 33]. Malignant effusion typically indicates disseminated or advanced disease, with a shortened life expectancy. Lung cancer is a major cause of malignant pleural effusion, accounting for more than a third of cases. However, pleural effusion can also result from obstructive pneumonia, atelectasis, pulmonary embolism, hypoalbuminemia, among other causes [31]. In this study, pleural effusion on CT images in patients with lung adenocarcinoma is identified as a risk factor for bone metastasis and typically indicates a poor prognosis. Higher levels of lymph node involvement are associated with an increased incidence of BM in LUAD patients, consistent with previous research findings [34, 35].
Increasing evidence supports that nomograms provide a visual and easily interpretable method for predicting disease risk. While several models have been proposed to predict bone metastases (BM) in lung cancer, none have specifically addressed the risk of synchronous bone metastases in T1 stage lung adenocarcinoma. Compared to existing BM prediction models, our model presents several advantages. Firstly, we have visualized this predictive model as a nomogram and developed an interactive web interface (https://bmofluad.shinyapps.io/model/), significantly optimizing the computational process and enhancing its clinical usability. Secondly, our clinical model leverages preoperatively accessible clinical and CT features to predict the risk of bone metastases. Thirdly, our nomogram model consistently maintains an AUC value exceeding 0.90 even after internal cross-validation. Finally, we conducted a decision curve analysis (DCA) to assess the clinical utility of the nomogram, which demonstrated significant net benefits within a wide range of threshold, underscoring its considerable clinical value.
The nomogram model developed in this study aids clinicians in evaluating the risk of bone metastases in patients with T1 stage LUAD using preoperative data, thereby enhancing clinical decision-making. For high-risk patients, preventive interventions can be implemented to mitigate the incidence of skeletal-related events and improve patient outcomes., while avoiding unnecessary interventions and associated costs for low-risk patients. For example, if a patient is identified as high-risk prior to surgery, further imaging modalities such as ECT or PET may be recommended to confirm the presence of bone metastases and facilitate early targeted treatment. Conversely, for low-risk patients, we do not recommend routine screening in the absence of symptoms indicative of bone metastases. Furthermore, informing high-risk patients about the potential for extended hospitalizations and increased costs can improve their acceptance of medical decisions.
Our study exhibits several limitations. Firstly, this study was conducted with patients from a single medical institution, which presents challenges in mitigating selection and information biases inherent to single-center samples. The model has undergone only internal validation; thus, prospective multi-center studies are necessary for external validation prior to clinical implementation. Secondly, the data were sourced from a tertiary hospital, and given that CEA fluid/serum level testing is not universally routine, the applicability of our novel web-based nomogram model may be restricted in regions where such testing is not standard. Additionally, the limited sample size in this study may introduce unavoidable biases. Future research will aim to increase the sample size to enhance the model's reliability and stability. Consequently, there is an urgent need for multi-center validation involving larger populations to provide robust evidence for future clinical applications. This will facilitate the refinement of the model and enable its integration into electronic health systems, thereby advancing clinical decision-making.