Risk factors and predictive nomograms for early mortality in patients with thyroid cancer lung metastasis based on the SEER database and a Chinese population study

Introduction

The most common endocrine system malignancy is thyroid cancer (TC) (1,2), and its annual incidence continues to increase (3). Distant metastases, which primarily occur in the lung, affect approximately 15% of patients (4,5) and significantly impact patient outcomes and survival (6). Caring for these patients is often challenging and requires substantial resources, emphasizing the need for effective models to predict patient outcomes.

For cancer patients, the term “early death” lacks established meaning, whereas in research and clinical trials, it is often defined as death shortly after diagnosis or treatment (7,8). Based on earlier research classifications, we characterized early mortality from TC as death within three months after the first diagnosis (9,10). We defined early death as survival ≤3 months based on clinical relevance and methodological consistency. This critical post-diagnosis window reflects initial treatment tolerance and disease aggressiveness in metastatic cancer. The cutoff aligns with SEER-based studies of other cancers (9-11), ensuring comparability. Although TC generally has a good prognosis, its course accelerates after metastasis, particularly in anaplastic or radioactive iodine-refractory (RAI-R) cases. A shorter-term cutoff enables prompt identification of the highest-risk patients for aggressive intervention or early palliative care, supporting timely clinical decision-making. Patients with thyroid cancer lung metastasis (TCLM) tend to have poorer nutritional and physical health than those with nonmetastatic TC (12). TCLM also increases the risk of treatment-related complications, including chest discomfort, malignant pleural effusion, pneumonia, dyspnea, and malnutrition (13). These complications may result in early mortality, which emphasizes the importance of evaluating clinicopathological features, therapeutic efficacy, safety, and patient health when developing treatment regimens for patients with TCLM (14).

The early mortality rate among individuals with TCLM remains high, primarily due to insufficient therapeutic treatment options. Radioiodine-131 therapy is the primary treatment option for these patients (15); however, resistance to radioactive iodine-131 significantly affects patient survival rates. Thus, attempts to modify and improve TCLM treatments have been made in recent years, with locoregional excision of lung metastases (LMs) offering improved survival benefits to patients (16,17). Small-molecule tyrosine kinase inhibitors (TKIs) can effectively control the disease and increase progression-free survival (PFS) (18,19). In a phase 3 clinical trial, lenvatinib, donafenib, and sorafenib, which are licensed to treat RAI-R TC, showed considerably greater response rates and PFS than the placebo (20-22). However, the survival rate for TCLM remains low due to its diagnosis usually occurring at later disease stages. Retrospective studies have demonstrated that early detection of asymptomatic distant metastases improves the long-term survival of patients with TC (23). Thus, identifying high-risk TC patients with LM is critical for improving their outcomes.

Although tremendous progress has been made in studying the molecular mechanisms underlying TCLM (24,25), these studies have only advanced our understanding of the involved pathways, and translation into therapeutic applications remains elusive. Most basic studies cannot accurately evaluate the outcomes of patients with TCLM in clinical practice. Therefore, a simple and easy-to-use model is urgently needed to accurately assess the outcomes of patients with TCLM.

Studies on predicting early death in TC patients are scarce (26,27), and no studies have attempted to identify relevant risk factors in individuals with TCLM. In this study, data from the Surveillance, Epidemiology, and End Results (SEER) database (28) and a Chinese cohort were analyzed to identify risk factors for early death in patients with TCLM and construct predictive nomograms (29). Nomograms are valuable tools for visualizing multivariable prediction models in oncology, aiding prognostic prediction and individualized treatment decisions (30). Their reliability hinges on rigorous development and validation. As nomogram studies proliferate, systematic reviews—such as Antonini et al.’s analysis of models predicting pathological complete response to neoadjuvant chemotherapy in early breast cancer (31)—highlight key methodological steps: variable selection, performance evaluation, and validation of clinical utility. These reviews stress that robust internal and external validation is essential for generalizability (32). In developing our nomogram to predict early death in TCLM, we have adhered to such established methodological frameworks to enhance credibility and applicability (33). These technologies assist physicians in identifying high-risk patients, developing tailored treatment regimens quickly, extending patient life expectancy, improving patient quality of life (QoL), and decreasing the associated cost burden on society and families. We present this article in accordance with the TRIPOD reporting checklist (available at https://gs.amegroups.com/article/view/10.21037/gs-2025-328/rc).

Methods

Information source

We conducted a retrospective study utilizing data from the SEER database, which provides survival, clinicopathological, and demographic information for cancer patients. Specifically, we selected data from 17 registries (2000–2019) from the SEER Research Plus database, published in November 2022, for our investigation. The SEER*Stat program (version 8.4.3) was used to retrieve patient data related to TC from the SEER database. Because cancer information needs to be reported in every state in the United States, the data in the SEER database do not require patient consent for inclusion. Participant information is confidential. We have got permission to access the research data file in the SEER program and analysis of the data does not require informed consent from patients.

Patient selection

Patients with a TC code indicative of a diagnosis of TCLM (C73.0–C73.9) between 2010 and 2015 whose disease was histopathologically confirmed were included in this study. The patients were classified according to the International Classification of Diseases for Oncology, Third Edition (ICD-O-3). Some patients were excluded because they did not meet specific requirements. The exclusion criteria were as follows: (I) death identified via a death certificate or autopsy; (II) diagnosis not based on histology; (III) Tis, T0, or an unknown disease stage; (IV) unknown distant metastases; (V) incomplete follow-up; and (VI) anonymized variable information. A total of 945 patients with TCLM were included and analyzed in our study. Patients were randomly categorized into two groups, the training group (n=661, 70%) and the internal validation group (n=284, 30%), to guarantee the accuracy of the study outcomes. For external validation, we included the data of 141 patients with TCLM from Nantong First People’s Hospital and the Affiliated Huaian No. 1 People’s Hospital of Nanjing Medical University. Figure 1 shows a flowchart of our patient selection procedure. The initial histological or cytological diagnosis of TCLM served as the basis for computing the survival time. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study was approved by the Ethics Committee of Nantong First People’s Hospital and the Affiliated Huaian No. 1 People’s Hospital of Nanjing Medical University (approval No. YW-2024-015-06), and individual consent for this retrospective analysis was waived.

Figure 1 Flowchart of patient eligibility and cohort division in forecasting ACED and CSED. ACED, all-cause early death; CSED, cancer-specific early death.

Several factors were extracted from the SEER database and the data of the Chinese population in this study, including baseline information (age, marital status, sex, and race), tumor characteristics [histology, tumor (T) stage, node (N) stage, grade, bone metastases, liver metastases, brain metastases, metastatic type, tumor size, cancer-specific (CS) extension, and multifocality], treatment data (surgery, radiotherapy, and chemotherapy), cause of death (all-cause or cancer-specific), and survival time. Two types of thyroid radiotherapy were included: external and RAI. The former treats anaplastic thyroid cancer (ATC), whereas the latter treats differentiated thyroid cancer (DTC). Organ transfer data, which are gathered annually, were also recorded from the two cohorts.

Statistical analysis

Patients from the SEER database were randomly divided into a training cohort (70%) and an internal validation cohort (30%). An additional external validation cohort consisted of 141 patients from Chinese hospitals. Baseline characteristics across the three cohorts were compared using the Pearson chi-square test. To identify risk factors for all-cause early death (ACED) and cancer-specific early death (CSED), we first performed univariable logistic regression analyses. Variables with a significance level of P<0.05 in the univariable analysis were subsequently included in a multivariable logistic regression analysis with backward stepwise selection. The final independent risk factors derived from the multivariable analysis were used to construct nomograms for predicting ACED and CSED probabilities. The models’ discriminating power was evaluated using receiver operating characteristic (ROC) curve analysis, and the area under the curve (AUC) was reported. Calibration curves were used to assess the agreement between expected probability and observed results. Decision curve analysis (DCA) (34) was used to calculate the clinical net benefit of the nomograms at various threshold probabilities. Only the ACED nomogram was investigated for external validation since the external cohort lacked accurate cause-of-death information for CSED. Its performance was similarly evaluated with ROC analysis, calibration curves, and DCA. All statistical analyses were performed using R software (version 4.3.0) and SPSS (version 29.0). A two-sided P value <0.05 indicated statistical significance.

Handling of missing data and sensitivity analysis

Data entries identified as ‘Unknown’, ‘Blank’, or ‘N/A’ in the SEER database, along with incomplete cases from our institutional records, were treated as missing. To address this, we applied a complete-case analysis approach: during patient selection, individuals with missing information on key variables (e.g., distant metastasis status or incomplete follow-up) were excluded, as outlined in Figure 1. This ensured that all analyzed patients had complete data for all covariates used in nomogram construction across the training, internal validation, and external validation sets, thereby avoiding potential biases associated with data imputation.

To evaluate potential selection bias, a sensitivity analysis was performed comparing available demographic characteristics (age and sex) between SEER patients included in the final cohort and those excluded due to missing data. No significant differences were observed (P>0.05), indicating that the exclusion had a limited impact on the representativeness of the study population.

Results

Patient characteristics: clinical and demographic features

This study initially collected the data of all patients with TCLM identified in the SEER database. Following the application of the exclusion criteria, the data of 945 patients were ultimately included, among whom 636 (67.3%) experienced ACED and 335 (35.4%) experienced CSED. Most early deaths occurred in patients over age 70 years (38.5%), in women (50.8%), and in white individuals (76.7%). Papillary thyroid cancer (PTC) was the most common cancer among individuals who experienced early death, accounting for 59.9% of patients who experienced early death, followed by follicular thyroid cancer (FTC) (13.5%). The most common T stage for the tumors was T4 (44.0%). The second most prevalent organ for distant metastases after the lungs was bone (21.7%), followed by the liver (5.8%) and brain (3.1%). Although relatively few patients underwent chemotherapy (21.0%), more than half of the patients underwent radiotherapy or surgery (56.6% and 56.5%). To externally validate our nomograms, the data of 141 patients from Nantong First People’s Hospital and the Affiliated Huaian No. 1 People’s Hospital of Nanjing Medical University were included in this study. Like the SEER patients, most patients in the external validation cohort (44.7%) were over 70 years old, almost 51.1% were female, and among those with distant metastases, bone accounted for the second most common metastatic site (23.4%). Tables 1,2 display the clinicopathological and demographic features of patients with TCLM in the SEER cohort and the Chinese cohort, respectively.

Identification of risk factors associated with premature death

Following randomization, the training cohort consisted of 661 patients, whereas the internal validation set consisted of 284 individuals. There were no significant differences between the two groups terms of relevant variables (P>0.05), except for race. The notable disparity in racial distribution may have resulted from the inclusion of only Chinese individuals in our external validation cohort. Most patients in the Chinese cohort (44.0%) had PTC, similar to the patients in the SEER cohort, and the most common T stage was T4 (50.35%). Finally, the second most prevalent location of metastases was in bone tissue (23.4%). Table 2 displays more specific baseline data for these two cohorts.

Univariable logistic regression analysis demonstrated that age, histological type, grade, T stage, N stage, type of surgery, radiotherapy, brain metastasis, liver metastasis, CS extension, and multifocality were significantly different between patients who did and did not experience ACED. The variables from the univariable logistic regression analysis were included in the multivariable logistic analysis, the results of which showed that grade, type of surgery, radiation, chemotherapy, brain metastases, and CS extension were independent factors for ACED in patients with TCLM. In the CSED analysis, univariable logistic regression analysis revealed that age, sex, grade, type of surgery, radiotherapy, chemotherapy, and bone metastasis were risk factors in patients with TCLM. Multivariate logistic regression analysis suggested that all seven of these variables were independent factors for CSED. The outcomes of the univariable and multivariable logic analyses are shown in Tables 3,4.

Construction of the nomograms

We developed two nomograms based on the independent risk factors identified via multivariable logistic regression analysis to predict the likelihood of early mortality from cancer or other causes in patients with TCLMs. A score was assigned for each risk factor, and the sum of the scores for all the prognostic factors was used to determine the final score. These scores were used to construct nomograms for supporting clinical decision-making and assessing patient outcomes (Figure 2A,2B). A vertical line was drawn down from the score line to the probability line (with values ranging from 0.1 to 0.9) to indicate the likelihood of early death. Notably, not all of the scores have corresponding probability values. Analysis of the nomograms indicated that the most significant predictive factor for early mortality was the GX-G4 grading system.

Figure 2 Nomograms to predict early mortality in patients with lung metastases from thyroid cancer [all-cause mortality in (A) and cancer-specific mortality in (B)]. CS extension, cancer-specific extension; gETE, gross extrathyroidal extension; mETE, minimal extrathyroidal extension; TT, total thyroidectomy.

Assessment of the nomograms

ACED and CSED were both well predicted by the nomograms according to the ROC curve analysis. Specifically, AUCs of the two nomograms in the training set were 0.912 [95% confidence interval (CI): 0.889–0.931] and 0.732 (95% CI: 0.691–0.776), respectively, as shown in Figure 3A,3B, whereas Figure 3C,3D show the ROC curves in the validation set, with AUCs for predicting ACED and CSED of 0.904 (95% CI: 0.872–0.938) and 0.746 (95% CI: 0.692–0.789), respectively. Figure 4A-4D shows a great match between the projected probabilities and the actual observed values in the training and validation sets, as evidenced by the calibration curves of both modal plots being close to the diagonal. We created two nomograms to predict the likelihood of early death from cancer or other causes in patients with TCLM based on the independent risk factors identified by the multivariable logistic regression analysis. Moreover, DCA revealed that compared with the individual risk factors, the nomograms offered greater net benefits in predicting ACED and CSED in both the training and internal validation sets and demonstrated the therapeutic utility of the nomograms (Figure 5). Taken together, these findings indicate that the nomograms have the potential to be very useful in predicting early death in patients with TCLM.

Figure 3 ROC curves in the training cohort (A,B) and internal validation cohort (C,D) for differentiating nomograms in predicting early mortality [all-cause (A,C) and cancer-specific (B,D)]. Data on the blue line represent the cutoff value (specificity, sensitivity), and the expression of the AUC is the AUC (95% CI). (A) AUC =0.912 (95% CI: 0.889–0.931); (B) AUC =0.732 (95% CI: 0.691–0.776); (C) AUC =0.904 (95% CI: 0.872–0.938); (D) AUC =0.746 (95% CI: 0.692–0.789). AUC, area under the curve; CI, confidence interval; ROC, receiver operating characteristic.

Figure 4 Calibration curves for assessing the calibration of the nomograms to predict early death (all-cause and cancer-specific) in the training cohort (A,B) and the internal validation cohort (C,D).

Figure 5 DCA was performed on nomograms to predict early death (all-cause and cancer-specific) in the training cohort (A,B) and internal validation cohort (C,D). DCA, decision curve analysis.

To preserve the study’s neutrality and rigor, external verification was restricted to the ACED nomograms because many patients’ relatives were unable to provide accurate information about the actual cause of death. The AUC of the nomogram in the external validation set (Figure 6A) was 0.781 (95% CI: 0.742–0.833), demonstrating that the nomogram has good predictive power with external data. Furthermore, the calibration curve (Figure 6B) demonstrated that the nomogram predictions remained consistent with the actual patient outcomes. Moreover, the results of DCA revealed that the model provided excellent net benefit while offering clinical applicability (Figure 6C).

Figure 6 Validation in the Chinese population. ROC curve (A) for the external validation of the ACED nomogram. Calibration curve (B) for the external validation of the ACED nomogram, and the DCA curve (C) for the external validation of the ACED nomogram. Data on the blue line represent the cutoff value (specificity, sensitivity), and the expression of the AUC is the AUC (95% CI). AUC =0.781 (95% CI: 0.742–0.833). ACED, all-cause early death; AUC, area under the curve; CI, confidence interval; DCA, decision curve analysis; ROC, receiver operating characteristic.

Discussion

Despite advances in the treatment of TC, including targeted drugs, advanced surgical procedures, and ablative therapy, the distant metastasis rates for DTC are 10%, those for medullary thyroid cancer (MTC) are 5% to 10%, and those for ATC are 50% (35-37). Patients with TC are more likely to die if they develop LMs. TCLM is more prevalent than bone, brain, or liver metastases (38). The discovery of relevant prognostic indicators is essential to identify people at greater risk of premature death, and these patients should be given greater priority to improve outcomes and clinical treatment. Numerous studies have investigated early mortality in people with metastatic cancers (39,40), but few studies have investigated premature death in patients with TCLMs. In this study, we sought to close this gap by identifying predictive markers and creating nomograms to predict early death in TCLM patients. To our knowledge, this study is the first of its kind. The conclusions of this study will help in the formulation of appropriate treatment plans for TC patients and in minimizing the risk of premature death.

Among the 945 people in the SEER cohort, 309 did not die early, and cancer-related causes accounted for 335 of the 636 early deaths. Early tumor treatment is critical, particularly for patients with TC that has metastasized to the lungs. Furthermore, the substantial increase in cancer-related mortality reported in the literature implies that people with TCLMs are more likely to die as a result of their disease. These findings underscore the need to treat each patient individually and regularly monitor the course of their disease. Notably, a prior study revealed that among patients with advanced TC who died within a year of initial diagnosis, the proportion whose death was attributable to cancer was only 68.0% (41). Additional research is needed to fully understand the reasons for premature mortality in patients with TCLM.

After reviewing data from individuals with metastatic lung cancer, we discovered numerous features that were independently related to early death from any cause, including cancer. According to our findings, patients younger than 55 years of age and with grade IV disease were more likely to experience CSED. In contrast, we found that the probability of experiencing CSED was reduced in patients who received radiotherapy and chemotherapy. These findings emphasize the need for early detection and therapy for TCLM patients.

Several treatments, including radiotherapy, chemotherapy, and surgery, can help to reduce the risk of CSED during treatment for TCLMs. Options for managing LMs include pulmonary wedge resection, pulmonary segmentectomy, lobectomy, radiofrequency ablation, and transcatheter arterial chemoembolization (13). In a study conducted by Moneke et al., 31 (72%) of 43 patients who underwent R0 metastasectomy had 5- and 10-year disease-specific survival rates of 100% and 77%, respectively, indicating that metastasectomy benefits RAI-R TC patients (17). Among all histological types of head and neck cancer, the 5-year overall survival rate after resection of metastatic tumors is approximately 50.0%, according to Yano et al.’s findings (16). Before surgery, the following should be considered: a lengthy disease-free period between the initial identification of TC and TCLM, a low visceral tumor load, and effective systemic treatment of lymph nodes and distant metastases. However, clinical data are currently insufficient to support partial pneumonectomy as the recommended course of treatment for TCLMs.

Chemotherapy is an important factor that affects the prognosis of patients with TCLMs. In the literature, 20% of RAI-R TC patients achieve a partial response with cisplatin chemotherapy (42). Thirteen of 32 patients with LMs from Hurthle cell TC received chemotherapy, and 90% of those who experienced CSED achieved local control of the disease (43). Clinicians should evaluate and interpret all available data to reduce the chances of early death.

Our data indicate that patients with TCLMs who are older than 70 years have a greater risk of early death. This finding may be due to the aging of the immune system and the co-occurrence of disorders such as type 2 diabetes, cardiovascular disease, or renal insufficiency, all of which constitute challenges in cancer therapy (44). Furthermore, a lack of social support and self-care for older cancer patients may impair their ability to cope with the psychological and physical stresses of cancer therapy (45). Therefore, a personalized treatment plan tailored to each TCLM patient’s unique needs and health status needs to be developed. Such a plan may require adjusting chemotherapy doses or schedules, considering options for treatment with fewer side effects, or giving more weight to medical assistance to control symptoms, improve QoL, and ultimately reduce the chance of early death (46,47). Optimizing clinical outcomes requires recognizing the unique issues experienced by older patients with TCLM and providing thorough, patient-centered care (48,49).

The ATC subtype has been significantly linked to an elevated risk of early mortality. Despite a variety of treatments, the prognosis for ATC is bleak, with 80% of patients dying within a year of diagnosis (50). However, except for patients with LMs, people with ATC have worse outcomes than patients with other subtypes of TC (51). ATC should be treated with chemotherapy or targeted therapy, as well as radiotherapy for cervical decompression (52). If the BRAF^V600E mutation is detected, a combination of darfenib and trametinib is advised (53). ATC is also related to a greater risk of LMs, possibly because many genes in the ATC lung metastatic microenvironment are involved in the extracellular matrix and angiogenesis, which increases the risk of early death in these patients (54).

Our findings suggest that patients with TC that metastasizes to the brain, liver, or bone are more likely to die early. This finding is similar to that of another study in which individuals with TC bone metastases had the highest survival rate, whereas patients with brain metastases had the poorest prognosis (55). Bone metastases are often less invasive than liver metastases or LMs, but they can cause discomfort, poor physical health, and bone-related adverse events (38). Furthermore, the restricted permeability of the blood‒brain barrier makes treating brain metastases in TC difficult (56). As a result, the joint presence of lung and brain metastases should be thoroughly investigated when choosing therapies for patients with TCLMs.

Early mortality has been observed in a variety of cancers, including breast, lung, colorectal, ATC, and liver cancer, and these cancers may be resistant to the three primary cancer treatment methods, namely, surgery, chemotherapy, and radiation (9,11,26,57,58). However, the specific explanation for this phenomenon remains uncertain. It is reasonable to believe that if the host is unable to properly regulate the inflammatory and stress-inducing circumstances associated with cancer, therapy, environmental or dietary variables, and persistent infections, cancer development may accelerate after treatment. Age is also a risk factor for early mortality from advanced liver, breast, colorectal, and lung cancers, as is histological grade (9,11,57,58). The three primary methods of cancer treatment (surgery, radiation, and chemotherapy) increase the chance of dying early from advanced colorectal, lung, and esophageal cancers (11,58,59). Brain metastasis increases the likelihood of early death from advanced colorectal cancer (58). Despite prompt cancer detection and the execution of staged therapy, the reason for premature mortality linked with disease progression in patients remains unknown. The stage and histological characteristics of the malignancy cannot fully explain the frequency of early mortality. Nevertheless, early death may occur when the patient cannot perform regular physiological tasks. This issue has historically been attributed to an inflammatory response caused by malignancy or treatment-induced damage.

Notably, mutations in the TP53 gene represent the most common genetic alterations observed in breast cancer, particularly in cases involving triple-negative and hormone receptor/human epidermal growth factor receptor 2 (Her-2) positive subtypes (60). These mutations have been associated with an increased risk of early death among these patients. Furthermore, research has indicated that miR-574-3p and LINC01003 can serve as reliable biomarkers for predicting early mortality in patients with cervical cancer (61). The roles of critical genes, noncoding RNAs, and signaling pathways that influence early death have been established. Nevertheless, further biological validation, tailored definitions of early mortality for various cancer types and treatments, and improved interdisciplinary collaboration are essential to gain a more comprehensive understanding of the complexities surrounding early death in cancer patients.

The pathological features of high-grade TC (particularly ATC), including cellular atypia, high proliferative index, necrosis, and vascular invasion, are strongly associated with aggressive behavior, early metastasis, and poor prognosis. High-grade tumors often harbor molecular alterations (62,63), such as TP53 and TERT promoter mutations, as well as the BRAF^V600E mutation, which drive disease progression and early death through mechanisms including epithelial-mesenchymal transition, angiogenesis, and immune evasion. Recent advances in the molecular mechanisms of TCLM (64,65), relating to key signaling pathways (e.g., PI3K/AKT, MAPK), non-coding RNAs (e.g., miR-451a, LINC01003), and the tumor microenvironment (e.g., extracellular matrix remodeling, immune cell infiltration) in promoting metastasis and early death. This integration helps bridge statistical associations with potential biological mechanisms, enhancing the depth and persuasiveness of the discussion. Lastly, the role of thyroid cytopathology reporting systems (e.g., the Bethesda System) and molecular profiling in risk assessment (66,67), emphasizes that future models should integrate histomorphological features with molecular biomarkers for more precise risk stratification.

Nomograms have emerged as valuable tools for providing customized prognostic information for TC (68). These tools allow visualization of predictions from quantitative data, assisting patients and physicians in making educated treatment decisions (69). In this study, we developed nomograms to predict premature death in individuals with TCLM through an extensive and excellent-quality population from the SEER database. While the nomograms showed good predictive performance in internal and external validation, their generalizability requires cautious interpretation. Our nomogram demonstrated superior predictive performance, with AUCs of 0.912 (training, 95% CI: 0.889–0.931) and 0.904 (internal validation, 95% CI: 0.872–0.938) for ACED. This exceeds most TC prediction models [e.g., Cui et al. (26): AUC =0.816; Wang et al. (68): AUC =0.853] and compares favorably with early death models for other cancers [e.g., Zhang et al. (39): AUC =0.892; Liu et al. (40): AUC =0.845]. Despite population and endpoint differences, these comparisons confirm our model’s excellent discriminatory power and support its clinical applicability. This nomogram offers practical utility in multidisciplinary management of TCLM. By incorporating readily available clinical parameters, it provides quantifiable risk assessment to guide team-based decisions. For high-risk patients, this may prompt earlier systemic therapy and integration of palliative care, while supporting aggressive local strategies in lower-risk cases. As a decision-support tool, it facilitates risk-stratified treatment selection, optimizing both resource allocation and patient outcomes. The relatively small external validation cohort (n=141) may limit evaluation robustness and pose overfitting risks. Although stepwise regression mitigated multicollinearity and validation in a Chinese cohort preliminarily supports applicability, we recommend further evaluation in larger, multi-center prospective studies. Future work should incorporate more diverse populations and additional clinical variables to improve model universality and clinical utility.

Our model’s calibration performance aligns with recent SEER-based nomograms predicting early death in other cancers. Zhang et al. (39) and Yang et al. (11) both reported strong calibration in their models for advanced liver cancer and lung cancer with brain metastases, respectively. Similarly, our nomograms showed calibration curves closely following the 45-degree line in internal and external validation, demonstrating comparable reliability in predictive accuracy. Our research focused on early mortality in individuals with TCLM; we mitigated the effect of multicollinearity between factors with stepwise regression, yielding a more credible predictive model. Our model is focused on early-risk prediction and to state that future research will employ time-to-event models to investigate dynamic risk factors associated with long-term survival. According to previous studies, the susceptibility of patients with TCLMs to early death is determined by their illness subtype (70). Understanding TC subtypes is critical for predicting the possibility of LM in patients. Bi et al. created a nomogram to predict survival in TCLM patients with the ATC subtype who underwent surgical excision of low-risk metastases (71). Clinicians can utilize these nomograms to predict whether a patient will develop LM, thus assisting in making better-informed treatment decisions (72).

Although this study offers insightful information, its limitations must be recognized:

• The lack of data on disease recurrence or secondary tumor sites in the SEER database limits our ability to analyze data from individuals with newly diagnosed TCLM.
• Selection bias may have been introduced by the exclusion of certain patients due to insufficient data. The primary focus of this study was to predict early death in patients with TCLMs. Because our findings were internally and externally validated with prospective studies, larger sample sizes are needed to further confirm the performance of the nomograms.
• Regression was performed via backward stepwise regression in the multifactor analysis.

According to the statistical significance of each predictor, this method iteratively reduces the scale of the variables starting from the predictors of all the models. However, because certain important variables may be excluded from the final model, this approach may result in the loss of raw data and predictive power. These biases may have influenced the validity of the analysis in this study.

Conclusions

Severe TCLM is associated with a high rate of early death. Using data from the SEER database and a Chinese population, we developed and tested predictive models for early death in patients with TCLMs. Our findings revealed that in addition to liver metastases, brain and bone metastases predict ACED and CSED in individuals with TCLM, respectively. We developed nomograms by incorporating multiple prognostic indicators that are easily obtained from medical records and that have significant implications for early death in patients with TCLM. Our predictive model is based on stepwise regression, which helps mitigate the effect of multicollinearity between the variables.

Acknowledgments

The authors are grateful to all the patients, researchers, and institutions participating in the SEER database and the Chinese population.

Footnote

Reporting Checklist: The authors have completed the TRIPOD reporting checklist. Available at https://gs.amegroups.com/article/view/10.21037/gs-2025-328/rc

Data Sharing Statement: Available at https://gs.amegroups.com/article/view/10.21037/gs-2025-328/dss

Peer Review File: Available at https://gs.amegroups.com/article/view/10.21037/gs-2025-328/prf

Funding: This study was supported by the Huai’an Basic Research Plan Program (No. HABL2023063). This study received the Young Investigator Award at the 19th Congress of the Asian Association of Endocrine Surgeons (AsAES), Seoul, Korea.

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://gs.amegroups.com/article/view/10.21037/gs-2025-328/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 Ethics Committee of Nantong First People’s Hospital and the Affiliated Huaian No. 1 People’s Hospital of Nanjing Medical University (approval No. YW-2024-015-06), and individual consent for this retrospective analysis was waived.

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. Siegel RL, Kratzer TB, Giaquinto AN, et al. Cancer statistics, 2025. CA Cancer J Clin 2025;75:10-45. [Crossref] [PubMed]
3. Megwalu UC, Moon PK. Thyroid Cancer Incidence and Mortality Trends in the United States: 2000-2018. Thyroid 2022;32:560-70. [Crossref] [PubMed]
4. Park H, Yang H, Heo J, et al. Long-Term Outcomes and Causes of Death among Medullary Thyroid Carcinoma Patients with Distant Metastases. Cancers (Basel) 2021;13:4670. [Crossref] [PubMed]
5. Reverter JL. Thyroid cancer. Med Clin (Barc) 2025;164:421-8. [Crossref] [PubMed]
6. Tahara M, Kiyota N, Hoff AO, et al. Impact of lung metastases on overall survival in the phase 3 SELECT study of lenvatinib in patients with radioiodine-refractory differentiated thyroid cancer. Eur J Cancer 2021;147:51-7. [Crossref] [PubMed]
7. Zhang L, Jing H, Tang S, et al. Benefits of Dynamic Nomogram Models for Elderly Diffuse Large B-Cell Lymphoma Patients' Early Death Prediction and Clinical Application. Mediators Inflamm 2023;2023:7874239. [Crossref] [PubMed]
8. Frontino G, Delvecchio M, Prudente S, et al. SID/SIEDP expert consensus on optimizing clinical strategies for early detection and management of wolfram syndrome. J Endocrinol Invest 2025;48:507-25. [Crossref] [PubMed]
9. Li Q, Sun T, Zhang Z. Early death prediction model for breast cancer with synchronous lung metastases: an analysis of the SEER database. Gland Surg 2024;13:1708-28. [Crossref] [PubMed]
10. Wu Y, Li T, Zhang R, et al. Establishment of nomogram of early death in elderly pancreatic cancer patients with liver metastasis. Discov Oncol 2025;16:333. [Crossref] [PubMed]
11. Yang F, Gao L, Wang Q, et al. Development and validation of nomograms to predict early death in non-small cell lung cancer patients with brain metastasis: a retrospective study in the SEER database. Transl Cancer Res 2023;12:473-89. [Crossref] [PubMed]
12. Gerull WD, Puri V, Kozower BD. The epidemiology and biology of pulmonary metastases. J Thorac Dis 2021;13:2585-9. [Crossref] [PubMed]
13. Nezami N, Khorshidi F, Mansur A, et al. Primary and Metastatic Lung Cancer: Rationale, Indications, and Outcomes of Thermal Ablation. Clin Lung Cancer 2023;24:389-400. [Crossref] [PubMed]
14. Li J, Shi W, Xiong J, et al. Toxicity spectrum and risk factors for chemo-immunotherapy in locally advanced or metastatic lung cancer. Clin Exp Immunol 2023;214:162-9. [Crossref] [PubMed]
15. Zhang S, Zhu M, Zhang H, et al. The Effect of Radioiodine Therapy on the Prognosis of Differentiated Thyroid Cancer with Lung Metastases. Biomedicines 2024;12:532. [Crossref] [PubMed]
16. Yano T, Haro A, Fukuyama S, et al. Chest surgery in patients with pulmonary metastasis from head and neck cancer. Kyobu Geka 2013;66:266-70.
17. Moneke I, Kaifi JT, Kloeser R, et al. Pulmonary metastasectomy for thyroid cancer as salvage therapy for radioactive iodine-refractory metastases. Eur J Cardiothorac Surg 2018;53:625-30. [Crossref] [PubMed]
18. Yu J, Liu Z, Su Y, et al. Tyrosine kinase inhibitors for radioiodine refractory differentiated thyroid cancer: A systematic review and meta-analysis. Clin Endocrinol (Oxf) 2024;100:379-88. [Crossref] [PubMed]
19. Porcelli T, Luongo C, Cerbone A, et al. Did selective kinase inhibitors change the management of patients with radioiodine-refractory thyroid cancer? Eur Thyroid J 2025;14:e240332. [Crossref] [PubMed]
20. Lin Y, Qin S, Li Z, et al. Apatinib vs Placebo in Patients With Locally Advanced or Metastatic, Radioactive Iodine-Refractory Differentiated Thyroid Cancer: The REALITY Randomized Clinical Trial. JAMA Oncol 2022;8:242-50. [Crossref] [PubMed]
21. Lin Y, Qin S, Yang H, et al. Multicenter Randomized Double-Blind Phase III Trial of Donafenib in Progressive Radioactive Iodine-Refractory Differentiated Thyroid Cancer. Clin Cancer Res 2023;29:2791-9. [Crossref] [PubMed]
22. Majidova N, Aghamaliyeva S, Guliyev M, et al. Sorafenib vs chemotherapy in the treatment of locally advanced or metastatic, radioactive iodine-refractory differentiated thyroid cancer: real-world data from Turkey. J Chemother 2025;37:735-40. [Crossref] [PubMed]
23. Talebpour M, Zabihi Mahmoudabadi H, Najjari K, et al. Distant Metastasis as the First Presentation of Thyroid Carcinoma. Indian J Otolaryngol Head Neck Surg 2023;75:1020-3. [Crossref] [PubMed]
24. Zhong Y, Yu F, Yang L, et al. HOXD9/miR-451a/PSMB8 axis is implicated in the regulation of cell proliferation and metastasis via PI3K/AKT signaling pathway in human anaplastic thyroid carcinoma. J Transl Med 2023;21:817. [Crossref] [PubMed]
25. Hu Y, Xu Z, Zhou D, et al. CXCR4 promotes migration, invasion, and epithelial-mesenchymal transition of papillary thyroid carcinoma by activating STAT3 signaling pathway. J Cancer Res Ther 2024;20:1241-50. [Crossref] [PubMed]
26. Cui H, Wang R, Zhao X, et al. Development and validation of a nomogram for predicting the early death of anaplastic thyroid cancer: a SEER population-based study. J Cancer Res Clin Oncol 2023;149:16001-13. [Crossref] [PubMed]
27. Tichanek F, Försti A, Liska V, et al. Early mortality critically impedes improvements in thyroid cancer survival through a half century. Eur J Endocrinol 2023;189:355-62. [Crossref] [PubMed]
28. Friedman S, Negoita S. History of the Surveillance, Epidemiology, and End Results (SEER) Program. J Natl Cancer Inst Monogr 2024;2024:105-9. [Crossref] [PubMed]
29. Balachandran VP, Gonen M, Smith JJ, et al. Nomograms in oncology: more than meets the eye. Lancet Oncol 2015;16:e173-80. [Crossref] [PubMed]
30. Tsai CY, Tian JH, Lee CC, et al. Building Dual AI Models and Nomograms Using Noninvasive Parameters for Aiding Male Bladder Outlet Obstruction Diagnosis and Minimizing the Need for Invasive Video-Urodynamic Studies: Development and Validation Study. J Med Internet Res 2024;26:e58599. [Crossref] [PubMed]
31. Antonini M, Pannain GD, Mattar A, et al. Systematic Review of Nomograms Used for Predicting Pathological Complete Response in Early Breast Cancer. Curr Oncol 2023;30:9168-80. [Crossref] [PubMed]
32. Riley RD, Ensor J, Snell KI, et al. External validation of clinical prediction models using big datasets from e-health records or IPD meta-analysis: opportunities and challenges. BMJ 2016;353:i3140. [Crossref] [PubMed]
33. Ma M, Gu W, Liang Y, et al. A novel model for predicting postoperative liver metastasis in R0 resected pancreatic neuroendocrine tumors: integrating computational pathology and deep learning-radiomics. J Transl Med 2024;22:768. [Crossref] [PubMed]
34. Vickers AJ, Holland F. Decision curve analysis to evaluate the clinical benefit of prediction models. Spine J 2021;21:1643-8. [Crossref] [PubMed]
35. Wells SA Jr, Asa SL, Dralle H, et al. Revised American Thyroid Association guidelines for the management of medullary thyroid carcinoma. Thyroid 2015;25:567-610. [Crossref] [PubMed]
36. Araque KA, Gubbi S, Klubo-Gwiezdzinska J. Updates on the Management of Thyroid Cancer. Horm Metab Res 2020;52:562-77. [Crossref] [PubMed]
37. Luis PO, Lucía MA, Hugo RC, et al. Differentiated Thyroid Carcinoma Long-Term Prognostic Factors. Int J Surg Oncol 2024;2024:1067447. [Crossref] [PubMed]
38. Zhong M, Khan FZ, He X, et al. Impact of Lung Metastasis versus Metastasis of Bone, Brain, or Liver on Overall Survival and Thyroid Cancer-Specific Survival of Thyroid Cancer Patients: A Population-Based Study. Cancers (Basel) 2022;14:3133. [Crossref] [PubMed]
39. Zhang H, Dong H, Pan Z, et al. Risk factors and predictive nomograms for early death of patients with pancreatic cancer liver metastasis: A large cohort study based on the SEER database and Chinese population. Front Oncol 2022;12:998445. [Crossref] [PubMed]
40. Liu S, Jia Y, Chai J, et al. A Predictive Model for the Early Death of Breast Cancer With Synchronous Liver Metastases: A Population-Based Study. Cancer Control 2023;30:10732748231202851. [Crossref] [PubMed]
41. Dueñas-Disotuar S, Piñar-Gutiérrez A, de Lara-Rodríguez I, et al. Bone metastasis in differentiated thyroid cancer: Spanish multicenter study of clinical characteristics, survival and prognostic factors. Front Endocrinol (Lausanne) 2024;15:1466245. [Crossref] [PubMed]
42. Hu Y, Zhou N, Wang R. Apatinib strengthens the anti-tumor effect of cisplatin in thyroid carcinoma through downregulating VEGFR2. J BUON 2021;26:613-9.
43. Besic N, Schwarzbartl-Pevec A, Vidergar-Kralj B, et al. Treatment and outcome of 32 patients with distant metastases of Hürthle cell thyroid carcinoma: a single-institution experience. BMC Cancer 2016;16:162. [Crossref] [PubMed]
44. Xu ML, Wang YY, Xie LP, et al. Liver metastases in thyroid cancer: epidemiology, risk stratification and survival outcomes in the immunotherapy era. Front Immunol 2025;16:1624181. [Crossref] [PubMed]
45. Sun X, Jiang Y, Wang J, et al. Effects of a mobile health intervention based on a multitheoretical model of health behavior change on anxiety and depression, fear of cancer progression, and quality of life in patients with differentiated thyroid cancer: A randomized controlled trial. BMC Med 2024;22:466. [Crossref] [PubMed]
46. Gonzalez-Ling A, Galindo Vázquez O, Espinoza Bello M, et al. Quality of life, anxiety, depression, and distress in patients with advanced and metastatic lung cancer. Palliat Support Care 2023;21:608-15. [Crossref] [PubMed]
47. Tefferi A, Vannucchi AM, Barbui T. Essential thrombocythemia: 2024 update on diagnosis, risk stratification, and management. Am J Hematol 2024;99:697-718. [Crossref] [PubMed]
48. Billingy NE, Tromp VNMF, van den Hurk CJG, et al. Health-Related Quality of Life and Survival in Metastasized Non-Small Cell Lung Cancer Patients with and without a Targetable Driver Mutation. Cancers (Basel) 2021;13:4282. [Crossref] [PubMed]
49. Matsumoto Y, Umemura S, Okizaki A, et al. Early specialized palliative care for patients with metastatic lung cancer receiving chemotherapy: a feasibility study of a nurse-led screening-triggered programme. Jpn J Clin Oncol 2022;52:375-82. [Crossref] [PubMed]
50. Bible KC, Kebebew E, Brierley J, et al. 2021 American Thyroid Association Guidelines for Management of Patients with Anaplastic Thyroid Cancer. Thyroid 2021;31:337-86. [Crossref] [PubMed]
51. Boucai L, Zafereo M, Cabanillas ME. Thyroid Cancer: A Review. JAMA 2024;331:425-35. [Crossref] [PubMed]
52. Are C, Shaha AR. Anaplastic thyroid carcinoma: biology, pathogenesis, prognostic factors, and treatment approaches. Ann Surg Oncol 2006;13:453-64. [Crossref] [PubMed]
53. Hamidi S, Dadu R, Zafereo ME, et al. Initial Management of BRAF V600E-Variant Anaplastic Thyroid Cancer: The FAST Multidisciplinary Group Consensus Statement. JAMA Oncol 2024;10:1264-71. [Crossref] [PubMed]
54. Restle D, Dux J, Li X, et al. Organ-specific heterogeneity in tumor-infiltrating immune cells and cancer antigen expression in primary and autologous metastatic lung adenocarcinoma. J Immunother Cancer 2023;11:e006609. [Crossref] [PubMed]
55. Liu WC, Li MP, Hong WY, et al. A practical dynamic nomogram model for predicting bone metastasis in patients with thyroid cancer. Front Endocrinol (Lausanne) 2023;14:1142796. [Crossref] [PubMed]
56. Wolff L, Steindl A, Popov P, et al. Clinical characteristics, treatment, and long-term outcome of patients with brain metastases from thyroid cancer. Clin Exp Metastasis 2023;40:217-26. [Crossref] [PubMed]
57. Zhang H, Du X, Dong H, et al. Risk factors and predictive nomograms for early death of patients with advanced hepatocellular carcinoma: a large retrospective study based on the SEER database. BMC Gastroenterol 2022;22:348. [Crossref] [PubMed]
58. Zhang Y, Zhang Z, Wei L, et al. Construction and validation of nomograms combined with novel machine learning algorithms to predict early death of patients with metastatic colorectal cancer. Front Public Health 2022;10:1008137. [Crossref] [PubMed]
59. Shi M, Zhai GQ. Models for Predicting Early Death in Patients With Stage IV Esophageal Cancer: A Surveillance, Epidemiology, and End Results-Based Cohort Study. Cancer Control 2022;29:10732748211072976. [Crossref] [PubMed]
60. Van Swearingen AED, Lee MR, Rogers LW, et al. Genomic and immune profiling of breast cancer brain metastases. Acta Neuropathol Commun 2025;13:99. [Crossref] [PubMed]
61. Cho O. Post-Radiotherapy Exosomal Non-Coding RNA and Hemograms for Early Death Prediction in Patients with Cervical Cancer. Int J Mol Sci 2023;25:126. [Crossref] [PubMed]
62. Lebbink CA, Links TP, Czarniecka A, et al. 2022 European Thyroid Association Guidelines for the management of pediatric thyroid nodules and differentiated thyroid carcinoma. Eur Thyroid J 2022;11:e220146. [Crossref] [PubMed]
63. Kitahara CM, Schneider AB. Epidemiology of Thyroid Cancer. Cancer Epidemiol Biomarkers Prev 2022;31:1284-97. [Crossref] [PubMed]
64. Patel J, Klopper J, Cottrill EE. Molecular diagnostics in the evaluation of thyroid nodules: Current use and prospective opportunities. Front Endocrinol (Lausanne) 2023;14:1101410. [Crossref] [PubMed]
65. Dell'Aquila M, Gravina C, Cocomazzi A, et al. A large series of hyalinizing trabecular tumors: Cytomorphology and ancillary techniques on fine needle aspiration. Cancer Cytopathol 2019;127:390-8. [Crossref] [PubMed]
66. Fiorentino V, Dell'Aquila M, Musarra T, et al. The Role of Cytology in the Diagnosis of Subcentimeter Thyroid Lesions. Diagnostics (Basel) 2021;11:1043. [Crossref] [PubMed]
67. Dell'Aquila M, Granitto A, Martini M, et al. PD-L1 and thyroid cytology: A possible diagnostic and prognostic marker. Cancer Cytopathol 2020;128:177-89. [Crossref] [PubMed]
68. Wang W, Liu J, Xu X, et al. A High-Quality Nomogram for Predicting Lung Metastasis in Newly Diagnosed Stage IV Thyroid Cancer: A Population-Based Study. Technol Cancer Res Treat 2023;22:15330338231167807. [Crossref] [PubMed]
69. Chen X, Zhao P, He Y, et al. Development and validation of survival nomograms for patients with anaplastic thyroid carcinoma: a SEER program-based study. Discov Oncol 2024;15:650. [Crossref] [PubMed]
70. Murayama D, Yamamoto Y, Matsui A, et al. Lung cavitation in patients with anaplastic thyroid cancer treated with lenvatinib. Gland Surg 2022;11:963-9. [Crossref] [PubMed]
71. Bi J, Zhang H. Nomogram predicts risk and prognostic factors for lung metastasis of anaplastic thyroid carcinoma: a retrospective study in the Surveillance Epidemiology and End Results (SEER) database. Transl Cancer Res 2023;12:3547-64. [Crossref] [PubMed]
72. Zullo L, Filippiadis D, Hendriks LEL, et al. Lung metastases. Nat Rev Dis Primers 2025;11:60. [Crossref] [PubMed]