Development and validation of a novel nomogram for predicting long-term survival in gallbladder cancer: a large population-based study using the SEER database

Introduction

Gallbladder cancer (GBC) is a relatively rare but highly aggressive malignancy of the biliary tract. Although it only accounts for a small fraction of cancer cases worldwide, its incidence has been rising over recent decades, especially among younger populations in the United States. The 5year overall survival (OS) remains dismally low at 5–15% (1,2). One major challenge is the disease’s silent progression: most patients already present with advanced‑stage disease, and fewer than 35% of cases are amenable to curativeintent resection at diagnosis (3). Even among those who undergo radical resection, the risk of recurrence stays alarmingly high between 46% and 61% and longterm survival outcomes have improved only marginally over time (4).

In recent years, several metaanalyses and systematic reviews have consolidated established prognostic factors for GBC, including lymph node status, T stage, invasion, tumor location, surgical margin status, and the survival benefits of adjuvant chemotherapy and radiotherapy (5,6). However, the American Joint Committee on Cancer (AJCC) tumor-node-metastasis (TNM) staging system-currently the clinical benchmark for prognostic assessment-still offers relatively coarse risk stratification and does not incorporate individuallevel factors such as treatment modalities or patient demographics. This leaves a clear need for more refined, individualized prediction tools (7). To address this limitation, a number of nomograms have been built using the Surveillance, Epidemiology and End Results (SEER) database to provide personalized survival estimates for GBC patients, with most models focusing on 1-, 3- and 5year outcomes (8,9). To our knowledge, although several SEER‑based nomograms for GBC already exist, the majority target 1-, 3- and 5year survival and do not formally quantify the incremental prognostic value using metrics like net reclassification improvement (NRI) and integrated discrimination improvement (IDI) (10,11). Moreover, long‑term predictions extending to 8 years are rarely reported. To fill these gaps, we extracted GBC cases from the SEER database (2000–2021) to develop a prognostic nomogram for 3-, 5- and 8‑year OS. The model’s performance was systematically evaluated using the C‑index, time‑dependent receiver operating characteristic (ROC), NRI, IDI, calibration curves and decision curve analysis (DCA) with the goal of providing a more comprehensive and clinically useful prognostic tool. We present this article in accordance with the TRIPOD reporting checklist (available at https://jgo.amegroups.com/article/view/10.21037/jgo-2026-0515/rc).

Methods

Data source and study population

This study extracted GBC patients’ data from 17 regions covered by the US Cancer Surveillance, Epidemiology and End Results (SEER) database from 2000 to 2021, totaling 3,892 cases were included for final analysis. Using a random sampling method, these subjects were divided into a training set and an internal validation set in a 7:3 ratio. (Figure 1). The internal validation was further reinforced by 500 bootstrap resamples to assess model stability.

Figure 1 Flowchart of data cleaning process for gallbladder cancer patients in the SEER database. SEER, Surveillance, Epidemiology, and End Results.

Ethics approval and consent to participate

This study did not involve direct interaction with human subjects nor the collection of any personally identifiable information. The research was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. As the study involved no human intervention, the requirement for informed consent was waived.

Inclusion criteria

(I) Primary GBC (ICD-O-3 site code C23.9); (II) histologically confirmed diagnosis; (III) available survival information (survival time and vital status).

Exclusion criteria

(I) Missing demographic variables (age, sex, race); (II) missing key clinicopathological variables (T stage, N stage, M stage, tumor grade); (III) missing treatment information (surgery, radiotherapy, chemotherapy coded as “unknown”); (IV) missing survival time or survival status; (V) prior or synchronous other primary malignancies.

Although patients with incomplete key variables were excluded to ensure the integrity of the multivariable regression analysis, the proportion of excluded cases was relatively modest. Notably, baseline demographic and clinicopathological characteristics were well balanced between the training and internal validation cohorts, with no significant differences observed (all P>0.05, Table 1). These findings suggest that the exclusion of missing data is unlikely to have introduced substantial selection bias, and the analytical cohort remains adequately representative of the general GBC population within the SEER registry.

Variables and definitions

Candidate predictor variables were selected based on prior literature and clinical relevance. They included demographic characteristics [age at diagnosis, sex, race, vital status (alive/dead), follow‑up time], tumor characteristics (AJCC grade, T stage, N stage, M stage, tumor size), and treatment variables (surgery, radiotherapy, chemotherapy). Importantly, vital status and follow‑up time were used solely to define the survival endpoint (i.e., the survival object in the Cox model) and for descriptive baseline comparisons; they were not entered as predictors in the regression models. Race was categorized as White, Black, and other (including Asian, Pacific Islander, American Indian/Alaska Native). T stage, N stage, and M stage were defined according to the AJCC 8th edition criteria. Tumor size was measured as the maximum dimension of the primary tumor (in millimeters) and was treated as a continuous variable in the Cox model. Surgery, radiotherapy, and chemotherapy were recorded as binary variables (yes/no). Follow‑up time was measured in months and survival status was recorded as either alive or dead.

The full cohort was randomly split into a training set (70%) and an internal validation set (30%). The log‑rank test was used to assess differences in survival curves between the two groups. To quantify predictive accuracy and reliability of the established model, the concordance index (C‑index) was calculated in both the training and internal validation cohorts; a higher C‑index indicates better discriminative ability and model robustness. Regression coefficients from the final Cox proportional hazards model were then used to construct a prognostic index (PI) for each patient. The PI was computed as a linear combination of the selected predictors, each weighted by its corresponding regression coefficient. Patients were subsequently divided into low‑risk and high‑risk groups using the median PI value as the cut‑off. This risk stratification approach was further validated via Kaplan‑Meier survival analysis and log‑rank tests in both the training and internal validation sets.

We plotted the nomogram, NRI, IDI, ROC, and DCA curves. Specifically, the nomogram was drawn to illustrate the predicted 3-, 5- and 8‑year survival rates of the training set population. Using the AJCC tumor staging system as a reference, the IDI and NRI were calculated in both the training and validation sets to quantify the incremental prognostic value of our new prediction model. Time‑dependent ROC curves, DCA, and calibration curves were used to evaluate predictive accuracy, clinical net benefit, and calibration performance.

Statistical analysis

All statistical analyses and graphical visualizations were carried out using R software (version 4.2.3, R Foundation for Statistical Computing, Vienna, Austria). The log‑rank test was applied to compare survival differences between subgroups. A prognostic nomogram was built based on the Cox regression model in the training set. The model’s discriminative ability was assessed by the C‑index in both the training and internal validation cohorts, along with ROC curve analysis. Bootstrapping with 500 repetitions was performed for internal validation of calibration and optimism correction. According to the PI stratification, Kaplan‑Meier survival curves for 3-, 5- and 8‑year endpoints were plotted. NRI and IDI were calculated to compare the proposed model with the conventional AJCC staging system. Calibration curves were constructed to evaluate the agreement between predicted and actual survival probabilities. DCA was also performed to assess the clinical net benefit and utility of the nomogram relative to the AJCC staging system. A two‑sided P value <0.05 was considered statistically significant.

Results

Patient characteristics

A total of 3,892 eligible GBC patients diagnosed between 2000 and 2021 were included in this study and divided into training and validation sets at a ratio of 7:3. Baseline data comparison indicated no significant difference between the groups (P>0.05), as shown in Table 1.

Development of the prognostic nomogram

Based on the results of multivariable Cox regression analysis, independent prognostic factors (including age, tumor grade, T stage, N stage, M stage, surgery, and radiotherapy) were incorporated to construct a prognostic model. Subsequently, a novel nomogram was established for individualized prediction of 3-, 5- and 8-year OS probabilities in GBC. Each factor was assigned a point score on a scale of 0 to 100. The total points corresponded to predicted survival probabilities ranging from 10% to 80% at each time point. The total points from all variables were used to estimate individualized survival probabilities. Higher total scores were associated with progressively lower survival rates (Figure 2). The full Cox model coefficients, hazard ratios, and baseline survival estimates are provided in Table 2, enabling complete reproducibility of the prediction formula.

Figure 2 Nomogram for predicting 3-, 5- and 8-year overall survival probability in patients with GBC. The nomogram integrates key clinical variables to generate a total point score, which is then used to estimate individualized survival rates. AJCC, American Joint Committee on Cancer; GBC, gallbladder cancer.

Using the regression coefficients derived from the final Cox model, a PI was calculated for each patient. The PI values ranged from 0.67 to 4.34 with a median of 1.83 in the train set and 0.62 to 4.31 with a median of 1.81 in the internal validation set. Patients were stratified into low-risk and high-risk groups based on the median PI as the cut-off point. Kaplan-Meier survival analysis demonstrated that patients in the low-risk group had significantly better OS compared with those in the high-risk group (log-rank test, P<0.001). The survival curves were clearly separated between the two groups throughout the follow-up period, confirming that the nomogram-based PI could effectively stratify patients into distinct prognostic subgroups (Figure 3).

Figure 3 Kaplan-Meier survival curves according to the PI. Patients were divided into low-risk (blue line) and high-risk (red line) groups using the median PI (1.81) as the cut-off. PI, prognostic index.

Discrimination performance of the model

The discriminative ability was evaluated using the C-index and time-dependent ROC curves. In the training cohort, the C-index of the nomogram was 0.708 [95% confidence interval (CI): 0.670–0.745]. In the internal validation cohort, the C-index was 0.623 (95% CI: 0.583–0.662), indicating moderate discrimination.

In the training set, the area under the curve (AUC) values were 0.752, 0.757 and 0.763 for 3-, 5- and 8-year survival, respectively. In the internal validation set, the corresponding AUC values were 0.758, 0.767 and 0.771, further confirming the satisfactory discriminative capacity of the model (Figure 4).

Figure 4 ROC curves of the nomogram for predicting 3-, 5- and 8-year OS in GBC. (A) ROC curves of the training set; (B) ROC curves of the validation set. AUC, area under the curve; GBC, gallbladder cancer; OS, overall survival; PI, prognostic index; ROC, receiver operating characteristic.

Incremental prognostic value assessed by NRI and IDI

The NRI was calculated to quantify the improvement in risk reclassification of the newly developed nomogram compared with the conventional AJCC staging system. As shown in Table 3, the nomogram yielded significant NRI values across all three time points in both the training and internal validation cohorts.

For 3-year OS, the NRI was 0.322 (95% CI: 0.18–0.45, P<0.001) in the training set and 0.460 (95% CI: 0.32–0.57, P<0.001) in the internal validation set. At 5 years, the NRI was 0.267 (95% CI: 0.13–0.40, P<0.001) in the training set and 0.467 (95% CI: 0.31–0.60, P<0.001) in the validation set. For 8-year survival, the training set NRI was 0.259 (95% CI: 0.11–0.38, P<0.001), and the internal validation set NRI was 0.494 (95% CI: 0.29–0.59, P<0.001). These results indicate that the nomogram significantly improved risk stratification compared with the AJCC system at all time points, with particularly strong reclassification performance in the validation cohort.

The IDI was calculated to evaluate the incremental predictive value of the nomogram over the conventional staging system. In the training set, the nomogram yielded IDI values of 0.036, 0.030 and 0.027 for 3-, 5- and 8-year OS (P<0.001). Similar significant improvements were observed in the internal validation set with corresponding IDI values of 0.058, 0.044 and 0.041 (P<0.001). These results indicated that the nomogram significantly improved risk stratification compared with the conventional staging system across all follow-up time points (Figure 5).

Figure 5 NRI analysis of the prognostic nomogram versus the standard AJCC staging system in the training and validation cohorts. (A) Training cohort; (B) validation cohort. From left to right, the panels represent NRI analyses at 3-, 5-, 8-year follow-up, respectively. The x-axis shows predicted probabilities from the standard AJCC staging model and the y-axis shows predicted probabilities from the new combined nomogram. Points are color-coded by outcome status (Case: events; Control: non-events; Censored: censored observations). The solid diagonal line indicates no net reclassification, while the dashed lines denote the boundaries for predefined risk category shifts. All NRI analyses were performed using the nricens package with 200 bootstrap resamples to assess the improvement in risk reclassification performance of the nomogram compared with the AJCC staging system. AJCC, American Joint Committee on Cancer; NRI, net reclassification improvement.

Calibration performance of the model

Calibration of the nomogram was evaluated using bootstrap resampling (500 repetitions). Calibration curves for 3-, 5- and 8-year OS were plotted in both the training and internal validation cohorts. The curves were close to the 45° diagonal line, with small prediction errors, indicating satisfactory agreement between predicted and actual survival probabilities (Figure 6).

Figure 6 Calibration curves of the nomogram for predicting overall survival. Calibration curves for 3, 5 and 8-year overall survival are shown in the training set (A) and validation set (B). The solid line represents the ideal reference line and the dashed line represents the observed nomogram-predicted probabilities. Closer alignment between the two lines indicates better calibration performance.

Clinical utility evaluated by DCA

Across all three time points (3-, 5- and 8-year survival), the DCA curves showed that the nomogram consistently yielded a higher net benefit than both the “treat all” and “treat none” strategies across a wide range of threshold probabilities. The pattern of clinical benefit was similar in the internal validation cohort, confirming that the model’s utility was not driven by overfitting. To further compare the clinical practicality of the nomogram with the conventional AJCC staging system, we performed DCA for 3-, 5- and 8-year survival. In both the training and validation cohorts, the full-variable nomogram showed a significantly higher net benefit than the AJCC model at most threshold probabilities. These findings indicated that the nomogram could guide more precise risk-stratified interventions, avoiding unnecessary overtreatment while ensuring timely care for high-risk patients (Figure 7).

Figure 7 Decision curve analysis of the prediction models. (A,B) Compare the full-variable model with the AJCC model; (C,D) the performance of the single prediction model. The black solid line indicates the net benefit of the established model, the horizontal line represents the “treat none” strategy and the dashed line represents the “treat all” strategy. In the training cohort (A,C), the model curves highly overlapped with the “treat all” reference line, suggesting that the net benefit of the model was similar to that of the universal intervention strategy in the training dataset. In the validation cohort (B,D), the model curves were obviously higher than the two reference strategies within a wide range of threshold probabilities, which confirms the reliable clinical decision-making value of these models. AJCC, American Joint Committee on Cancer.

Discussion

Comparison with existing SEER‑based GBC nomograms

Several SEER‑derived GBC nomograms have been published in recent years. Feng et al. developed a prognostic model for patients undergoing radical surgery, reporting a C‑index of 0.737 and an NRI of 0.79–0.82 for 1-, 3- and 6‑year cancerspecific survival (12). Liu and Huang constructed a postoperative nomogram with good discrimination and performed external validation using a Chinese cohort (13,14). Our model differs from and complements these studies in three important respects. First, unlike the CSS focus of Feng et al., our model predicts OS is an endpoint more directly relevant to allcause mortality and patient counseling. Second, we extended the prediction horizon to 8 years for OS, whereas most existing models stop at 5 years (15,16). Third, we included both surgical and non‑surgical patients, which makes our model more representative of the real‑world GBC spectrum.

Prognostic factors and biological plausibility

The independent predictors in our model—age, T/N/M stage, tumor grade, surgery and radiotherapy—are all well established in the literature. Two recent large‑scale metaanalyses quantified the pooled HRs for these factors: lymph node metastasis [hazard ratio (HR) ≈2.0], T3/T4 stage (HR ≈2.4), and positive margins (HR ≈2.7) (5). Although margin status is not available in the SEER database, the moderate performance of our model suggests that the included variables capture most of the prognostic information relevant for long‑term survival. The inclusion of treatment variables is particularly important, because SEERbased nomograms that omit them (15) may underestimate the effect of therapeutic interventions. Dong et al. recently demonstrated that neoadjuvant therapy can improve OS in patients with nonmetastatic stage III/IV GBC and is an independent prognostic factor (17). Yang et al. and Jiang et al. investigated lymph node involvement in elderly GBC patients (≥65 years) and found that after adjustment, OS was compromised for those with nodal involvement (HR =2.238, P<0.01), and the presence of more than two metastatic lymph nodes was associated with even worse survival (HR =3.305, P<0.01) (18,19). These findings support the inclusion of both lymph node status and age in our model. Additionally, a machine learning study by Wang et al. using SEER data identified independent risk factors for distant metastasis in T1/T2 GBC and developed a nomogram for DM prediction, further validating the importance of earlystage risk stratification (20).

Risk stratification using the PI

Beyond the nomogram itself, we derived a PI as a continuous variable from the Cox model. Using the median PI as a cutoff, we stratified patients into high‑risk and low‑risk groups. Kaplan‑Meier survival curves showed a clear and statistically significant separation between the two groups in both the training and internal validation sets (P<0.001), confirming the robust risk‑stratification ability of the nomogram‑based PI. The ability to dichotomize patients into distinct prognostic categories enhances the clinical utility of our model, allowing clinicians to quickly identify high‑risk patients who may require more aggressive surveillance or adjuvant therapy, while low‑risk patients could be spared from overtreatment. This type of risk classification has been shown to improve clinical decisionmaking in GBC management (19).

Interpretation of NRI and IDI

NRI and IDI have become standard metrics for quantifying the incremental prognostic value of a new model over a reference standard (20,21). In our internal validation set, NRI values reached 0.460, 0.467, and 0.494 for 3‑, 5‑, and 8‑year survival, meaning that more than half of the patients were correctly reclassified by our nomogram compared with AJCC staging. Notably, the NRI values were slightly higher in the validation set than in the training set (e.g., 0.46 vs. 0.32 at 3 years), which argues against overfitting.

The IDI values, although modest (e.g., 0.041 at 8 years in the validation set), were all statistically significant (P<0.001). As Pencina et al. noted, given the low baseline probability of 8year survival in GBC (approximately 10–15%), this absolute IDI of 0.041 translates into a meaningful relative improvement in the model’s ability to correctly assign risk probabilities. Even small IDI values can be clinically meaningful when the baseline event rate is low, as in GBC (22). The persistence of significant NRI and IDI at 8 years indicates that the model’s added prognostic value is not limited to short‑term followup (23). This is consistent with findings from other SEERbased nomogram studies that reported significant NRI improvements over TNM staging (24).

Clinical utility and DCA

DCA has become the standard approach for assessing whether a prediction model leads to better clinical decisions than alternative strategies (25). Our DCA showed that the nomogram provides a net benefit superior to the AJCC system across a wide range of threshold probabilities. This means that using the nomogram to guide adjuvant therapy—for example, offering chemoradiotherapy to patients identified as high‑risk would theoretically improve outcomes without increasing unnecessary interventions. However, it must be emphasized that this is a prognostic, not a predictive, model; thus, it cannot directly inform the choice of specific treatments. Its value lies in risk communication and identification of extreme- risk patients who may be candidates for more intensive monitoring or enrollment in clinical trials.

Recent SEER‑based studies have confirmed that adjuvant chemotherapy confers survival benefits in resected GBC patients, particularly in those with advancedstage disease (26). Wang et al. conducted an updated retrospective cohort analysis of 2,782 patients from SEER and demonstrated that the adjuvant chemotherapy cohort exhibited a significant survival advantage compared with the nonadjuvant chemotherapy cohort, with age at diagnosis, grade, AJCC T stage, AJCC N stage, and adjuvant chemotherapy identified as independent prognostic factors (20). Similarly, Li et al. compared different treatment models for advanced GBC using SEER data and found that radical surgery combined with chemoradiotherapy maximally improved cancerspecific survival (27). Our nomogram could effectively identify patients who are most likely to benefit from such intensive treatment regimens, but these decisions should always be made in a multidisciplinary setting.

Long‑term predictions (8‑year survival)

Most existing GBC nomograms stop at 5year predictions (28). However, GBC patients who survive beyond 5 years still face a nonnegligible risk of late recurrence and competing mortality (29). Li et al. reported that early recurrence after radical resection occurs in 46–61% of GBC patients and is a major determinant of longterm survival (4). Our study provides 8‑year predictions and, more importantly, shows that the model’s discriminative ability remains stable from 3 to 8 years (AUC 0.745–0.752 in the internal validation set). This stability is a key strength, as it indicates that the baseline factors captured by the nomogram retain their prognostic power over an extended follow‑up period without rapid decay. To our knowledge, such a long‑term evaluation has rarely been reported in SEER‑based GBC nomogram studies. Furthermore, recent studies have highlighted the importance of long‑term surveillance in GBC patients. Romatoski et al. reported that after propensity score matching, oncologic resection demonstrated improved OS compared with cholecystectomy alone at 1‑year (69.2% vs. 47.2%), 3‑year (42.8% vs. 21.1%), and 5‑year (37.5% vs. 17.4%), underscoring the need for longterm outcome tracking (2). Our 8‑year prediction capability aligns with this clinical need.

Limitations

Several limitations should be acknowledged. First, this is a retrospective study using the SEER database, which is subject to selection bias and missing data. A considerable proportion of patients were excluded due to incomplete records, and this may have introduced bias toward a healthier or more completely documented population. We did not perform multiple imputation, which could have mitigated some of this bias. Second, the SEER database lacks many clinically important variables such as Eastern Cooperative Oncology Group performance status, surgical margin status (R0 vs. R1/R2), presence of residual disease, recurrence patterns, detailed chemotherapy regimens and dosing, radiation fields, and molecular biomarkers [e.g., human epidermal growth factor receptor 2 (HER2), programmed death-ligand 1 (PD‑L1), microsatellite instability (MSI) status, fibroblast growth factor receptor (FGFR) fusions]. These factors are known to affect prognosis and could have improved model performance if available. Third, although we performed internal validation with bootstrapping, we did not have an external independent cohort, which limits the generalizability of our findings. External validation in other populations (e.g., Asian cohorts, European registries) is urgently needed. Fourth, the C‑index in the internal validation set was only 0.623, indicating moderate discrimination; thus, the model should be used with caution and not as a sole basis for critical clinical decisions. Fifth, as with all SEER‑based studies, coding errors and treatment information (especially chemotherapy) may be underreported or misclassified. Despite these limitations, our nomogram offers a simple, user‑friendly tool for long‑term risk estimation that outperforms the AJCC staging system alone.

Conclusions

The nomogram provides additional prognostic information on top of TNM staging. While its predictive accuracy is only moderate (AUC around 0.76–0.77 in the internal validation set, C-index 0.62), it may still serve as a complementary tool for individualized risk communication in clinical practice, particularly for identifying patients with extremely high or low risk. Future studies with external validation and inclusion of molecular data are warranted to refine the model.

Acknowledgments

This study was conducted using oncology population data from the Surveillance, Epidemiology and End Results (SEER) database.

Footnote

Reporting Checklist: The authors have completed the TRIPOD reporting checklist. Available at https://jgo.amegroups.com/article/view/10.21037/jgo-2026-0515/rc

Peer Review File: Available at https://jgo.amegroups.com/article/view/10.21037/jgo-2026-0515/prf

Funding: This work was supported by the District-level Scientific Research Project of Medical and Health Institutions in Longhua District, Shenzhen (No. 2025017).

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://jgo.amegroups.com/article/view/10.21037/jgo-2026-0515/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 research was conducted in accordance with the Declaration of Helsinki and its subsequent amendments.

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. Hu Q, Tong X, Wan D, et al. The incidence of gallbladder carcinoma is increasing in the younger U.S. population: a SEER-based study. HPB (Oxford) 2024;26:1200-2.
2. Romatoski KS, Chung SH, Sawhney V, et al. Factors Associated With Surgical Management in Gallbladder Cancer-A Surveillance, Epidemiology, and End Results Medicare-Based Study. J Surg Res 2024;304:9-18. [Crossref] [PubMed]
3. Hu X, Zeng D, Wen N, et al. Prognostic factors in gallbladder cancer: a comprehensive systematic review and meta-analysis. Hepatobiliary Surg Nutr 2025;14:374-97. [Crossref] [PubMed]
4. Li Q, Li N, Gao Q, et al. The clinical impact of early recurrence and its recurrence patterns in patients with gallbladder carcinoma after radical resection. Eur J Surg Oncol 2023;49:106959. [Crossref] [PubMed]
5. Lv TR, Wang JK, Li FY, et al. Prognostic factors for resected cases with gallbladder carcinoma: a systematic review and meta-analysis. Int J Surg 2024;110:4342-55. [Crossref] [PubMed]
6. Adsay NV, Bagci P, Tajiri T, et al. Pathologic staging of pancreatic, ampullary, biliary, and gallbladder cancers: pitfalls and practical limitations of the current AJCC/UICC TNM staging system and opportunities for improvement. Semin Diagn Pathol 2012;29:127-41. [Crossref] [PubMed]
7. Liu Y, Song T. Development and validation of a postoperative nomogram for predicting survival in gallbladder cancer patients: insights from SEER and external cohorts. J Clin Oncol 2025;43:e16314.
8. Liu Y, Zhu K, Tian X, et al. Individualized Prediction of Postoperative Survival in Gallbladder Cancer: A Nomogram Based on SEER Data and External Validation. Cancers (Basel) 2025;17:1919. [Crossref] [PubMed]
9. Wang Z, Cheng Y, Seaberg EC, et al. Quantifying diagnostic accuracy improvement of new biomarkers for competing risk outcomes. Biostatistics 2020;23:666-82. [Crossref] [PubMed]
10. Che WQ, Li YJ, Tsang CK, et al. How to use the Surveillance, Epidemiology, and End Results (SEER) data: research design and methodology. Mil Med Res 2023;10:50. [Crossref] [PubMed]
11. Xu Y, Zheng X, Li Y, et al. Exploring patient medication adherence and data mining methods in clinical big data: A contemporary review. J Evid Based Med 2023;16:342-75. [Crossref] [PubMed]
12. Feng Y, Yang J, Wang A, et al. A prognostic model and novel risk classification system for radical gallbladder cancer surgery: A population-based study and external validation. Heliyon 2024;10:e35551. [Crossref] [PubMed]
13. Huang J, Qiu Y, Bai X, He X. Lymph node involvement is associated with overall survival for elderly patients with non‑metastatic gallbladder adenocarcinoma. Front Surg 2024;11:1414870. [Crossref] [PubMed]
14. Liu R, Zhang C, Shen Y, et al. Establishment and validation of a novel prognostic nomogram for gallbladder cancer patients. Eur J Med Res 2025;30:331. [Crossref] [PubMed]
15. Zhou N, Zhang K, Qiao B, et al. Personalized risk prediction of mortality and rehospitalization for heart failure in patients undergoing mitral valve repair surgery. Front Cardiovasc Med 2024;11:1470987. [Crossref] [PubMed]
16. Guo Z, Zhang Z, Liu L, et al. Machine Learning Algorithm for Predicting Distant Metastasis of T1 and T2 Gallbladder Cancer Based on SEER Database. Bioengineering (Basel) 2024;11:927. [Crossref] [PubMed]
17. Dong J, Zhu Z. Efficacy of neoadjuvant therapy and lymph node dissection in advanced gallbladder cancer without distant metastases: a SEER database analysis. Front Oncol 2024;14:1511583. [Crossref] [PubMed]
19. Jiang S, Zhang J, Zhang L, et al. A novel nomogram based on log odds of positive lymph nodes to predict survival for non-metastatic gallbladder adenocarcinoma after surgery. Sci Rep 2022;12:16466. [Crossref] [PubMed]
20. Wang Y, Kong Y, Yang Q, et al. Survival benefit of adjuvant chemotherapy in patients with resected gallbladder adenocarcinoma: An updated retrospective cohort analysis. Eur J Surg Oncol 2024;50:108047. [Crossref] [PubMed]
21. Steyerberg EW, Vergouwe Y. Towards better clinical prediction models: seven steps for development and validation. Eur Heart J 2024;35:1925-31.
22. Pencina MJ, D'Agostino RB Sr, Steyerberg EW. Extensions of net reclassification improvement calculations to measure usefulness of new biomarkers. Stat Med 2011;30:11-21. [Crossref] [PubMed]
23. Liu F, Qin Y, Li L, et al. Effects of radiotherapy, chemotherapy, and chemoradiotherapy on survival outcomes in patients with gallbladder carcinoma: a real population-based study. Transl Cancer Res 2025;14:3627-41. [Crossref] [PubMed]
24. Jin H, Du D, Xie Y, et al. The role of marital status in gallbladder cancer: a real-world competing risk analysis. BMC Gastroenterol 2024;24:276. [Crossref] [PubMed]
25. Xu X, He M, Wang H, et al. Development and validation of a prognostic nomogram for gallbladder cancer patients after surgery. BMC Gastroenterol 2022;22:200. [Crossref] [PubMed]
26. Pirenne S, Manzano-Núñez F, Loriot A, et al. Spatial transcriptomics profiling of gallbladder adenocarcinoma: a detailed two-case study of progression from precursor lesions to cancer. BMC Cancer 2024;24:1025. [Crossref] [PubMed]
27. Li R, Chen X, Wang B, et al. Comparison of treatment models for single primary advanced gallbladder cancer. Front Immunol 2024;15:1500091. [Crossref] [PubMed]
28. Siwakoti K, Harmon C, Al-Obaidi M, et al. Association of frailty with health-related quality of life and survival among older adults with prostate cancer. J Geriatr Oncol 2024;15:101812. [Crossref] [PubMed]
29. Chen C, Wang J, Kang M, et al. Identification of a novel MEF2C::SS18L1 fusion in childhood acute B-lymphoblastic leukemia. J Cancer Res Clin Oncol 2024;150:314. [Crossref] [PubMed]