Development and validation of a prognostic scoring model for mortality risk stratification in patients with recurrent or metastatic gastric carcinoma

Abstract

Background Survival times differ among patients with advanced gastric carcinoma. A precise and universal prognostic evaluation strategy has not yet been established. The current study aimed to construct a prognostic scoring model for mortality risk stratification in patients with advanced gastric carcinoma. Methods Patients with advanced gastric carcinoma from two hospitals (development and validation cohort) were included. Cox proportional hazards regression analysis was conducted to identify independent risk factors for survival. A prognostic nomogram model was developed using R statistics and validated both in bootstrap and external cohort. The concordance index and calibration curves were plotted to determine the discrimination and calibration of the model, respectively. The nomogram score and a simplified scoring system were developed to stratify patients in the two cohorts. Results Development and validation cohort was comprised of 401 and 214 gastric cancer patients, respectively. Mucinous or non-mucinous histology, ECOG score, bone metastasis, ascites, hemoglobin concentration, serum albumin level, lactate dehydrogenase level, carcinoembryonic antigen level, and chemotherapy were finally incorporated into prognostic nomogram. The concordance indices were 0.689 (95% CI: 0.664 ~ 0.714) and 0.673 (95% CI: 0.632 ~ 0.714) for bootstrap and external validation. 100 and 200 were set as the cut-off values of nomogram score, patients in development cohort were stratified into low-, intermediate- and high-risk groups with median overall survival time 15.8 (95% CI: 12.2 ~ 19.5), 8.4 (95% CI: 6.7 ~ 10.2), and 3.9 (95% CI: 2.7 ~ 5.2) months, respectively; the cut-off values also worked well in validation cohort with different survival time in subgroups. A simplified model was also established and showed good consistency with the nomogram scoring model in both of development and validation cohorts. Conclusion The prognostic scoring model and its simplified surrogate can be used as tools for mortality risk stratification in patients with advanced gastric carcinoma.

Machine learning models for predicting one-year survival in patients with metastatic gastric cancer who experienced upfront radical gastrectomy

“…Here, this issue is addressed by repeating the holdout CV five times, so the performance is unbiased in the local cohort, as measured by AUROC, recall, and precision. All the AUROCs of the three models exceed 0.75, which is higher than those in the previous reports ( Koo et al, 2011 ; Custodio et al, 2017 ; Ma et al, 2021 ). In addition, our earlier work selected the patients with AGC receiving first-line chemotherapy from the same local cohort and compared the performances of seven survival prediction models on it, showing the best model having an AUROC of 0.60 ( Xu et al, 2021 ).…”

“…A Korean single-center study constructed a score-based model with an accuracy of 0.58 ( Koo et al, 2011 ), then externally validated another three models ( Chau et al, 2004 ; Lee et al, 2007 ; Kim et al, 2008 ), and showed similar performances in the same population ( Koo et al, 2011 ). Our previous work also developed a score-based model in a Chinese cohort with a c-index of 0.67 ( Ma et al, 2021 ). Meanwhile, we validated seven published models ( Chau et al, 2004 ; Lee et al, 2007 ; Kim et al, 2008 ; Koo et al, 2011 ; Takahari et al, 2014 ; Wang et al, 2016 ; Kim et al, 2020 ) in a Chinese population, and the results showed that the area under receiver operating characteristic (AUROC) curves was only about 0.60 ( Xu et al, 2021 ).…”

“…The traditional prognostic model is frequently built by the logistic or Cox regression analysis on the basis of the well-known clinical and pathological variables, for example, performance status ( Chau et al, 2004 ; Lee et al, 2007 ; Kim et al, 2008 ; Takahari et al, 2014 ), tumor differentiation ( Custodio et al, 2017 ; Kim et al, 2020 ), metastatic sites ( Kim et al, 2008 ; Takahari et al, 2014 ; Wang et al, 2016 ), and routine laboratory tests ( Koo et al, 2011 ; Custodio et al, 2017 ; Kim et al, 2020 ; Ma et al, 2021 ). The selection of candidate variables is typically guided by the clinical experience and previous literature.…”

A Nomogram for Predicting the Cancer-Specific Survival of Patients with Initially Diagnosed Metastatic Gastric Cancer

“…While not designed for patients initially diagnosed with metastatic GC, Ma et al built a nomogram model for mortality risk stratification in patients with advanced GC that included mucinous or non-mucinous histology, Eastern Cooperative Oncology Group score, bone metastasis, ascites, hemoglobin concentration, serum albumin level, lactate dehydrogenase level, carcinoembryonic antigen level, and chemotherapy treatment. 21 Hannah et al reported that the Viennese risk prediction score for Advanced Gastroesophageal carcinoma based on Alarm Symptoms could be used at the time of initial diagnosis of patients with metastatic gastro-esophageal cancer. The score system comprises five factors: stenosis in endoscopy, weight loss, HER2 positivity, dyspepsia, and ulcer or active bleeding; the model was applied not only to GC but also to esophageal cancer.…”

Treatment Outcome and Safety of the TCX Regimen for Advanced Gastric Cancer: A Prospective Cohort Study

“…In a meta-analysis that included 14,651 patients, increased pretreatment serum CEA levels nearly doubled the risk of mortality in patients with gastric cancer. 52 In a study on 615 Chinese patients with advanced gastric cancer, a CEA level of ≥8 ng/mL was associated with higher death risk compared to that of <8 ng/mL, and therefore, CEA has been incorporated into a prognostic scoring model for mortality risk stratification. 53 However, because other studies have indicated otherwise, [54][55][56][57][58] CEA is not considered an independent predictor for patients with gastric cancer, and curative approach should not be tailored based on CEA level.…”