Construction and validation the first prognostic models of progression-free survival in gastric cancer patients after gastrectomy with deficient mismatch repair: Nomogram and three machine learning models approaches

Construction and validation the first prognostic models of progression-free survival in gastric cancer patients after gastrectomy with deficient mismatch repair: Nomogram and three machine learning models approaches | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Construction and validation the first prognostic models of progression-free survival in gastric cancer patients after gastrectomy with deficient mismatch repair: Nomogram and three machine learning models approaches Jinfeng Ma, Wenhua Cheng, Yifan Li This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-4639290/v1 This work is licensed under a CC BY 4.0 License Status: Published Journal Publication published 24 Jan, 2025 Read the published version in BMC Cancer → Version 1 posted 4 You are reading this latest preprint version Abstract Objective To assess the effectiveness of a machine learning framework and nomogram in predicting progression-free survival (PFS) post radical gastrectomy in patients with dMMR. Method An observational study conducted at Shanxi Cancer Hospital from 2002 to 2020 focused on developing and evaluating three machine learning models and nomogram to forecast PFS in patients undergoing radical gastrectomy for nonmetastatic gastric cancer with dMMR. Independent risk factors were identified using Cox regression analysis to develop the nomogram. The performance of the models was assessed through C-index, time receiver operating characteristic (T-ROC) curves, calibration curves, and decision curve analysis (DCA) curves in both training and validation cohorts. Subsequently, patients were categorized into high-risk and low-risk groups based on the nomogram's risk scores. Results Among the 582 patients studied, machine learning models exhibited higher c-index values compared to the nomogram. RSF demonstrated the highest c-index (0.968), followed by XG boosting (0.945), DST (0.924), the nomogram (0.808), and 8th TNM staging (0.757). Age, positive lymph nodes, neural invasion, and Ki67 were identified as key factors and integrated into the prognostic nomogram. Calibration and DCA curves provided evidence of the accuracy and clinical benefits of both machine learning and nomogram models. Conclusion Our study first successfully developed and validated machine learning and nomogram model based on clinical parameters for predicting 3-, 5-year PFS among dMMR gastric patients following gastrectomy. The nomogram exhibited a remarkable capability in identifying high-risk patients, furnishing clinicians with invaluable insights for postoperative surveillance and tailored therapeutic interventions. gastric carcinoma mismatch repair progression-free survival machine learning nomogram Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 Figure 8 Figure 9 Figure 10 Figure 11 Figure 12 Figure 13 Figure 14 Figure 15 Introduction Despite being a significant issue globally, gastric cancer has now become the fifth most prevalent form of cancer and the fourth most common cause of mortality [ 1 ] . Thanks to advancements in medical technology, there are now numerous effective treatments available for those diagnosed with this condition [ 2 – 6 ] . The prognosis for patients with gastric cancer has been greatly improved by various treatments such as endoscopy, surgery, radiotherapy, chemotherapy, and immunotherapy. One area of research interest is the investigation of tumors that have deficient mismatch repair (dMMR), also known as high-frequency microsatellite instability (MSI-H). These tumors lack mismatch repair proteins in tumor cell nuclei and/or have reduced MMR activity [ 7 ] . Immunohistochemistry (IHC) and microsatellite instability (MSI) testing can be used to assess the MMR status of tumor DNA. Research suggests that dMMR can enhance the immune response against tumor cells and hinder their growth in colorectal cancer, usually due to mutations or epigenetic alterations in DNA MMR genes [ 8 – 9 ] . However, the role of dMMR in gastric cancer is still unclear, making it challenging to establish its clinical significance [ 10 – 11 ] . Some studies suggest a link between dMMR status and improved long-term survival after gastrectomy, but recent clinical trials have contradicted this finding. It is evident that dMMR is not a standalone factor in the outcomes of gastric cancer patients. Despite progress in gastric cancer treatment, the exact role and clinical significance of dMMR in this cancer type remain unknown [ 12 , 7 ] . Further investigation is necessary to fully explore the potential of dMMR as a prognostic or predictive marker in gastric cancer, which could ultimately enhance treatment strategies and outcomes for patients. The beauty of artificial intelligence and machine learning is their ability to continuously evolve and refine themselves autonomously. These advanced technologies go beyond traditional methods, offering a new way to analyze data. Machine learning can uncover hidden connections between variables that may not be easily apparent [ 13 ] . In the field of medicine, machine learning is gaining popularity, especially in medical imaging [ 14 ] , precise cancer prognosis [ 15 ] , and predicting surgical complication [ 16 ] s. This study focused on creating accurate machine learning models to predict the progression-free survival of patients with gastric carcinoma and deficient mismatch repair (dMMR). By utilizing machine learning, researchers aim to provide precise and efficient predictions of patient outcomes, leading to improved personalized treatment strategies. Artificial intelligence and machine learning have the potential to revolutionize the analysis and interpretation of medical data, promising a bright future for advancing healthcare practices. Patients and methods Data source and study population The study focused on patients who underwent radical gastrectomy for nonmetastatic gastric cancer at Shanxi Cancer Hospital between May 1, 2002, and December 31, 2020. Diagnosis of gastric carcinoma was confirmed through histological examination, and patients had comprehensive clinicopathological details available. Follow-up data was collected for patients with no severe organ damage, no prior history of other malignancies, and no unrelated causes of death. Exclusion criteria included systemic tumors, incomplete clinical information, palliative procedures, or non-gastric cancer cases. Tumor staging followed the AJCC 8th TNM classification. The study protocol was approved by the Ethics Committee of Shanxi Cancer Hospital and complied with the Declaration of Helsinki. Patient confidentiality was maintained through anonymization of data, and informed consent was obtained from all participants, along with written permission for publication and use of accompanying visuals. Refer to Figs. 1 and 2 for an overview of the research methodology. The research participants were randomly split into two distinct groups: a training cohort representing 70% of the patients, which was utilized for the development of machine learning algorithms, and a validation cohort encompassing the remaining 30% of patients, which served as the testing ground for the algorithms. Demographic details of the study population were presented through frequencies and percentages for categorical variables, while mean and standard deviation (SD) were used for continuous variables. Predictive variables Table 1 showcases all the variables integrated into the machine learning models. These variables encompassed a wide range of standard data categories, such as demographics, laboratory results, surgical specifics, tumor characteristics, proteomics data, and pathology reports. The researchers undertook an exhaustive analysis of various factors that significantly impacted the surgical outcomes. The investigated factors encompassed gender, age at surgery, presence of vascular and neural invasion, tumor stage (pT stage), number of positive lymph nodes, Lauren classification, maximum tumor diameter, type of gastrectomy undertaken, omentum metastasis status, surgical margin outcomes, complication severity as per the Clavien-Dindo classification, expression levels of specific biomarkers (including AE1/AE3, CK20, CDX-2, SATB-2, SYN, CGA, CD56, MLH1, PMS2, Her-2, MSH2, and MSH6), and progression-free survival (PFS). Follow-up protocol and outcome Patients included in this study had to meet specific criteria, which involved receiving a confirmed diagnosis of gastric cancer through histological examination and undergoing surgery aimed at curing the disease (R0-R1). The researchers explored various factors that could potentially impact the surgical outcome. In order to be eligible for the study, patients needed to meet certain requirements. These requirements included having a confirmed diagnosis of gastric cancer through histological examination and undergoing surgery with the goal of curing the disease (R0-R1). The researchers utilized electronic medical records and collaborated with oncologists to determine the duration of follow-up. Postoperative follow-up protocols varied depending on the institution/physician, but generally followed national and international guidelines. This included a 3-month postoperative outpatient visit in the first year, followed by appointments every 6 months for the next 3 years, and then annually for an additional 2 years. Follow-up examinations involved physical assessments and contrast-enhanced computed tomography scans of the chest, abdomen, and pelvis. The primary objective of the model prediction in this study was progression-free survival (PFS). PFS was defined as the period from surgery to disease recurrence (including both local and distant recurrences) or death from any cause. MMR Status Detected by Immunohistochemistry(IHC) The tissue specimens were initially preserved in a 10% formalin solution and then embedded in paraffin, following a standard protocol. Immunohistochemistry (IHC) was performed on the MMR protein using Ventana anti-MLH1, anti-PMS2, MSH2, and MSH6 mouse monoclonal antibodies from reputable suppliers. The staining process was carried out on the VENTANA Bench Mark ULTRA platform, with 2 µm sections of the specimens examined for the presence of MLH1, MSH2, MSH6, and PMS2 [ 17 ] . In order to evaluate the results, two senior pathologists assessed the staining patterns, with a third senior pathologist brought in to resolve any disagreements. A lack of staining in the tumor cell nuclei, alongside positive staining in adjacent normal cells such as fibroblasts and lymphocytes, was considered a negative assessment. The absence or presence of nuclear staining was used to distinguish between intact and lost signals. A tumor was classified as deficient in MMR protein expression if the cancerous epithelial cells lacked nuclear staining, while surrounding non-neoplastic cells exhibited positive staining. Conversely, if all four MMR proteins were present in the tumor cells, it was categorized as proficient MMR (pMMR). Any level of nuclear staining above the background level of cancer cells indicated MMR protein expression. In cases where one of the proteins was absent, the tumor was labeled as deficient MMR (dMMR). Machine learning models In the study, three survival machine learning algorithms - Random Survival Forests (RSF), Decision survival tree (DST), and Extreme Gradient Boosting (XG boosting) were used. The optimization of machine learning parameters involved a randomized search of various parameter settings, followed by testing with a ten-fold cross validation in order to maximize the concordance index (c-index). The parameters explored for each model and the final parameter combinations used for analysis can be found in Table 2 . Nomogram development This research sought to develop nomograms that can effectively predict progression-free survival (PFS) by incorporating variables identified through Cox regression analysis. To achieve this, the researchers conducted Cox regression analysis on a range of PFS-related parameters and repeated the process 10,000 times to ensure robustness. Model performance, fit, and validation assessments The model's performance was evaluated using discrimination, calibration, and decision curve analysis (DCA). Discrimination, which assesses the model's ability to differentiate between patients who will experience an event and those who will not, was measured using the c-index method by Uno et al [ 18 ] , for specific time points and by Harrell et al for the overall time assessment [ 19 ] . Calibration, which evaluates the model's ability to make accurate predictions close to actual events, was gauged using the Brier score. The Brier score calculates the mean square of the difference between predicted and observed probabilities, with a score of 0 indicating perfect calibration and 1 indicating the worst calibration [ 20 ] . DCA was used to determine if the new model added value by increasing the net benefit across a realistic range of threshold probabilities [ 21 ] . Furthermore, a sensitivity analysis was conducted to explore the relationship between dataset size and model performance, as the optimal amount of data needed to train a machine learning model is not yet established. All analyses and visualizations were performed using the Google Collaboratory Pro environment ( https://research.google.com/colaboratory ) with the Python 3.8.3 programming language ( https://www.python.org ) and the 'scikit-survival' package (v. 0.16.1; https://scikit-survival.readthedocs.io ). Statistical analysis We utilized descriptive statistics to succinctly encapsulate our data, portraying categorical variables as absolute numerical counts and continuous variables as mean and standard deviation (SD). Our results were elucidated through hazard ratios, 95% confidence intervals (CI), and corresponding P values. A significance threshold of 0.05 was employed to ascertain statistical significance. This comprehensive analysis was executed using Python 3.8.3, R software (version 4.3.2), and SPSS 25.0. Results Cohort characteristics A total of 582 patients were included in this study (Table 1 ). The incidence of gastric cancer in males was nearly five times higher than in females, regardless of whether they were in the training or validation cohort. The mean age of patients in the training and validation cohorts was 58.78 ± 8.929 and 58.47 ± 9.29 years, respectively. The mean maximum diameter of the tumor was 4.05 ± 2.42 cm in the training cohort and 4.63 ± 2.75 cm in the validation cohort. After a median follow-up period of 61 months (interquartile range 12–120), 86 (14.8%) patients experienced recurrence. The recurrence rates were 14.6% in the training cohort and 15.3% in the validation cohort. Overall survival and progression-free survival were similar in both groups. All variables were well balanced between the training and validation cohorts (P > 0.05). Patients were divided into training (n = 405) and validation (n = 177) cohorts, with both groups showing a balance in clinical and pathological variables, follow-up duration, and proteomic data (Refer to Fig. 3 for Kaplan-Meier curves showing progression-free survival in the training and validation sets). Machine learning model evaluation and validation Table 2 illustrates the c-index values for predicting 3-year and 5-year PFS using classical and machine learning prognostic models in the validation cohort. Among the machine learning models, RSF exhibited the highest discrimination with a c-index of 0.936, followed by XG boosting (0.904) and DST (0.884). The AUC values further confirmed the superior discriminative performance of all machine learning models, with RSF leading (0.968), followed by XG boosting (0.945) and DST (0.924). Given the potential changes in patients' status over time, we assessed the performance of the models at various time points and observed a decrease in performance over time. The c-index accuracy peaked at 60 months for all machine learning models, while the prediction error was minimized at this time point, especially for RSF and XG boosting(Fig. 4 ). The integrated Brier score, used to evaluate model calibration over a 60-month period, indicated good calibration with low integrated Brier scores for all models (all < 0.1; Table 2 ). To gain insights into the risk predictions provided by the models, we utilized SHAP values to extract global-level explanations of risk(Fig. 5 – 7 ). The top three important features for predicting PFS in gastric cancer with dMMR remained consistent across all models: number of positive lymph nodes, Ki67, and pT stage. Given the exceptional accuracy of the RSF model (Fig. 5 ), our analysis focused on this model. Importance scores of the top twenty factors were plotted to assess face validity, with variables higher on the graph considered more influential. The findings emphasized number of positive lymph nodes, Ki67, pT stage, Lauren classification, and maximum tumor diameter as the most discriminating factors for RSF. Notably, removing the number of positive lymph nodes from the model led to an average c-index drop of 0.06 points. Our study delved into the practicality of various models using Decision Curve Analysis (DCA) (Fig. 8 ) to assess their efficacy in guiding post-surgery care decisions, such as the choice between surveillance and adjuvant treatment. DCA visually showcases the clinical advantages of a model by mapping out a spectrum of risk thresholds (x-axis) and the net benefit of stratifying patients according to the model (y-axis) compared to assuming no relapse will occur in any patient. Our investigation revealed that DST models yielded lower benefits, while RSF models demonstrated the highest gains, consistently all machine learning models in both the training and validation cohorts (Fig. 8 ). Particularly notable was the performance of RSF models at risk thresholds between 0.2 and 0.5. The calibration curves depicted in Fig. 9 A and 9 B showcase the strong correlation between projected and observed outcomes of all machine learning model. (Fig. 9 ). Sensitivity analysis of machine learning model The optimal amount of training data required to build an effective machine learning predictive model remains an ongoing topic of debate. To address this issue, we conducted a sensitivity analysis to assess the impact of dataset size on model performance. By utilizing the learning curve plot, we were able to visualize how the addition of more training and validation data influenced the accuracy of various models. Our findings revealed that the accuracy scores varied as the size of the training samples changed for both the training and validation cohorts. Among the models considered, XG boosting consistently demonstrated the highest sensitivity analysis, outperforming RSF and DST (Fig. 10 ). Additionally, we observed that the sensitivity scores for the training cohort were notably higher than those for the validation cohort. Specifically, in the case of RSF, the sensitivity score ranged from 0.83 to 0.97 for the training cohort and from 0.52 to 0.71 for the validation cohort. Similarly, for DST, the sensitivity score ranged from 0.82 to 0.95 for the training cohort and from 0.65 to 0.77 for the validation cohort. In contrast, for XG boosting, the sensitivity score ranged from 0.90 to 1.00 for the training cohort and from 0.60 to 0.75 for the validation cohort. Moreover, as the size of the sample increased, we observed an enhancement in the predictive accuracy of the machine learning models. This suggests that incorporating more data leads to greater stability in the model's performance. Overall, our sensitivity analysis highlights the importance of dataset size in optimizing the accuracy and effectiveness of machine learning predictive models. Designing and Validating a Prediction nomogram Model for PFS A comprehensive analysis using Multivariate Cox regression was conducted to assess the factors influencing PFS in gastric cancer patients. The results from the study, which included variables such as age, positive lymph nodes, neural invasion, and Ki67, were particularly significant. This analysis was carried out on a group of 405 individuals with gastric cancer, and by incorporating these key factors into a nomogram model, accurate predictions of 3-year and 5-year PFS probabilities were made possible. The nomogram model accounts for various factors known to impact outcomes, allowing for precise estimation of a patient's PFS over these defined time periods. The effectiveness of the nomogram in forecasting positive outcomes for gastric cancer patients, considering variables affecting 3-year and 5-year PFS, is illustrated in Fig. 11 . The evaluation of the nomogram's performance related to PFS in the training group revealed a strong predictive capability with a C-index of 0.808 (95% CI: 0.769–0.847), surpassing the AJCC 8th edition TNM staging with a C-index of 0.757 (95%CI: 0.717–0.797). Calibration curves depicted in Fig. 12 A, 12 B, 12 C, and 12 D demonstrated a strong correlation between projected and observed outcomes, further validated internally and externally.The time-dependent ROC analysis during internal validation reaffirmed the nomogram's exceptional discriminatory abilities, with AUC values of 0.849 (95%CI: 0.705–0.872) for 3-year PFS and 0.846 (95%CI: 0.800–0.890) for 5-year PFS. External validation also yielded promising results, with AUC values of 0.830 (95%CI: 0.754–0.892) for 3-year PFS and 0.804 (95%CI: 0.747–0.928) for 5-year PFS (Fig. 13 A, 13 B). To determine the clinical utility of the nomogram, a decision analysis curve (DCA) was utilized to compare 5-year and 3-year PFS predictions between the nomogram and the AJCC 8th edition TNM staging. The internal validation C-index for the nomogram was superior at 0.808 (95%CI: 0.769–0.847) compared to the AJCC 8th edition TNM staging at 0.757 (95%CI: 0.717–0.797). Similarly, the external validation C-index for the nomogram outperformed the AJCC 8th edition TNM staging with a value of 0.755 (95%CI: 0.698–0.811) compared to 0.729 (95%CI: 0.675–0.783). The higher C-index values of the nomogram in both internal and external validations highlight its superior predictive accuracy over the AJCC 8th edition TNM staging (Fig. 14 A, 14 B, 14 C, 14 D). Risk Assessment in a Stratified System for Prediction of PFS based on nomogram By employing advanced nomogram modeling techniques, each patient underwent categorization and scoring to evaluate their risk of disease progression. The X-tile software was used in the initial cohort of 405 patients to determine the optimal cutoff value for PFS scores. Subsequent analysis using the log-rank test revealed significant differences in survival times across various risk groups. The prognostic nomogram was then applied to calculate overall scores, leading to the clear division of the 582 participants into low-risk and high-risk groups using a threshold value of 175.97. Visualization of the cohort in Fig. 15 A, 15 B, and 15 C demonstrated the effective separation of individuals based on their likelihood of disease progression. The low-risk group, comprising 316 patients from the training set and 124 patients from the validation set, contrasted with the high-risk group, which included 89 patients from the training set and 53 patients from the validation set. Analysis of PFS curves in Fig. 15 indicated highly significant differences with P values below 0.001. Notably, the median PFS for the low-risk group in both the training and validation cohorts was not reached, emphasizing their favorable prognosis. In contrast, the high-risk group exhibited median PFS values of 27, 24, and 33 months across the overall, training, and validation cohorts, respectively. These distinct outcomes underscore the robust predictive capability of the model. The plot further illustrated the variability in PFS among patients with dMMR status, culminating in the stratification of risk levels based on individual total scores. This comprehensive approach to risk assessment enhances our understanding of disease progression and aids in personalized treatment strategies for each patient. Discussion Radical gastrectomy and subsequent adjuvant chemotherapy have been shown to effectively treat early-stage gastric cancer [ 22 – 23 ] . Despite optimal multimodality therapy, 30–40% of patients may experience a relapse within 5 years [ 22 – 23 ] . The delicate balance of nature is illuminated by the vital function of the mismatch repair system. This intricate cellular mechanism is pivotal in identifying and correcting any mispaired bases resulting from inaccuracies in DNA replication, genetic recombination, or external influences like exposure to toxins or injury [ 24 ] . The biological system involving enzymes like MLH1, MLH3, MSH2, MSH3, MSH6, PMS1, and PMS2 is essential for repairing DNA mismatches. During DNA replication, complexes such as MSH2/MSH6 and MSH2/MSH3 detect and bind to errors in DNA. The MLH1/PMS2 heterodimers then remove and replace incorrect DNA bases at these sites. Disruptions in the function or expression of these enzymes can result in deficient complexes that are unable to properly repair DNA mismatches. As accurately classifying and selecting patients for personalized treatment is crucial, numerous studies have focused on the relationship between dMMR and various pathological features [ 25 ] . Previous reports have already demonstrated the prognostic role of dMMR in gastric cancer [ 26 – 29 ] . However, there is limited research on predicting the prognosis of dMMR cohorts in gastric cancer. In our study, we are the first to investigate different machine learning models to predict the prognosis of dMMR in Chinese gastric cancer patients and evaluate their merits and faults in terms of prediction ability. Our ultimate objective is to confirm the most precise machine learning model for dMMR patients based on the findings of this study while minimizing the negative effects of a limited sample size. Therefore, sensitivity analysis was applied to evaluate the stability of the model as more sample sizes were incorporated. Additionally, we conducted a comprehensive, long-term follow-up within the same study population, meticulously documenting their prognostic data. In our study, we utilized a combination of clinical features, pathological indicators, and tumor molecular markers to predict the progression-free survival (PFS) of deficient mismatch repair (dMMR) patients with gastric cancer. Our main goal was to evaluate different machine learning models for forecasting PFS in patients with dMMR. The detailed analysis for dMMR patients included a wide range of variables such as gender, age at surgery, presence of vascular and neural invasion, tumor stage, number of positive lymph nodes, Lauren classification, maximum tumor diameter, type of gastrectomy performed, omentum metastasis status, surgical margin outcomes, complication severity, and various proteomic markers associated with gastric cancer. Recognizing the potential impact of diverse factors on patient outcomes following radical gastrectomy in different medical settings, we conducted rigorous internal and external validations to assess the effectiveness of our model. These validations confirmed that our model provided strong predictive accuracy, calibration, discrimination, and clinical utility. Our holistic approach aims to offer valuable guidance for healthcare professionals and improve communication between patients and physicians. By incorporating multiple factors and conducting thorough validations, our model can assist clinicians in predicting PFS for dMMR patients, allowing for informed treatment decisions and personalized care plans. Currently, the TNM staging system is widely used as the main method for predicting the risk of clinical outcomes. However, its effectiveness and reliability are limited, leading to a significant decrease in its practical use [ 30 – 32 ] . Numerous studies have shown that using bar charts in assessments can improve accuracy and reduce the need for unnecessary tests in patients with gastric cancer [ 33 ] . Research has focused on factors such as age, gender, tumor size, lymph node status, tumor depth, location, Lauren and histological classification, and biomarkers to predict outcomes. Prognostic models have been developed with training and validation groups consistently yielding significant results (P < 0.001) [ 34 – 38 ] . Our team's innovative approach combines the transparency and reliability of the nomogram with the predictive power of machine learning models to create a comprehensive tool for prognosis in gastric cancer with dMMR. This hybrid system not only provides personalized and reliable prognostic information but also enhances shared decision-making between patients and clinicians. By accurately differentiating between different probabilities of progression-free survival in patients with dMMR, we can tailor treatment plans to individual patients' needs. High probability survival patients may not require additional postoperative treatment, while low probability survival patients with PD-L1 highly expressed conditions can benefit from targeted immunotherapies. This groundbreaking model has significant implications for personalized patient care and decision-making in oncology. By combining the strengths of both the nomogram and machine learning models, we can improve care, provide valuable information to patients, and aid clinicians in making management decisions for tongue cancer. Although the study yielded promising results, it is important to acknowledge its limitations. Firstly, the model construction and validation only utilized data from a single center, which raises the need for validation using data from other medical centers to ensure the generalizability of the findings. Secondly, there seems to be a discrepancy between the effectiveness of the line plot model in predicting the three-year progression-free survival (PFS) and the actual data. This inconsistency warrants further investigation to understand the underlying reasons behind it. Thirdly, the study failed to differentiate between patients with early and late-stage gastric carcinoma, which may have led to varying predictive performance among patients at different stages of the disease. When it comes to predicting patient outcomes in gastric cancer, it is generally observed that patients with dMMR tumors have a more favorable overall survival than those with pMMR tumors. This difference in prognosis could be due to the higher rate of immune infiltration in dMMR tumors, leading to increased tumor-infiltrating lymphocytes. Additionally, dMMR tumors are less likely to present adverse prognostic factors, such as lymphatic invasion and perineural invasion. These findings highlight the significant prognostic implications of MMR status in gastric cancer, not only in predicting patient outcomes but also in guiding treatment decisions. The presence of dMMR tumors in gastric cancer patients suggests a better response to immune checkpoint inhibitors, such as pembrolizumab. Pembrolizumab has already been approved for the treatment of MMR-deficient solid tumors, including gastric cancer. Therefore, understanding a patient's MMR status in gastric cancer becomes crucial in predicting prognosis and determining the most effective treatment options. In conclusion, recognizing the MMR status of gastric cancer patients plays a vital role in predicting their prognosis and selecting appropriate treatment strategies. The observation that dMMR tumors are associated with more favorable outcomes and higher responsiveness to immunotherapy highlights the importance of considering MMR status when managing gastric cancer patients. Abbreviations Progression free survival(PFS), deficient mismatch repair(dMMR), decision curve analysis (DCA), time-dependent receiver operating characteristic (t-ROC), Random Survival Forests (RSF), Decision Survival Tree (DST), Extreme Gradient Boosting (XG boosting) Declarations Ethics approval and consent to participate This retrospective study utilized clinical data collected for clinical purposes. Extensive consultation with the Ethics Committee of Shanxi Cancer Hospital was conducted, and ethical approval was granted by the Ethics Committee of Shanxi Carcinoma Hospital (No: 2022JC23). All methods were carried out in accordance with relevant guidelines and regulations. The procedures involving human participants were conducted in compliance with ethical standards set by our institutional research committee. Informed consent was obtained from all patients, indicating their voluntary and informed agreement to participate in the study. Patients also provided written consent for publication and any accompanying images. A copy of the consent form is available for review by the journal's Editor-in-Chief upon request. Registration of research Our study has been officially recorded on the Research Registry under the identifying number Research Registry 9867 (https://researchregistry.knack.com/research-registry#user-researchregistry/ ). Authors’ Contributions Yifan Li wrote the main manuscript text and prepared figures 1-15, while Jinfeng Ma and Wenhua Cheng reviewed and edited the manuscript. Consent for publication Not applicable. Data availability Data is provided within the manuscript or supplementary information files. Funding Supported by the Science and Education Cultivation Fund of the National Cancer and Regional Medical center of Shanxi Provincial Cancer hospital (SD2023005) References Sung H, Ferlay J, Siegel RL, Laversanne M, et al. Global Cancer Statistics 2020: GLOBOCAN Estimates of Incidence and Mortality Worldwide for 36 Cancers in 185 Countries. CA Cancer J Clin. 2021 May;71(3):209-249. Ajani JA, Lee J, Sano T, et al. Gastric adenocarcinoma. Nat Rev Dis Primers. 2017 Jun 1;3:17036. Bourke MJ, Neuhaus H, Bergman JJ. Endoscopic Submucosal Dissection: Indications and Application in Western Endoscopy Practice. Gastroenterology. 2018 May;154(7):1887-1900.e5. Salati M, Orsi G, Smyth E, et al. Gastric cancer: Translating novel concepts into clinical practice. Cancer Treat Rev. 2019 Sep;79:101889. Karimi P, Islami F, Anandasabapathy S, et al. Gastric cancer: descriptive epidemiology, risk factors, screening, and prevention. Cancer Epidemiol Biomarkers Prev. 2014 May;23(5):700-13. Biagioni A, Skalamera I, Peri S, et al. Update on gastric cancer treatments and gene therapies. Cancer Metastasis Rev. 2019 Sep;38(3):537-548. Smyth EC, Wotherspoon A, Peckitt C, et al. Mismatch Repair Deficiency, Microsatellite Instability, and Survival: An Exploratory Analysis of the Medical Research Council Adjuvant Gastric Infusional Chemotherapy (MAGIC) Trial. JAMA Oncol. 2017 Sep 1;3(9):1197-1203. André T, Lonardi S, Wong KYM, et al. Nivolumab plus low-dose ipilimumab in previously treated patients with microsatellite instability-high/mismatch repair-deficient metastatic colorectal cancer: 4-year follow-up from CheckMate 142. Ann Oncol. 2022 Oct;33(10):1052-1060. Overman MJ, Lonardi S, Wong KYM, et al. Durable Clinical Benefit With Nivolumab Plus Ipilimumab in DNA Mismatch Repair-Deficient/Microsatellite Instability-High Metastatic Colorectal Cancer. J Clin Oncol. 2018 Mar 10;36(8):773-779. Choi YY, Kim H, Shin SJ, et al. Microsatellite Instability and Programmed Cell Death-Ligand 1 Expression in Stage II/III Gastric Cancer: Post Hoc Analysis of the CLASSIC Randomized Controlled Study. Ann Surg. 2019 Aug;270(2):309-316. Tsai CY, Lin TA, Huang SC, et al. Is Adjuvant Chemotherapy Necessary for Patients with Deficient Mismatch Repair Gastric Cancer?-Autophagy Inhibition Matches the Mismatched. Oncologist. 2020 Jul;25(7):e1021-e1030. Zhao L, Fu Y, Niu P, Zhang F, et al. Perioperative Chemotherapy Could Not Improve the Prognosis of Gastric Cancer Patients With Mismatch Repair Deficiency: A Multicenter, Real-World Study. Oncologist. 2023 Oct 3;28(10):e891-e901 Hwang TJ, Kesselheim AS, Vokinger KN. Lifecycle Regulation of Artificial Intelligence- and Machine Learning-Based Software Devices in Medicine. JAMA. 2019 Dec 17;322(23):2285-2286. Liu X, Faes L, Kale AU, Wagner SK, et al. A comparison of deep learning performance against health-care professionals in detecting diseases from medical imaging: a systematic review and meta-analysis. Lancet Digit Health. 2019 Oct;1(6):e271-e297. Lee KH, Choi GH, Yun J, et al. Machine learning-based clinical decision support system for treatment recommendation and overall survival prediction of hepatocellular carcinoma: a multi-center study. NPJ Digit Med. 2024 Jan 5;7(1):2. Xue B, Li D, Lu C, et al. Use of Machine Learning to Develop and Evaluate Models Using Preoperative and Intraoperative Data to Identify Risks of Postoperative Complications. JAMA Netw Open. 2021 Mar 1;4(3):e212240. Powell MA. Immunohistochemistry to determine mismatch repair-deficiency in endometrial cancer: the appropriate standard. Ann Oncol. 2017 Jan 1;28(1):9-10. Uno H, Cai T, Pencina MJ, et al. On the C-statistics for evaluating overall adequacy of risk prediction procedures with censored survival data. Stat Med. 2011 May 10;30(10):1105-17. Harrell FE Jr, Lee KL, Mark DB. Multivariable prognostic models: issues in developing models, evaluating assumptions and adequacy, and measuring and reducing errors. Stat Med. 1996 Feb 28;15(4):361-87. Li Y, Zhang X. Prognostic nomograms for gastric carcinoma after surgery to assist decision-making for postoperative treatment with chemotherapy cycles <9 or chemotherapy cycles ≥9. Front Surg. 2022 Aug 26;9:916483. Li Y, Bai M, Gao Y. Prognostic nomograms for gastric carcinoma after D2 + total gastrectomy to assist decision-making for postoperative treatment: based on Lasso regression. World J Surg Oncol. 2023 Jul 20;21(1):207. Park SH, Sohn TS, Lee J, et al. Phase III Trial to Compare Adjuvant Chemotherapy With Capecitabine and Cisplatin Versus Concurrent Chemoradiotherapy in Gastric Cancer: Final Report of the Adjuvant Chemoradiotherapy in Stomach Tumors Trial, Including Survival and Subset Analyses. J Clin Oncol. 2015 Oct 1;33(28):3130-6. Noh SH, Park SR, Yang HK, et al. Adjuvant capecitabine plus oxaliplatin for gastric cancer after D2 gastrectomy (CLASSIC): 5-year follow-up of an open-label, randomized phase 3 trial. Lancet Oncol. 2014 Nov;15(12):1389-96. Ruiz-Bañobre J, Goel A. DNA Mismatch Repair Deficiency and Immune Checkpoint Inhibitors in Gastrointestinal Cancers. Gastroenterology. 2019 Mar;156(4):890-903. Polom K, Marano L, Marrelli D, et al. Meta-analysis of microsatellite instability about clinicopathological characteristics and overall survival in gastric cancer. Br J Surg. 2018 Feb;105(3):159-167. Pietrantonio F, Miceli R, Raimondi A, et al. Individual Patient Data Meta-Analysis of the Value of Microsatellite Instability As a Biomarker in Gastric Cancer. J Clin Oncol. 2019 Dec 10;37(35):3392-3400. Quaas A, Biesma HD, Wagner AD, et al. Microsatellite instability and sex differences in resectable gastric cancer - A pooled analysis of three European cohorts. Eur J Cancer. 2022 Sep;173:95-104. Yu HY, Li CP, Huang YH, et al. Microsatellite Instability, Epstein-Barr Virus, and Programmed Cell Death Ligand 1 as Predictive Markers for Immunotherapy in Gastric Cancer. Cancers (Basel). 2022 Jan 3;14(1):218. Kohlruss M, Ott K, Grosser B, et al. Sexual Difference Matters: Females with High Microsatellite Instability Show Increased Survival after Neoadjuvant Chemotherapy in Gastric Cancer. Cancers (Basel). 2021 Mar 2;13(5):1048. Wang F, Sun GP, Zou YF, et al. Helicobacter pylori infection predicts favorable outcomes in patients with gastric cancer. Curr Oncol. 2013; 20(5): e388–e395. Lu J, Zheng ZF, Wang W, et al. A novel TNM staging system for gastric cancer based on the metro-ticket paradigm: a comparative study with the AJCC-TNM staging system. Gastric Cancer. 2019 Jul;22(4):759-768. Zhou YY, Kang YT, Chen C, et al. Combination of TNM staging and pathway-based risk score models in patients with gastric cancer. J Cell Biochem. 2018 Apr;119(4):3608-3617. Lu J, Zheng ZF, Xie JW, et al. Is the 8th Edition of the AJCC TNM Staging System Sufficiently Reasonable for All Patients with Noncardia Gastric Cancer? A 12,549-Patient International Database Study. Ann Surg Oncol. 2018 Jul;25(7):2002-2011. Iasonos A, Schrag D, Raj GV, Panageas KS. How to build and interpret a nomogram for cancer prognosis. J Clin Oncol. 2008 Mar 10;26(8):1364-70. Necchi A, Sonpavde G, Lo Vullo S, et al. Nomogram-based Prediction of Overall Survival in Patients with Metastatic Urothelial Carcinoma Receiving First-line Platinum-based Chemotherapy: Retrospective International Study of Invasive/Advanced Cancer of the Urothelium (RISC). Eur Urol. 2017 Feb;71(2):281-289. Yip PL, Lee SF, Choi CH, et al. External Validation of a Nomogram to Predict Survival and Benefit of Concurrent Chemoradiation for Stage II Nasopharyngeal Carcinoma. Cancers (Basel). 2021 Aug 25;13(17):4286. Raghav K, Hwang H, Jácome AA, et al. Development and Validation of a Novel Nomogram for Individualized Prediction of Survival in Cancer of Unknown Primary. Clin Cancer Res. 2021 Jun 15;27(12):3414-3421. Chang YR, Huang WK, Wang SY, et al. A Nomogram Predicting Progression Free Survival in Patients with Gastrointestinal Stromal Tumor Receiving Sunitinib: Incorporating Pre-Treatment and Post-Treatment Parameters. Cancers (Basel). 2021 May 25;13(11):2587. Liu S, Yu X, Yang S, et al. Machine Learning-Based Radiomics Nomogram for Detecting Extramural Venous Invasion in Rectal Cancer. Front Oncol. 2021 Mar 26;11:610338. Huang L, Lin W, Xie D, et al. Development and validation of a preoperative CT-based radiomic nomogram to predict pathology invasiveness in patients with a solitary pulmonary nodule: a machine learning approach, multicenter, diagnostic study. Eur Radiol. 2022 Mar;32(3):1983-1996. Alabi RO, Mäkitie AA, Pirinen M, et al. Comparison of nomogram with machine learning techniques for prediction of overall survival in patients with tongue cancer. Int J Med Inform. 2021 Jan;145:104313. Tables Table1. Baseline of study population Variables Training cohort (n=405) Mean±SD / No(%) Validation cohort (n=177) Mean±SD /No (%) P Gender 0.843 Male 322(79.5%) 142(80.2%) Female 83(20.5%) 35(19.8%) Age (years) 58.78±8.93 58.47±9.29 0.677 pT stage 0.431 T1 155(38.3%) 66(37.3%) T2 25(6.2%) 6(3.4%) T3 89(22.0%) 20(21.5%) T4 136(33.6%) 42(37.9%) Number of positive lymph nodes 0.138 0 224(55.3%) 82(46.3%) 1-2 76(18.8%) 33(18.6%) 3-6 32(7.9%) 20(11.3%) ≥7 73(18.0%) 42(23.7%) pTNM stage 0.125 Ⅰ 174(43.0%) 67(37.9%) Ⅱ 93(23.0%) 37(20.9%) Ⅲ 138(34.1%) 73(41.2%) Vascular invasion 0.482 Negative 255(63.0%) 106(59.9%) Positive 150(37.0%) 71(40.1%) Neural invasion 0.197 Negative 274(67.7%) 110(62.1%) Positive 131(32.3%) 67(37.9%) Lauren classification 0.951 Intestinal 222(54.8%) 96(54.2%) Diffused 94(23.2%) 45(25.4%) Mixed 89(22.0%) 36(20.3%) Overall survival(months) 42.53±21.29 43.06±23.13 0.448 Type of gastrectomy 0.103 Primal 61(15.1%) 17(9.6%) Distal 158(39.0%) 69(39.0%) Total 185(45.7%) 90(50.8%) PPG 1(0.2%) 1(0.6%) Omentum metastasis 0.295 Negative 402(99.3%) 174(98.7%) Positive 3(0.7%) 3(1.7%) Surgical margin 0.477 Negative 386(95.3%) 171(96.6%) Positive 19(4.7%) 6(3.4%) Her-2 0.913 - 245(60.5%) 106(59.9%) + 134(33.1%) 60(33.9%) ++ 19(4.7%) 8(4.5%) +++ 7(1.7%) 3(1.7%) Recurrence 0.83 Negative 346(85.4%) 150(84.7%) Positive 59(14.6%) 27(15.3%) Progress Free Survival(months) 39.53±22.18 38.55±23.99 0.452 AE1/AE3 0.733 Negative 203(50.1%) 86(48.6%) Positive 202(49.9%) 91(51.4%) Ki67(%) 37.83±29.80 41.73±29.64 0.115 CK7 0.572 Negative 239(59.0%) 100(56.5%) Positive 166(41.0%) 77(43.5%) CK20 0.999 Negative 254(62.7%) 111(59.0%) Positive 151(37.3%) 66(41.0%) CDX-2 0.557 Negative 271(66.9%) 114(64.4%) Positive 134(33.1%) 63(35.6%) SATB-2 0.404 Negative 344(84.9%) 155(87.6%) Positive 61(15.1%) 22(12.4%) SYN 0.166 Negative 280(69.1%) 112(63.3%) Positive 125(30.9%) 65(36.7%) CGA 0.853 Negative 276(68.1%) 122(68.9%) Positive 129(31.9%) 55(31.1%) CD56 0.804 Negative 305(75.3%) 135(76.3%) Positive 100(24.7%) 42(23.7%) MLH1 0.333 Negative 152(37.5%) 59(33.3%) Positive 253(62.5%) 118(66.7%) PMS2 0.896 Negative 320(79.0%) 139(78.5%) Positive 85(21.0%) 38(21.5%) MSH2 0.654 Negative 159(39.3%) 66(37.3%) Positive 246(60.7%) 111(62.7%) MSH6 0.48 Negative 152(37.5%) 61(34.5%) Positive 253(62.5%) 116(65.5%) Maximum diameter of tumor(cm) 4.05±2.42 4.63±2.75 0.14 Tumor location 0.981 Upper 1/3 187(46.2%) 79(44.6%) Middle 1/3 68(16.8%) 34(19.2%) Lower 1/3 147(36.3%) 63(35.6%) Multiple 3(0.7%) 1(0.6%) Table 2 Discrimination and calibration of each model predicting progression-free survival at 60 months Machine learning model C-Index 95%CI AUC 95%CI Brier score Radom survival forest 0.936 0.935-0.937 0.968 0.966-0.968 0.067 Decision survival tree 0.884 0.883-0.887 0.924 0.924-0.926 0.030 Gradient Boosting survival 0.904 0.904-0.907 0.945 0.944-0.946 0.029 Table 3 Multivariate analysis of PFS of training cohort of dMMR and analyzed by Cox regression B SE Wald df P HR HR (95%CI) Age 0.023 0.011 4.683 1 0.030 1.023 1.002-1.045 Number of positive lymph nodes 46.954 3 <0.001 0 Vs 1-2 0.740 0.302 6.015 1 0.014 2.097 1.160-3.788 0 Vs 3-6 0.699 0.381 3.375 1 0.066 2.013 0.954-4.245 0 VS ≥7 1.817 0.292 38.648 1 <0.001 6.153 3.470-10.911 Neural invasion Negative Vs Positive 0.570 0.198 8.262 1 0.004 1.769 1.199-2.610 Ki67 0.014 0.004 9.438 1 0.002 1.014 1.005-1.023 Abbreviations: B, regression coefficient; SE, standard error; df, degree of freedom; HR, hazard ratio; CI, confidence interval Additional Declarations No competing interests reported. Cite Share Download PDF Status: Published Journal Publication published 24 Jan, 2025 Read the published version in BMC Cancer → Version 1 posted Editorial decision: Revision requested 27 Jun, 2024 Editor assigned by journal 26 Jun, 2024 Submission checks completed at journal 26 Jun, 2024 First submitted to journal 25 Jun, 2024 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-4639290","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":319679518,"identity":"c2beba77-da45-4fba-978e-275ff21cc01c","order_by":0,"name":"Jinfeng Ma","email":"","orcid":"","institution":"Shanxi Cancer Hospital","correspondingAuthor":false,"prefix":"","firstName":"Jinfeng","middleName":"","lastName":"Ma","suffix":""},{"id":319679519,"identity":"e14d7c7b-766a-498b-9d6a-3ad5d304ab0e","order_by":1,"name":"Wenhua Cheng","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA80lEQVRIie3PsUvEMBTH8ZRAuqSX9ZXC+S+8EqgnlPtbLIVMHXS76Qwc3HR4a/+Mjo6pwU4VVweHSsG5bg43iLPS6OaQz/y+PH6EeN5/RMXQvp9gKcTxdZgwX7uTkAdjrFcyrnua1leqdCeCU4z0pmh0xRI+3QfaWeyiDuI7CBryaGWOhpLQPjRzCdiFguse6Dm9LcYKXxaEK/U8+8aSDIABu9BGygrfKAGezSZnXwlnwNFcZskKbaBdCVouMdoDoKmyhPwmSS0rRugB47or0wOqkrm2LJ+saafN9uYodu3wccrXIrTd/Pxv2N/OPc/zvJ98ArztS4hFM9HbAAAAAElFTkSuQmCC","orcid":"","institution":"Shanxi Cancer Hospital","correspondingAuthor":true,"prefix":"","firstName":"Wenhua","middleName":"","lastName":"Cheng","suffix":""},{"id":319679520,"identity":"14f76245-8eb1-4f1d-a987-628b99187c00","order_by":2,"name":"Yifan Li","email":"","orcid":"","institution":"Shanxi Cancer Hospital","correspondingAuthor":false,"prefix":"","firstName":"Yifan","middleName":"","lastName":"Li","suffix":""}],"badges":[],"createdAt":"2024-06-26 02:14:55","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-4639290/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-4639290/v1","draftVersion":[],"editorialEvents":[{"content":"https://doi.org/10.1186/s12885-025-13542-0","type":"published","date":"2025-01-24T15:57:12+00:00"}],"editorialNote":"","failedWorkflow":false,"files":[{"id":60709921,"identity":"e11d79c5-bbf0-42e0-b84c-f370a6da3b67","added_by":"auto","created_at":"2024-07-19 19:59:12","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":2605973,"visible":true,"origin":"","legend":"\u003cp\u003eThe flowchart of study population enrolment in the training and validation cohort of gastric cancer.\u003c/p\u003e","description":"","filename":"Figure1.png","url":"https://assets-eu.researchsquare.com/files/rs-4639290/v1/cfa12f59749fad621e589f3c.png"},{"id":60708855,"identity":"895267ae-2e86-42a0-ac0b-66afc33d95a9","added_by":"auto","created_at":"2024-07-19 19:51:12","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":1756832,"visible":true,"origin":"","legend":"\u003cp\u003eWorkflow for data management and gastric cancer PFS prediction model development\u003c/p\u003e","description":"","filename":"Figure2.png","url":"https://assets-eu.researchsquare.com/files/rs-4639290/v1/a44cb38dcfa7c8180bf20fe0.png"},{"id":60708853,"identity":"3f880670-c81c-4679-949e-4a64214e0767","added_by":"auto","created_at":"2024-07-19 19:51:12","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":127032,"visible":true,"origin":"","legend":"\u003cp\u003eKaplan-Meier survival curves for PFS of training and validation cohorts\u003c/p\u003e","description":"","filename":"Figure3.png","url":"https://assets-eu.researchsquare.com/files/rs-4639290/v1/75adc78df7ea5804a8073ffd.png"},{"id":60709923,"identity":"2a67cb87-f410-4d1b-9240-c8f0808754ec","added_by":"auto","created_at":"2024-07-19 19:59:12","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":3411111,"visible":true,"origin":"","legend":"\u003cp\u003eA. Machine learning models’ discrimination measures for in 60 months. B. Prediction error curves show the Brier score in 60 months\u003c/p\u003e","description":"","filename":"Figure4.png","url":"https://assets-eu.researchsquare.com/files/rs-4639290/v1/bbca34a8742cefc200a1cc88.png"},{"id":60709922,"identity":"93910a00-6872-465e-bd14-e059eeaa237e","added_by":"auto","created_at":"2024-07-19 19:59:12","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":3383904,"visible":true,"origin":"","legend":"\u003cp\u003eA. The SHAP plot of the RSF model. B. Feature importance of RSF model\u003c/p\u003e","description":"","filename":"Figure5.png","url":"https://assets-eu.researchsquare.com/files/rs-4639290/v1/fa37e013826a8bcee2db9661.png"},{"id":60708865,"identity":"ff6e6c87-a671-4624-98a2-090e433f5fa5","added_by":"auto","created_at":"2024-07-19 19:51:13","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":3458604,"visible":true,"origin":"","legend":"\u003cp\u003eA. The SHAP plot of the DST model. B. Feature importance of DST model\u003c/p\u003e","description":"","filename":"Figure6.png","url":"https://assets-eu.researchsquare.com/files/rs-4639290/v1/f8cf5800970424bb3986dfc6.png"},{"id":60708866,"identity":"bd071493-4a31-4fa1-a656-9066c7c0fc21","added_by":"auto","created_at":"2024-07-19 19:51:13","extension":"png","order_by":7,"title":"Figure 7","display":"","copyAsset":false,"role":"figure","size":3359158,"visible":true,"origin":"","legend":"\u003cp\u003eA. The SHAP plot of the XG boosting model. B. Feature importance of XG boosting model\u003c/p\u003e","description":"","filename":"Figure7.png","url":"https://assets-eu.researchsquare.com/files/rs-4639290/v1/805d44387875c4f40b1ca9af.png"},{"id":60710818,"identity":"209d6299-1d3b-4600-baef-d8df4f7cf08b","added_by":"auto","created_at":"2024-07-19 20:15:12","extension":"png","order_by":8,"title":"Figure 8","display":"","copyAsset":false,"role":"figure","size":3746682,"visible":true,"origin":"","legend":"\u003cp\u003eDecision curve analysis showing the net benefit associated with the use of machine learning prognostic models at 60 months to predict PFS A. Training cohort B. Validation cohort\u003c/p\u003e","description":"","filename":"Figure8.png","url":"https://assets-eu.researchsquare.com/files/rs-4639290/v1/ac05144056c20d7577148555.png"},{"id":60708857,"identity":"f7a3d1a3-6066-4030-b6ff-a71aa74f72d4","added_by":"auto","created_at":"2024-07-19 19:51:12","extension":"png","order_by":9,"title":"Figure 9","display":"","copyAsset":false,"role":"figure","size":838496,"visible":true,"origin":"","legend":"\u003cp\u003eCalibration curves of machine learning prognostic models A. Training cohort B. Validation cohort\u003c/p\u003e","description":"","filename":"Figure9.png","url":"https://assets-eu.researchsquare.com/files/rs-4639290/v1/b8e6c9a15b2b4d61981f2260.png"},{"id":60708863,"identity":"71a57fb3-79f4-4ea4-8d4a-b4007c7d18fd","added_by":"auto","created_at":"2024-07-19 19:51:12","extension":"png","order_by":10,"title":"Figure 10","display":"","copyAsset":false,"role":"figure","size":439994,"visible":true,"origin":"","legend":"\u003cp\u003eSensitivity analysis of machine learning model A. RSF B.DST C.XG boosting\u003c/p\u003e","description":"","filename":"Figure10.png","url":"https://assets-eu.researchsquare.com/files/rs-4639290/v1/3f02b93f18ea7bc54fb1cfed.png"},{"id":60708861,"identity":"a47264f8-21bf-42f0-8a7c-8ce02309a34a","added_by":"auto","created_at":"2024-07-19 19:51:12","extension":"png","order_by":11,"title":"Figure 11","display":"","copyAsset":false,"role":"figure","size":422425,"visible":true,"origin":"","legend":"\u003cp\u003eNomogram model to predict 3-year and 5-year PFS\u003c/p\u003e","description":"","filename":"Figure11.png","url":"https://assets-eu.researchsquare.com/files/rs-4639290/v1/e6b5dfd20a482c6247b4ff84.png"},{"id":60708864,"identity":"f5c22815-58e9-4031-ab4e-d2e9d265dd41","added_by":"auto","created_at":"2024-07-19 19:51:12","extension":"png","order_by":12,"title":"Figure 12","display":"","copyAsset":false,"role":"figure","size":1199031,"visible":true,"origin":"","legend":"\u003cp\u003eA. Calibration curves of internal validation to predict 3-year PFS B. Calibration curves of external validation to predict 3-year PFS C. Calibration curves of internal validation to predict 5-year PFS D. Calibration curves of external validation to predict 5- year PFS\u003c/p\u003e","description":"","filename":"Figure12.png","url":"https://assets-eu.researchsquare.com/files/rs-4639290/v1/e2cfe0cc28a462b6589b7444.png"},{"id":60710588,"identity":"e163a3f1-f7c0-4ae0-921f-9252e20a79cc","added_by":"auto","created_at":"2024-07-19 20:07:12","extension":"png","order_by":13,"title":"Figure 13","display":"","copyAsset":false,"role":"figure","size":2521274,"visible":true,"origin":"","legend":"\u003cp\u003eA. Time-dependent receiver operating characteristic (t-ROC) curves of internal validation to predict PFS B. Time-dependent receiver operating characteristic (t-ROC) curves of external validation to predict PFS\u003c/p\u003e","description":"","filename":"Figure13.png","url":"https://assets-eu.researchsquare.com/files/rs-4639290/v1/8307f8c1e9c800d60c6bfaa7.png"},{"id":60708867,"identity":"90930cea-2415-4918-99c2-504be6a23f22","added_by":"auto","created_at":"2024-07-19 19:51:13","extension":"png","order_by":14,"title":"Figure 14","display":"","copyAsset":false,"role":"figure","size":1840093,"visible":true,"origin":"","legend":"\u003cp\u003eA. Decision curve analysis (DCA) of internal validation to predict 3-year PFS B. Decision curve analysis (DCA) of internal validation to predict 5-year PFSC. Decision curve analysis (DCA) of external validation to predict 3- year PFS D. Decision curve analysis (DCA) of external validation to predict 5- year PFS\u003c/p\u003e","description":"","filename":"Figure14.png","url":"https://assets-eu.researchsquare.com/files/rs-4639290/v1/81a9d25e56c318aba90d46bd.png"},{"id":60708868,"identity":"36086c61-7a8c-42cd-992b-9afd7b834906","added_by":"auto","created_at":"2024-07-19 19:51:13","extension":"png","order_by":15,"title":"Figure 15","display":"","copyAsset":false,"role":"figure","size":2133096,"visible":true,"origin":"","legend":"\u003cp\u003eThe Kaplan-Meier survival curves patients with different scores. A. all cohort B. training cohort C. validation cohort\u003c/p\u003e","description":"","filename":"Figure15.png","url":"https://assets-eu.researchsquare.com/files/rs-4639290/v1/56f340b9725e73ba244c5784.png"},{"id":74858384,"identity":"ae5c8555-d540-46ed-8441-aa852004c2f7","added_by":"auto","created_at":"2025-01-27 16:08:49","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":45585161,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-4639290/v1/99f29a6e-be85-4233-96e8-513ae8be480a.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"Construction and validation the first prognostic models of progression-free survival in gastric cancer patients after gastrectomy with deficient mismatch repair: Nomogram and three machine learning models approaches","fulltext":[{"header":"Introduction","content":"\u003cp\u003eDespite being a significant issue globally, gastric cancer has now become the fifth most prevalent form of cancer and the fourth most common cause of mortality \u003csup\u003e[\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e]\u003c/sup\u003e. Thanks to advancements in medical technology, there are now numerous effective treatments available for those diagnosed with this condition \u003csup\u003e[\u003cspan additionalcitationids=\"CR3 CR4 CR5\" citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e]\u003c/sup\u003e. The prognosis for patients with gastric cancer has been greatly improved by various treatments such as endoscopy, surgery, radiotherapy, chemotherapy, and immunotherapy. One area of research interest is the investigation of tumors that have deficient mismatch repair (dMMR), also known as high-frequency microsatellite instability (MSI-H). These tumors lack mismatch repair proteins in tumor cell nuclei and/or have reduced MMR activity\u003csup\u003e[\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e]\u003c/sup\u003e. Immunohistochemistry (IHC) and microsatellite instability (MSI) testing can be used to assess the MMR status of tumor DNA. Research suggests that dMMR can enhance the immune response against tumor cells and hinder their growth in colorectal cancer, usually due to mutations or epigenetic alterations in DNA MMR genes\u003csup\u003e[\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e]\u003c/sup\u003e. However, the role of dMMR in gastric cancer is still unclear, making it challenging to establish its clinical significance\u003csup\u003e[\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e]\u003c/sup\u003e. Some studies suggest a link between dMMR status and improved long-term survival after gastrectomy, but recent clinical trials have contradicted this finding. It is evident that dMMR is not a standalone factor in the outcomes of gastric cancer patients. Despite progress in gastric cancer treatment, the exact role and clinical significance of dMMR in this cancer type remain unknown\u003csup\u003e[\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e, \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e]\u003c/sup\u003e. Further investigation is necessary to fully explore the potential of dMMR as a prognostic or predictive marker in gastric cancer, which could ultimately enhance treatment strategies and outcomes for patients.\u003c/p\u003e \u003cp\u003eThe beauty of artificial intelligence and machine learning is their ability to continuously evolve and refine themselves autonomously. These advanced technologies go beyond traditional methods, offering a new way to analyze data. Machine learning can uncover hidden connections between variables that may not be easily apparent\u003csup\u003e[\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e]\u003c/sup\u003e. In the field of medicine, machine learning is gaining popularity, especially in medical imaging\u003csup\u003e[\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e]\u003c/sup\u003e, precise cancer prognosis\u003csup\u003e[\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e]\u003c/sup\u003e, and predicting surgical complication\u003csup\u003e[\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e]\u003c/sup\u003es. This study focused on creating accurate machine learning models to predict the progression-free survival of patients with gastric carcinoma and deficient mismatch repair (dMMR). By utilizing machine learning, researchers aim to provide precise and efficient predictions of patient outcomes, leading to improved personalized treatment strategies. Artificial intelligence and machine learning have the potential to revolutionize the analysis and interpretation of medical data, promising a bright future for advancing healthcare practices.\u003c/p\u003e"},{"header":"Patients and methods","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e\n \u003ch2\u003eData source and study population\u003c/h2\u003e\n \u003cp\u003eThe study focused on patients who underwent radical gastrectomy for nonmetastatic gastric cancer at Shanxi Cancer Hospital between May 1, 2002, and December 31, 2020. Diagnosis of gastric carcinoma was confirmed through histological examination, and patients had comprehensive clinicopathological details available. Follow-up data was collected for patients with no severe organ damage, no prior history of other malignancies, and no unrelated causes of death. Exclusion criteria included systemic tumors, incomplete clinical information, palliative procedures, or non-gastric cancer cases. Tumor staging followed the AJCC 8th TNM classification. The study protocol was approved by the Ethics Committee of Shanxi Cancer Hospital and complied with the Declaration of Helsinki. Patient confidentiality was maintained through anonymization of data, and informed consent was obtained from all participants, along with written permission for publication and use of accompanying visuals. Refer to Figs. 1 and \u003cspan class=\"InternalRef\"\u003e2\u003c/span\u003e for an overview of the research methodology.\u003c/p\u003e\n \u003cp\u003eThe research participants were randomly split into two distinct groups: a training cohort representing 70% of the patients, which was utilized for the development of machine learning algorithms, and a validation cohort encompassing the remaining 30% of patients, which served as the testing ground for the algorithms. Demographic details of the study population were presented through frequencies and percentages for categorical variables, while mean and standard deviation (SD) were used for continuous variables.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec4\" class=\"Section2\"\u003e\n \u003ch2\u003ePredictive variables\u003c/h2\u003e\n \u003cp\u003eTable \u003cspan class=\"InternalRef\"\u003e1\u003c/span\u003e showcases all the variables integrated into the machine learning models. These variables encompassed a wide range of standard data categories, such as demographics, laboratory results, surgical specifics, tumor characteristics, proteomics data, and pathology reports. The researchers undertook an exhaustive analysis of various factors that significantly impacted the surgical outcomes. The investigated factors encompassed gender, age at surgery, presence of vascular and neural invasion, tumor stage (pT stage), number of positive lymph nodes, Lauren classification, maximum tumor diameter, type of gastrectomy undertaken, omentum metastasis status, surgical margin outcomes, complication severity as per the Clavien-Dindo classification, expression levels of specific biomarkers (including AE1/AE3, CK20, CDX-2, SATB-2, SYN, CGA, CD56, MLH1, PMS2, Her-2, MSH2, and MSH6), and progression-free survival (PFS).\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec5\" class=\"Section2\"\u003e\n \u003ch2\u003eFollow-up protocol and outcome\u003c/h2\u003e\n \u003cp\u003ePatients included in this study had to meet specific criteria, which involved receiving a confirmed diagnosis of gastric cancer through histological examination and undergoing surgery aimed at curing the disease (R0-R1). The researchers explored various factors that could potentially impact the surgical outcome. In order to be eligible for the study, patients needed to meet certain requirements. These requirements included having a confirmed diagnosis of gastric cancer through histological examination and undergoing surgery with the goal of curing the disease (R0-R1). The researchers utilized electronic medical records and collaborated with oncologists to determine the duration of follow-up. Postoperative follow-up protocols varied depending on the institution/physician, but generally followed national and international guidelines. This included a 3-month postoperative outpatient visit in the first year, followed by appointments every 6 months for the next 3 years, and then annually for an additional 2 years. Follow-up examinations involved physical assessments and contrast-enhanced computed tomography scans of the chest, abdomen, and pelvis. The primary objective of the model prediction in this study was progression-free survival (PFS). PFS was defined as the period from surgery to disease recurrence (including both local and distant recurrences) or death from any cause.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec6\" class=\"Section2\"\u003e\n \u003ch2\u003eMMR Status Detected by Immunohistochemistry(IHC)\u003c/h2\u003e\n \u003cp\u003eThe tissue specimens were initially preserved in a 10% formalin solution and then embedded in paraffin, following a standard protocol. Immunohistochemistry (IHC) was performed on the MMR protein using Ventana anti-MLH1, anti-PMS2, MSH2, and MSH6 mouse monoclonal antibodies from reputable suppliers. The staining process was carried out on the VENTANA Bench Mark ULTRA platform, with 2 \u0026micro;m sections of the specimens examined for the presence of MLH1, MSH2, MSH6, and PMS2\u003csup\u003e[\u003cspan class=\"CitationRef\"\u003e17\u003c/span\u003e]\u003c/sup\u003e. In order to evaluate the results, two senior pathologists assessed the staining patterns, with a third senior pathologist brought in to resolve any disagreements. A lack of staining in the tumor cell nuclei, alongside positive staining in adjacent normal cells such as fibroblasts and lymphocytes, was considered a negative assessment. The absence or presence of nuclear staining was used to distinguish between intact and lost signals. A tumor was classified as deficient in MMR protein expression if the cancerous epithelial cells lacked nuclear staining, while surrounding non-neoplastic cells exhibited positive staining. Conversely, if all four MMR proteins were present in the tumor cells, it was categorized as proficient MMR (pMMR). Any level of nuclear staining above the background level of cancer cells indicated MMR protein expression. In cases where one of the proteins was absent, the tumor was labeled as deficient MMR (dMMR).\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec7\" class=\"Section2\"\u003e\n \u003ch2\u003eMachine learning models\u003c/h2\u003e\n \u003cp\u003eIn the study, three survival machine learning algorithms - Random Survival Forests (RSF), Decision survival tree (DST), and Extreme Gradient Boosting (XG boosting) were used. The optimization of machine learning parameters involved a randomized search of various parameter settings, followed by testing with a ten-fold cross validation in order to maximize the concordance index (c-index). The parameters explored for each model and the final parameter combinations used for analysis can be found in Table \u003cspan class=\"InternalRef\"\u003e2\u003c/span\u003e.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec8\" class=\"Section2\"\u003e\n \u003ch2\u003eNomogram development\u003c/h2\u003e\n \u003cp\u003eThis research sought to develop nomograms that can effectively predict progression-free survival (PFS) by incorporating variables identified through Cox regression analysis. To achieve this, the researchers conducted Cox regression analysis on a range of PFS-related parameters and repeated the process 10,000 times to ensure robustness.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec9\" class=\"Section2\"\u003e\n \u003ch2\u003eModel performance, fit, and validation assessments\u003c/h2\u003e\n \u003cp\u003eThe model\u0026apos;s performance was evaluated using discrimination, calibration, and decision curve analysis (DCA). Discrimination, which assesses the model\u0026apos;s ability to differentiate between patients who will experience an event and those who will not, was measured using the c-index method by Uno et al \u003csup\u003e[\u003cspan class=\"CitationRef\"\u003e18\u003c/span\u003e]\u003c/sup\u003e, for specific time points and by Harrell et al for the overall time assessment \u003csup\u003e[\u003cspan class=\"CitationRef\"\u003e19\u003c/span\u003e]\u003c/sup\u003e. Calibration, which evaluates the model\u0026apos;s ability to make accurate predictions close to actual events, was gauged using the Brier score. The Brier score calculates the mean square of the difference between predicted and observed probabilities, with a score of 0 indicating perfect calibration and 1 indicating the worst calibration \u003csup\u003e[\u003cspan class=\"CitationRef\"\u003e20\u003c/span\u003e]\u003c/sup\u003e. DCA was used to determine if the new model added value by increasing the net benefit across a realistic range of threshold probabilities \u003csup\u003e[\u003cspan class=\"CitationRef\"\u003e21\u003c/span\u003e]\u003c/sup\u003e. Furthermore, a sensitivity analysis was conducted to explore the relationship between dataset size and model performance, as the optimal amount of data needed to train a machine learning model is not yet established. All analyses and visualizations were performed using the Google Collaboratory Pro environment (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://research.google.com/colaboratory\u003c/span\u003e\u003c/span\u003e) with the Python 3.8.3 programming language (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.python.org\u003c/span\u003e\u003c/span\u003e) and the \u0026apos;scikit-survival\u0026apos; package (v. 0.16.1; \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://scikit-survival.readthedocs.io\u003c/span\u003e\u003c/span\u003e).\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec10\" class=\"Section2\"\u003e\n \u003ch2\u003eStatistical analysis\u003c/h2\u003e\n \u003cp\u003eWe utilized descriptive statistics to succinctly encapsulate our data, portraying categorical variables as absolute numerical counts and continuous variables as mean and standard deviation (SD). Our results were elucidated through hazard ratios, 95% confidence intervals (CI), and corresponding P values. A significance threshold of 0.05 was employed to ascertain statistical significance. This comprehensive analysis was executed using Python 3.8.3, R software (version 4.3.2), and SPSS 25.0.\u003c/p\u003e\n\u003c/div\u003e"},{"header":"Results","content":"\u003cdiv id=\"Sec12\" class=\"Section2\"\u003e \u003ch2\u003eCohort characteristics\u003c/h2\u003e \u003cp\u003eA total of 582 patients were included in this study (Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e). The incidence of gastric cancer in males was nearly five times higher than in females, regardless of whether they were in the training or validation cohort. The mean age of patients in the training and validation cohorts was 58.78\u0026thinsp;\u0026plusmn;\u0026thinsp;8.929 and 58.47\u0026thinsp;\u0026plusmn;\u0026thinsp;9.29 years, respectively. The mean maximum diameter of the tumor was 4.05\u0026thinsp;\u0026plusmn;\u0026thinsp;2.42 cm in the training cohort and 4.63\u0026thinsp;\u0026plusmn;\u0026thinsp;2.75 cm in the validation cohort. After a median follow-up period of 61 months (interquartile range 12\u0026ndash;120), 86 (14.8%) patients experienced recurrence. The recurrence rates were 14.6% in the training cohort and 15.3% in the validation cohort. Overall survival and progression-free survival were similar in both groups. All variables were well balanced between the training and validation cohorts (P\u0026thinsp;\u0026gt;\u0026thinsp;0.05). Patients were divided into training (n\u0026thinsp;=\u0026thinsp;405) and validation (n\u0026thinsp;=\u0026thinsp;177) cohorts, with both groups showing a balance in clinical and pathological variables, follow-up duration, and proteomic data (Refer to Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e3\u003c/span\u003e for Kaplan-Meier curves showing progression-free survival in the training and validation sets).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec13\" class=\"Section2\"\u003e \u003ch2\u003eMachine learning model evaluation and validation\u003c/h2\u003e \u003cp\u003eTable\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e illustrates the c-index values for predicting 3-year and 5-year PFS using classical and machine learning prognostic models in the validation cohort. Among the machine learning models, RSF exhibited the highest discrimination with a c-index of 0.936, followed by XG boosting (0.904) and DST (0.884). The AUC values further confirmed the superior discriminative performance of all machine learning models, with RSF leading (0.968), followed by XG boosting (0.945) and DST (0.924). Given the potential changes in patients' status over time, we assessed the performance of the models at various time points and observed a decrease in performance over time. The c-index accuracy peaked at 60 months for all machine learning models, while the prediction error was minimized at this time point, especially for RSF and XG boosting(Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e4\u003c/span\u003e). The integrated Brier score, used to evaluate model calibration over a 60-month period, indicated good calibration with low integrated Brier scores for all models (all \u0026lt;\u0026thinsp;0.1; Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eTo gain insights into the risk predictions provided by the models, we utilized SHAP values to extract global-level explanations of risk(Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e5\u003c/span\u003e\u0026ndash;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e7\u003c/span\u003e). The top three important features for predicting PFS in gastric cancer with dMMR remained consistent across all models: number of positive lymph nodes, Ki67, and pT stage. Given the exceptional accuracy of the RSF model (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e5\u003c/span\u003e), our analysis focused on this model. Importance scores of the top twenty factors were plotted to assess face validity, with variables higher on the graph considered more influential. The findings emphasized number of positive lymph nodes, Ki67, pT stage, Lauren classification, and maximum tumor diameter as the most discriminating factors for RSF. Notably, removing the number of positive lymph nodes from the model led to an average c-index drop of 0.06 points.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eOur study delved into the practicality of various models using Decision Curve Analysis (DCA) (Fig.\u0026nbsp;\u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e8\u003c/span\u003e) to assess their efficacy in guiding post-surgery care decisions, such as the choice between surveillance and adjuvant treatment. DCA visually showcases the clinical advantages of a model by mapping out a spectrum of risk thresholds (x-axis) and the net benefit of stratifying patients according to the model (y-axis) compared to assuming no relapse will occur in any patient. Our investigation revealed that DST models yielded lower benefits, while RSF models demonstrated the highest gains, consistently all machine learning models in both the training and validation cohorts (Fig.\u0026nbsp;\u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e8\u003c/span\u003e). Particularly notable was the performance of RSF models at risk thresholds between 0.2 and 0.5. The calibration curves depicted in Fig.\u0026nbsp;\u003cspan refid=\"Fig8\" class=\"InternalRef\"\u003e9\u003c/span\u003eA and \u003cspan refid=\"Fig8\" class=\"InternalRef\"\u003e9\u003c/span\u003eB showcase the strong correlation between projected and observed outcomes of all machine learning model. (Fig.\u0026nbsp;\u003cspan refid=\"Fig8\" class=\"InternalRef\"\u003e9\u003c/span\u003e).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec14\" class=\"Section2\"\u003e \u003ch2\u003eSensitivity analysis of machine learning model\u003c/h2\u003e \u003cp\u003eThe optimal amount of training data required to build an effective machine learning predictive model remains an ongoing topic of debate. To address this issue, we conducted a sensitivity analysis to assess the impact of dataset size on model performance. By utilizing the learning curve plot, we were able to visualize how the addition of more training and validation data influenced the accuracy of various models. Our findings revealed that the accuracy scores varied as the size of the training samples changed for both the training and validation cohorts. Among the models considered, XG boosting consistently demonstrated the highest sensitivity analysis, outperforming RSF and DST (Fig.\u0026nbsp;\u003cspan refid=\"Fig9\" class=\"InternalRef\"\u003e10\u003c/span\u003e). Additionally, we observed that the sensitivity scores for the training cohort were notably higher than those for the validation cohort. Specifically, in the case of RSF, the sensitivity score ranged from 0.83 to 0.97 for the training cohort and from 0.52 to 0.71 for the validation cohort. Similarly, for DST, the sensitivity score ranged from 0.82 to 0.95 for the training cohort and from 0.65 to 0.77 for the validation cohort. In contrast, for XG boosting, the sensitivity score ranged from 0.90 to 1.00 for the training cohort and from 0.60 to 0.75 for the validation cohort. Moreover, as the size of the sample increased, we observed an enhancement in the predictive accuracy of the machine learning models. This suggests that incorporating more data leads to greater stability in the model's performance. Overall, our sensitivity analysis highlights the importance of dataset size in optimizing the accuracy and effectiveness of machine learning predictive models.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec15\" class=\"Section2\"\u003e \u003ch2\u003eDesigning and Validating a Prediction nomogram Model for PFS\u003c/h2\u003e \u003cp\u003eA comprehensive analysis using Multivariate Cox regression was conducted to assess the factors influencing PFS in gastric cancer patients. The results from the study, which included variables such as age, positive lymph nodes, neural invasion, and Ki67, were particularly significant. This analysis was carried out on a group of 405 individuals with gastric cancer, and by incorporating these key factors into a nomogram model, accurate predictions of 3-year and 5-year PFS probabilities were made possible. The nomogram model accounts for various factors known to impact outcomes, allowing for precise estimation of a patient's PFS over these defined time periods. The effectiveness of the nomogram in forecasting positive outcomes for gastric cancer patients, considering variables affecting 3-year and 5-year PFS, is illustrated in Fig.\u0026nbsp;\u003cspan refid=\"Fig10\" class=\"InternalRef\"\u003e11\u003c/span\u003e. The evaluation of the nomogram's performance related to PFS in the training group revealed a strong predictive capability with a C-index of 0.808 (95% CI: 0.769\u0026ndash;0.847), surpassing the AJCC 8th edition TNM staging with a C-index of 0.757 (95%CI: 0.717\u0026ndash;0.797). Calibration curves depicted in Fig.\u0026nbsp;\u003cspan refid=\"Fig11\" class=\"InternalRef\"\u003e12\u003c/span\u003eA, \u003cspan refid=\"Fig11\" class=\"InternalRef\"\u003e12\u003c/span\u003eB, \u003cspan refid=\"Fig11\" class=\"InternalRef\"\u003e12\u003c/span\u003eC, and \u003cspan refid=\"Fig11\" class=\"InternalRef\"\u003e12\u003c/span\u003eD demonstrated a strong correlation between projected and observed outcomes, further validated internally and externally.The time-dependent ROC analysis during internal validation reaffirmed the nomogram's exceptional discriminatory abilities, with AUC values of 0.849 (95%CI: 0.705\u0026ndash;0.872) for 3-year PFS and 0.846 (95%CI: 0.800\u0026ndash;0.890) for 5-year PFS. External validation also yielded promising results, with AUC values of 0.830 (95%CI: 0.754\u0026ndash;0.892) for 3-year PFS and 0.804 (95%CI: 0.747\u0026ndash;0.928) for 5-year PFS (Fig.\u0026nbsp;\u003cspan refid=\"Fig12\" class=\"InternalRef\"\u003e13\u003c/span\u003eA, \u003cspan refid=\"Fig12\" class=\"InternalRef\"\u003e13\u003c/span\u003eB).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eTo determine the clinical utility of the nomogram, a decision analysis curve (DCA) was utilized to compare 5-year and 3-year PFS predictions between the nomogram and the AJCC 8th edition TNM staging. The internal validation C-index for the nomogram was superior at 0.808 (95%CI: 0.769\u0026ndash;0.847) compared to the AJCC 8th edition TNM staging at 0.757 (95%CI: 0.717\u0026ndash;0.797). Similarly, the external validation C-index for the nomogram outperformed the AJCC 8th edition TNM staging with a value of 0.755 (95%CI: 0.698\u0026ndash;0.811) compared to 0.729 (95%CI: 0.675\u0026ndash;0.783). The higher C-index values of the nomogram in both internal and external validations highlight its superior predictive accuracy over the AJCC 8th edition TNM staging (Fig.\u0026nbsp;\u003cspan refid=\"Fig13\" class=\"InternalRef\"\u003e14\u003c/span\u003eA, \u003cspan refid=\"Fig13\" class=\"InternalRef\"\u003e14\u003c/span\u003eB, \u003cspan refid=\"Fig13\" class=\"InternalRef\"\u003e14\u003c/span\u003eC, \u003cspan refid=\"Fig13\" class=\"InternalRef\"\u003e14\u003c/span\u003eD).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec16\" class=\"Section2\"\u003e \u003ch2\u003eRisk Assessment in a Stratified System for Prediction of PFS based on nomogram\u003c/h2\u003e \u003cp\u003eBy employing advanced nomogram modeling techniques, each patient underwent categorization and scoring to evaluate their risk of disease progression. The X-tile software was used in the initial cohort of 405 patients to determine the optimal cutoff value for PFS scores. Subsequent analysis using the log-rank test revealed significant differences in survival times across various risk groups. The prognostic nomogram was then applied to calculate overall scores, leading to the clear division of the 582 participants into low-risk and high-risk groups using a threshold value of 175.97. Visualization of the cohort in Fig.\u0026nbsp;\u003cspan refid=\"Fig14\" class=\"InternalRef\"\u003e15\u003c/span\u003eA, \u003cspan refid=\"Fig14\" class=\"InternalRef\"\u003e15\u003c/span\u003eB, and \u003cspan refid=\"Fig14\" class=\"InternalRef\"\u003e15\u003c/span\u003eC demonstrated the effective separation of individuals based on their likelihood of disease progression. The low-risk group, comprising 316 patients from the training set and 124 patients from the validation set, contrasted with the high-risk group, which included 89 patients from the training set and 53 patients from the validation set. Analysis of PFS curves in Fig.\u0026nbsp;\u003cspan refid=\"Fig14\" class=\"InternalRef\"\u003e15\u003c/span\u003e indicated highly significant differences with P values below 0.001. Notably, the median PFS for the low-risk group in both the training and validation cohorts was not reached, emphasizing their favorable prognosis. In contrast, the high-risk group exhibited median PFS values of 27, 24, and 33 months across the overall, training, and validation cohorts, respectively. These distinct outcomes underscore the robust predictive capability of the model. The plot further illustrated the variability in PFS among patients with dMMR status, culminating in the stratification of risk levels based on individual total scores. This comprehensive approach to risk assessment enhances our understanding of disease progression and aids in personalized treatment strategies for each patient.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e"},{"header":"Discussion","content":"\u003cp\u003eRadical gastrectomy and subsequent adjuvant chemotherapy have been shown to effectively treat early-stage gastric cancer \u003csup\u003e[\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e]\u003c/sup\u003e. Despite optimal multimodality therapy, 30\u0026ndash;40% of patients may experience a relapse within 5 years \u003csup\u003e[\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e]\u003c/sup\u003e. The delicate balance of nature is illuminated by the vital function of the mismatch repair system. This intricate cellular mechanism is pivotal in identifying and correcting any mispaired bases resulting from inaccuracies in DNA replication, genetic recombination, or external influences like exposure to toxins or injury\u003csup\u003e[\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e]\u003c/sup\u003e. The biological system involving enzymes like MLH1, MLH3, MSH2, MSH3, MSH6, PMS1, and PMS2 is essential for repairing DNA mismatches. During DNA replication, complexes such as MSH2/MSH6 and MSH2/MSH3 detect and bind to errors in DNA. The MLH1/PMS2 heterodimers then remove and replace incorrect DNA bases at these sites. Disruptions in the function or expression of these enzymes can result in deficient complexes that are unable to properly repair DNA mismatches.\u003c/p\u003e \u003cp\u003eAs accurately classifying and selecting patients for personalized treatment is crucial, numerous studies have focused on the relationship between dMMR and various pathological features\u003csup\u003e[\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e]\u003c/sup\u003e. Previous reports have already demonstrated the prognostic role of dMMR in gastric cancer\u003csup\u003e[\u003cspan additionalcitationids=\"CR27 CR28\" citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e]\u003c/sup\u003e. However, there is limited research on predicting the prognosis of dMMR cohorts in gastric cancer. In our study, we are the first to investigate different machine learning models to predict the prognosis of dMMR in Chinese gastric cancer patients and evaluate their merits and faults in terms of prediction ability. Our ultimate objective is to confirm the most precise machine learning model for dMMR patients based on the findings of this study while minimizing the negative effects of a limited sample size. Therefore, sensitivity analysis was applied to evaluate the stability of the model as more sample sizes were incorporated. Additionally, we conducted a comprehensive, long-term follow-up within the same study population, meticulously documenting their prognostic data.\u003c/p\u003e \u003cp\u003eIn our study, we utilized a combination of clinical features, pathological indicators, and tumor molecular markers to predict the progression-free survival (PFS) of deficient mismatch repair (dMMR) patients with gastric cancer. Our main goal was to evaluate different machine learning models for forecasting PFS in patients with dMMR. The detailed analysis for dMMR patients included a wide range of variables such as gender, age at surgery, presence of vascular and neural invasion, tumor stage, number of positive lymph nodes, Lauren classification, maximum tumor diameter, type of gastrectomy performed, omentum metastasis status, surgical margin outcomes, complication severity, and various proteomic markers associated with gastric cancer. Recognizing the potential impact of diverse factors on patient outcomes following radical gastrectomy in different medical settings, we conducted rigorous internal and external validations to assess the effectiveness of our model. These validations confirmed that our model provided strong predictive accuracy, calibration, discrimination, and clinical utility. Our holistic approach aims to offer valuable guidance for healthcare professionals and improve communication between patients and physicians. By incorporating multiple factors and conducting thorough validations, our model can assist clinicians in predicting PFS for dMMR patients, allowing for informed treatment decisions and personalized care plans.\u003c/p\u003e \u003cp\u003eCurrently, the TNM staging system is widely used as the main method for predicting the risk of clinical outcomes. However, its effectiveness and reliability are limited, leading to a significant decrease in its practical use \u003csup\u003e[\u003cspan additionalcitationids=\"CR31\" citationid=\"CR30\" class=\"CitationRef\"\u003e30\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e32\u003c/span\u003e]\u003c/sup\u003e. Numerous studies have shown that using bar charts in assessments can improve accuracy and reduce the need for unnecessary tests in patients with gastric cancer\u003csup\u003e[\u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e33\u003c/span\u003e]\u003c/sup\u003e. Research has focused on factors such as age, gender, tumor size, lymph node status, tumor depth, location, Lauren and histological classification, and biomarkers to predict outcomes. Prognostic models have been developed with training and validation groups consistently yielding significant results (P\u0026thinsp;\u0026lt;\u0026thinsp;0.001) \u003csup\u003e[\u003cspan additionalcitationids=\"CR35 CR36 CR37\" citationid=\"CR34\" class=\"CitationRef\"\u003e34\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e38\u003c/span\u003e]\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eOur team's innovative approach combines the transparency and reliability of the nomogram with the predictive power of machine learning models to create a comprehensive tool for prognosis in gastric cancer with dMMR. This hybrid system not only provides personalized and reliable prognostic information but also enhances shared decision-making between patients and clinicians. By accurately differentiating between different probabilities of progression-free survival in patients with dMMR, we can tailor treatment plans to individual patients' needs. High probability survival patients may not require additional postoperative treatment, while low probability survival patients with PD-L1 highly expressed conditions can benefit from targeted immunotherapies. This groundbreaking model has significant implications for personalized patient care and decision-making in oncology. By combining the strengths of both the nomogram and machine learning models, we can improve care, provide valuable information to patients, and aid clinicians in making management decisions for tongue cancer.\u003c/p\u003e \u003cp\u003eAlthough the study yielded promising results, it is important to acknowledge its limitations. Firstly, the model construction and validation only utilized data from a single center, which raises the need for validation using data from other medical centers to ensure the generalizability of the findings. Secondly, there seems to be a discrepancy between the effectiveness of the line plot model in predicting the three-year progression-free survival (PFS) and the actual data. This inconsistency warrants further investigation to understand the underlying reasons behind it. Thirdly, the study failed to differentiate between patients with early and late-stage gastric carcinoma, which may have led to varying predictive performance among patients at different stages of the disease.\u003c/p\u003e \u003cp\u003eWhen it comes to predicting patient outcomes in gastric cancer, it is generally observed that patients with dMMR tumors have a more favorable overall survival than those with pMMR tumors. This difference in prognosis could be due to the higher rate of immune infiltration in dMMR tumors, leading to increased tumor-infiltrating lymphocytes. Additionally, dMMR tumors are less likely to present adverse prognostic factors, such as lymphatic invasion and perineural invasion. These findings highlight the significant prognostic implications of MMR status in gastric cancer, not only in predicting patient outcomes but also in guiding treatment decisions. The presence of dMMR tumors in gastric cancer patients suggests a better response to immune checkpoint inhibitors, such as pembrolizumab. Pembrolizumab has already been approved for the treatment of MMR-deficient solid tumors, including gastric cancer. Therefore, understanding a patient's MMR status in gastric cancer becomes crucial in predicting prognosis and determining the most effective treatment options. In conclusion, recognizing the MMR status of gastric cancer patients plays a vital role in predicting their prognosis and selecting appropriate treatment strategies. The observation that dMMR tumors are associated with more favorable outcomes and higher responsiveness to immunotherapy highlights the importance of considering MMR status when managing gastric cancer patients.\u003c/p\u003e"},{"header":"Abbreviations","content":"\u003cp\u003eProgression free survival(PFS), deficient mismatch repair(dMMR), decision curve analysis (DCA), time-dependent receiver operating characteristic (t-ROC), Random Survival Forests (RSF), Decision Survival Tree (DST), Extreme Gradient Boosting (XG boosting)\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eEthics approval and consent to participate\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis retrospective study utilized clinical data collected for clinical purposes. Extensive consultation with the Ethics Committee of Shanxi Cancer Hospital was conducted, and ethical approval was granted by the Ethics Committee of Shanxi Carcinoma Hospital (No: 2022JC23). All methods were carried out in accordance with relevant guidelines and regulations. The procedures involving human participants were conducted in compliance with ethical standards set by our institutional research committee. Informed consent was obtained from all patients, indicating their voluntary and informed agreement to participate in the study. Patients also provided written consent for publication and any accompanying images. A copy of the consent form is available for review by the journal\u0026apos;s Editor-in-Chief upon request.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eRegistration of research\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eOur study has been officially recorded on the Research Registry under the identifying number Research Registry 9867 (https://researchregistry.knack.com/research-registry#user-researchregistry/ ).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthors\u0026rsquo; Contributions\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eYifan Li wrote the main manuscript text and prepared figures 1-15, while Jinfeng Ma and Wenhua Cheng reviewed and edited the manuscript.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConsent for publication\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNot applicable.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eData availability\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eData is provided within the manuscript or supplementary information files.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFunding\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eSupported by the Science and Education Cultivation Fund of the National Cancer and Regional Medical center of Shanxi Provincial Cancer hospital (SD2023005)\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n \u003cli\u003eSung H, Ferlay J, Siegel RL, Laversanne M, et al. Global Cancer Statistics 2020: GLOBOCAN Estimates of Incidence and Mortality Worldwide for 36 Cancers in 185 Countries. CA Cancer J Clin. 2021 May;71(3):209-249.\u0026nbsp;\u003c/li\u003e\n \u003cli\u003eAjani JA, Lee J, Sano T, et al. Gastric adenocarcinoma. Nat Rev Dis Primers. 2017 Jun 1;3:17036.\u003c/li\u003e\n \u003cli\u003eBourke MJ, Neuhaus H, Bergman JJ. Endoscopic Submucosal Dissection: Indications and Application in Western Endoscopy Practice. Gastroenterology. 2018 May;154(7):1887-1900.e5.\u003c/li\u003e\n \u003cli\u003eSalati M, Orsi G, Smyth E, et al. Gastric cancer: Translating novel concepts into clinical practice. Cancer Treat Rev. 2019 Sep;79:101889.\u0026nbsp;\u003c/li\u003e\n \u003cli\u003eKarimi P, Islami F, Anandasabapathy S, et al. Gastric cancer: descriptive epidemiology, risk factors, screening, and prevention. Cancer Epidemiol Biomarkers Prev. 2014 May;23(5):700-13.\u003c/li\u003e\n \u003cli\u003eBiagioni A, Skalamera I, Peri S, et al. Update on gastric cancer treatments and gene therapies. Cancer Metastasis Rev. 2019 Sep;38(3):537-548.\u0026nbsp;\u003c/li\u003e\n \u003cli\u003eSmyth EC, Wotherspoon A, Peckitt C, et al. Mismatch Repair Deficiency, Microsatellite Instability, and Survival: An Exploratory Analysis of the Medical Research Council Adjuvant Gastric Infusional Chemotherapy (MAGIC) Trial. JAMA Oncol. 2017 Sep 1;3(9):1197-1203.\u0026nbsp;\u003c/li\u003e\n \u003cli\u003eAndr\u0026eacute; T, Lonardi S, Wong KYM, et al. Nivolumab plus low-dose ipilimumab in previously treated patients with microsatellite instability-high/mismatch repair-deficient metastatic colorectal cancer: 4-year follow-up from CheckMate 142. Ann Oncol. 2022 Oct;33(10):1052-1060.\u0026nbsp;\u003c/li\u003e\n \u003cli\u003eOverman MJ, Lonardi S, Wong KYM, et al. Durable Clinical Benefit With Nivolumab Plus Ipilimumab in DNA Mismatch Repair-Deficient/Microsatellite Instability-High Metastatic Colorectal Cancer. J Clin Oncol. 2018 Mar 10;36(8):773-779.\u0026nbsp;\u003c/li\u003e\n \u003cli\u003eChoi YY, Kim H, Shin SJ, et al. Microsatellite Instability and Programmed Cell Death-Ligand 1 Expression in Stage II/III Gastric Cancer: Post Hoc Analysis of the CLASSIC Randomized Controlled Study. Ann Surg. 2019 Aug;270(2):309-316.\u0026nbsp;\u003c/li\u003e\n \u003cli\u003eTsai CY, Lin TA, Huang SC, et al. Is Adjuvant Chemotherapy Necessary for Patients with Deficient Mismatch Repair Gastric Cancer?-Autophagy Inhibition Matches the Mismatched. Oncologist. 2020 Jul;25(7):e1021-e1030.\u003c/li\u003e\n \u003cli\u003eZhao L, Fu Y, Niu P, Zhang F, et al. Perioperative Chemotherapy Could Not Improve the Prognosis of Gastric Cancer Patients With Mismatch Repair Deficiency: A Multicenter, Real-World Study. Oncologist. 2023 Oct 3;28(10):e891-e901\u003c/li\u003e\n \u003cli\u003eHwang TJ, Kesselheim AS, Vokinger KN. Lifecycle Regulation of Artificial Intelligence- and Machine Learning-Based Software Devices in Medicine. JAMA. 2019 Dec 17;322(23):2285-2286.\u0026nbsp;\u003c/li\u003e\n \u003cli\u003eLiu X, Faes L, Kale AU, Wagner SK, et al. A comparison of deep learning performance against health-care professionals in detecting diseases from medical imaging: a systematic review and meta-analysis. Lancet Digit Health. 2019 Oct;1(6):e271-e297.\u0026nbsp;\u003c/li\u003e\n \u003cli\u003eLee KH, Choi GH, Yun J, et al. Machine learning-based clinical decision support system for treatment recommendation and overall survival prediction of hepatocellular carcinoma: a multi-center study. NPJ Digit Med. 2024 Jan 5;7(1):2.\u0026nbsp;\u003c/li\u003e\n \u003cli\u003eXue B, Li D, Lu C, et al. Use of Machine Learning to Develop and Evaluate Models Using Preoperative and Intraoperative Data to Identify Risks of Postoperative Complications. JAMA Netw Open. 2021 Mar 1;4(3):e212240.\u0026nbsp;\u003c/li\u003e\n \u003cli\u003ePowell MA. Immunohistochemistry to determine mismatch repair-deficiency in endometrial cancer: the appropriate standard. Ann Oncol. 2017 Jan 1;28(1):9-10.\u003c/li\u003e\n \u003cli\u003eUno H, Cai T, Pencina MJ, et al. On the C-statistics for evaluating overall adequacy of risk prediction procedures with censored survival data. Stat Med. 2011 May 10;30(10):1105-17.\u0026nbsp;\u003c/li\u003e\n \u003cli\u003eHarrell FE Jr, Lee KL, Mark DB. Multivariable prognostic models: issues in developing models, evaluating assumptions and adequacy, and measuring and reducing errors. Stat Med. 1996 Feb 28;15(4):361-87.\u003c/li\u003e\n \u003cli\u003eLi Y, Zhang X. Prognostic nomograms for gastric carcinoma after surgery to assist decision-making for postoperative treatment with chemotherapy cycles \u0026lt;9 or chemotherapy cycles \u0026ge;9. Front Surg. 2022 Aug 26;9:916483.\u003c/li\u003e\n \u003cli\u003eLi Y, Bai M, Gao Y. Prognostic nomograms for gastric carcinoma after D2\u0026thinsp;+\u0026thinsp;total gastrectomy to assist decision-making for postoperative treatment: based on Lasso regression. World J Surg Oncol. 2023 Jul 20;21(1):207.\u003c/li\u003e\n \u003cli\u003ePark SH, Sohn TS, Lee J, et al. Phase III Trial to Compare Adjuvant Chemotherapy With Capecitabine and Cisplatin Versus Concurrent Chemoradiotherapy in Gastric Cancer: Final Report of the Adjuvant Chemoradiotherapy in Stomach Tumors Trial, Including Survival and Subset Analyses. J Clin Oncol. 2015 Oct 1;33(28):3130-6.\u0026nbsp;\u003c/li\u003e\n \u003cli\u003eNoh SH, Park SR, Yang HK, et al. Adjuvant capecitabine plus oxaliplatin for gastric cancer after D2 gastrectomy (CLASSIC): 5-year follow-up of an open-label, randomized phase 3 trial. Lancet Oncol. 2014 Nov;15(12):1389-96.\u0026nbsp;\u003c/li\u003e\n \u003cli\u003eRuiz-Ba\u0026ntilde;obre J, Goel A. DNA Mismatch Repair Deficiency and Immune Checkpoint Inhibitors in Gastrointestinal Cancers. Gastroenterology. 2019 Mar;156(4):890-903.\u0026nbsp;\u003c/li\u003e\n \u003cli\u003ePolom K, Marano L, Marrelli D, et al. Meta-analysis of microsatellite instability about clinicopathological characteristics and overall survival in gastric cancer. Br J Surg. 2018 Feb;105(3):159-167.\u0026nbsp;\u003c/li\u003e\n \u003cli\u003ePietrantonio F, Miceli R, Raimondi A, et al. Individual Patient Data Meta-Analysis of the Value of Microsatellite Instability As a Biomarker in Gastric Cancer. J Clin Oncol. 2019 Dec 10;37(35):3392-3400.\u0026nbsp;\u003c/li\u003e\n \u003cli\u003eQuaas A, Biesma HD, Wagner AD, et al. Microsatellite instability and sex differences in resectable gastric cancer - A pooled analysis of three European cohorts. Eur J Cancer. 2022 Sep;173:95-104.\u0026nbsp;\u003c/li\u003e\n \u003cli\u003eYu HY, Li CP, Huang YH, et al. Microsatellite Instability, Epstein-Barr Virus, and Programmed Cell Death Ligand 1 as Predictive Markers for Immunotherapy in Gastric Cancer. Cancers (Basel). 2022 Jan 3;14(1):218.\u0026nbsp;\u003c/li\u003e\n \u003cli\u003eKohlruss M, Ott K, Grosser B, et al. Sexual Difference Matters: Females with High Microsatellite Instability Show Increased Survival after Neoadjuvant Chemotherapy in Gastric Cancer. Cancers (Basel). 2021 Mar 2;13(5):1048.\u003c/li\u003e\n \u003cli\u003eWang F, Sun GP, Zou YF, et al. Helicobacter pylori infection predicts favorable outcomes in patients with gastric cancer. Curr Oncol. 2013; 20(5): e388\u0026ndash;e395.\u003c/li\u003e\n \u003cli\u003eLu J, Zheng ZF, Wang W, et al. A novel TNM staging system for gastric cancer based on the metro-ticket paradigm: a comparative study with the AJCC-TNM staging system. Gastric Cancer. 2019 Jul;22(4):759-768.\u0026nbsp;\u003c/li\u003e\n \u003cli\u003eZhou YY, Kang YT, Chen C, et al. Combination of TNM staging and pathway-based risk score models in patients with gastric cancer. J Cell Biochem. 2018 Apr;119(4):3608-3617.\u003c/li\u003e\n \u003cli\u003eLu J, Zheng ZF, Xie JW, et al. Is the 8th Edition of the AJCC TNM Staging System Sufficiently Reasonable for All Patients with Noncardia Gastric Cancer? A 12,549-Patient International Database Study. Ann Surg Oncol. 2018 Jul;25(7):2002-2011.\u003c/li\u003e\n \u003cli\u003eIasonos A, Schrag D, Raj GV, Panageas KS. How to build and interpret a nomogram for cancer prognosis. J Clin Oncol. 2008 Mar 10;26(8):1364-70.\u0026nbsp;\u003c/li\u003e\n \u003cli\u003eNecchi A, Sonpavde G, Lo Vullo S, et al. Nomogram-based Prediction of Overall Survival in Patients with Metastatic Urothelial Carcinoma Receiving First-line Platinum-based Chemotherapy: Retrospective International Study of Invasive/Advanced Cancer of the Urothelium (RISC). Eur Urol. 2017 Feb;71(2):281-289.\u0026nbsp;\u003c/li\u003e\n \u003cli\u003eYip PL, Lee SF, Choi CH, et al. External Validation of a Nomogram to Predict Survival and Benefit of Concurrent Chemoradiation for Stage II Nasopharyngeal Carcinoma. Cancers (Basel). 2021 Aug 25;13(17):4286.\u003c/li\u003e\n \u003cli\u003eRaghav K, Hwang H, J\u0026aacute;come AA, et al. Development and Validation of a Novel Nomogram for Individualized Prediction of Survival in Cancer of Unknown Primary. Clin Cancer Res. 2021 Jun 15;27(12):3414-3421.\u0026nbsp;\u003c/li\u003e\n \u003cli\u003eChang YR, Huang WK, Wang SY, et al. A Nomogram Predicting Progression Free Survival in Patients with Gastrointestinal Stromal Tumor Receiving Sunitinib: Incorporating Pre-Treatment and Post-Treatment Parameters. Cancers (Basel). 2021 May 25;13(11):2587.\u003c/li\u003e\n \u003cli\u003eLiu S, Yu X, Yang S, et al. Machine Learning-Based Radiomics Nomogram for Detecting Extramural Venous Invasion in Rectal Cancer. Front Oncol. 2021 Mar 26;11:610338.\u0026nbsp;\u003c/li\u003e\n \u003cli\u003eHuang L, Lin W, Xie D, et al. Development and validation of a preoperative CT-based radiomic nomogram to predict pathology invasiveness in patients with a solitary pulmonary nodule: a machine learning approach, multicenter, diagnostic study. Eur Radiol. 2022 Mar;32(3):1983-1996.\u003c/li\u003e\n \u003cli\u003eAlabi RO, M\u0026auml;kitie AA, Pirinen M, et al. Comparison of nomogram with machine learning techniques for prediction of overall survival in patients with tongue cancer. Int J Med Inform. 2021 Jan;145:104313.\u0026nbsp;\u003c/li\u003e\n\u003c/ol\u003e"},{"header":"Tables","content":"\u003cp\u003eTable1. Baseline of study population\u003c/p\u003e\n\u003ctable border=\"0\" cellspacing=\"0\" cellpadding=\"0\" width=\"100%\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eVariables\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eTraining cohort\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003e(n=405)\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003eMean\u0026plusmn;SD / No(%)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eValidation cohort\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003e(n=177)\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003eMean\u0026plusmn;SD /No (%)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp; \u0026nbsp; \u0026nbsp;\u003cem\u003eP\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eGender\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e0.843\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eMale\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e322(79.5%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e142(80.2%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eFemale\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e83(20.5%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e35(19.8%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eAge (years)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e58.78\u0026plusmn;8.93\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e58.47\u0026plusmn;9.29\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e0.677\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003epT stage\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e0.431\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eT1\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e155(38.3%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e66(37.3%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eT2\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e25(6.2%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e6(3.4%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eT3\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e89(22.0%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e20(21.5%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eT4\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e136(33.6%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e42(37.9%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"50%\" colspan=\"2\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eNumber of positive lymph nodes\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e0.138\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003e0\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e224(55.3%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e82(46.3%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003e1-2\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e76(18.8%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e33(18.6%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003e3-6\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e32(7.9%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e20(11.3%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026ge;7\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e73(18.0%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e42(23.7%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003epTNM stage\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e0.125\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eⅠ\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e174(43.0%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e67(37.9%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eⅡ\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e93(23.0%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e37(20.9%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eⅢ\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e138(34.1%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e73(41.2%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eVascular invasion\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e0.482\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eNegative\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e255(63.0%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e106(59.9%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003ePositive\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e150(37.0%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e71(40.1%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eNeural invasion\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e0.197\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eNegative\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e274(67.7%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e110(62.1%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003ePositive\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e131(32.3%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e67(37.9%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eLauren classification\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e0.951\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eIntestinal\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e222(54.8%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e96(54.2%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eDiffused\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e94(23.2%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e45(25.4%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eMixed\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e89(22.0%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e36(20.3%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eOverall survival(months)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e42.53\u0026plusmn;21.29\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e43.06\u0026plusmn;23.13\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e0.448\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eType of gastrectomy\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e0.103\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003ePrimal\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e61(15.1%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e17(9.6%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eDistal\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e158(39.0%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e69(39.0%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eTotal\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e185(45.7%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e90(50.8%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003ePPG\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e1(0.2%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e1(0.6%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eOmentum metastasis\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e0.295\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eNegative\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e402(99.3%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e174(98.7%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003ePositive\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e3(0.7%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e3(1.7%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eSurgical margin\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e0.477\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eNegative\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e386(95.3%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e171(96.6%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003ePositive\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e19(4.7%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e6(3.4%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eHer-2\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e0.913\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003e-\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e245(60.5%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e106(59.9%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003e+\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e134(33.1%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e60(33.9%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003e++\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e19(4.7%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e8(4.5%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003e+++\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e7(1.7%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e3(1.7%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eRecurrence\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e0.83\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eNegative\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e346(85.4%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e150(84.7%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003ePositive\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e59(14.6%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e27(15.3%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eProgress Free Survival(months)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e39.53\u0026plusmn;22.18\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e38.55\u0026plusmn;23.99\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e0.452\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eAE1/AE3\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e0.733\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eNegative\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e203(50.1%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e86(48.6%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003ePositive\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e202(49.9%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e91(51.4%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eKi67(%)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e37.83\u0026plusmn;29.80\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e41.73\u0026plusmn;29.64\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e0.115\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eCK7\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e0.572\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eNegative\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e239(59.0%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e100(56.5%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003ePositive\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e166(41.0%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e77(43.5%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eCK20\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e0.999\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eNegative\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e254(62.7%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e111(59.0%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003ePositive\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e151(37.3%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e66(41.0%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eCDX-2\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e0.557\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eNegative\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e271(66.9%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e114(64.4%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003ePositive\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e134(33.1%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e63(35.6%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eSATB-2\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e0.404\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eNegative\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e344(84.9%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e155(87.6%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003ePositive\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e61(15.1%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e22(12.4%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eSYN\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e0.166\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eNegative\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e280(69.1%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e112(63.3%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003ePositive\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e125(30.9%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e65(36.7%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eCGA\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e0.853\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eNegative\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e276(68.1%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e122(68.9%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003ePositive\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e129(31.9%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e55(31.1%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eCD56\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e0.804\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eNegative\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e305(75.3%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e135(76.3%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003ePositive\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e100(24.7%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e42(23.7%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eMLH1 \u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e0.333\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eNegative\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e152(37.5%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e59(33.3%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003ePositive\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e253(62.5%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e118(66.7%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003ePMS2\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e0.896\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eNegative\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e320(79.0%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e139(78.5%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003ePositive\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e85(21.0%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e38(21.5%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eMSH2\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e0.654\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eNegative\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e159(39.3%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e66(37.3%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003ePositive\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e246(60.7%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e111(62.7%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eMSH6\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e0.48\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eNegative\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e152(37.5%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e61(34.5%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003ePositive\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e253(62.5%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e116(65.5%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eMaximum diameter of tumor(cm)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e4.05\u0026plusmn;2.42\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e4.63\u0026plusmn;2.75\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e0.14\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eTumor location\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e0.981\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eUpper 1/3\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e187(46.2%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e79(44.6%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eMiddle 1/3\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e68(16.8%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e34(19.2%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eLower 1/3\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e147(36.3%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e63(35.6%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eMultiple\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e3(0.7%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e1(0.6%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"25%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003e\u003cbr\u003e\u003c/p\u003e\n\u003cp\u003eTable 2 \u0026nbsp;Discrimination and calibration of each model predicting progression-free survival at 60 months\u003c/p\u003e\n\u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\" width=\"100%\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd width=\"34.02061855670103%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eMachine learning model\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"11.34020618556701%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eC-Index\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"15.463917525773196%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003e95%CI\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"8.24742268041237%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eAUC\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"15.463917525773196%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003e95%CI\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"15.463917525773196%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eBrier score\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"34.02061855670103%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eRadom survival forest\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"11.34020618556701%\" valign=\"top\"\u003e\n \u003cp\u003e0.936\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"15.463917525773196%\" valign=\"top\"\u003e\n \u003cp\u003e0.935-0.937\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"8.24742268041237%\" valign=\"top\"\u003e\n \u003cp\u003e0.968\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"15.463917525773196%\" valign=\"top\"\u003e\n \u003cp\u003e0.966-0.968\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"15.463917525773196%\" valign=\"top\"\u003e\n \u003cp\u003e0.067\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"34.02061855670103%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eDecision survival tree\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"11.34020618556701%\" valign=\"top\"\u003e\n \u003cp\u003e0.884\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"15.463917525773196%\" valign=\"top\"\u003e\n \u003cp\u003e0.883-0.887\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"8.24742268041237%\" valign=\"top\"\u003e\n \u003cp\u003e0.924\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"15.463917525773196%\" valign=\"top\"\u003e\n \u003cp\u003e0.924-0.926\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"15.463917525773196%\" valign=\"top\"\u003e\n \u003cp\u003e0.030\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"34.02061855670103%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eGradient Boosting survival\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"11.34020618556701%\" valign=\"top\"\u003e\n \u003cp\u003e0.904\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"15.463917525773196%\" valign=\"top\"\u003e\n \u003cp\u003e0.904-0.907\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"8.24742268041237%\" valign=\"top\"\u003e\n \u003cp\u003e0.945\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"15.463917525773196%\" valign=\"top\"\u003e\n \u003cp\u003e0.944-0.946\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"15.463917525773196%\" valign=\"top\"\u003e\n \u003cp\u003e0.029\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003e\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eTable 3 Multivariate analysis of PFS of training cohort of dMMR and analyzed by Cox regression\u003c/p\u003e\n\u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\" width=\"100%\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd width=\"29.18918918918919%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"9.36936936936937%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eB\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"9.36936936936937%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eSE\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"10.45045045045045%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eWald\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"5.945945945945946%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003edf\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"10.45045045045045%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eP\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"9.36936936936937%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eHR\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"15.855855855855856%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eHR (95%CI)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"29.18918918918919%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eAge\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"9.36936936936937%\" valign=\"top\"\u003e\n \u003cp\u003e0.023\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"9.36936936936937%\" valign=\"top\"\u003e\n \u003cp\u003e0.011\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"10.45045045045045%\" valign=\"top\"\u003e\n \u003cp\u003e4.683\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"5.945945945945946%\" valign=\"top\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"10.45045045045045%\" valign=\"top\"\u003e\n \u003cp\u003e0.030\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"9.36936936936937%\" valign=\"top\"\u003e\n \u003cp\u003e1.023\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"15.855855855855856%\" valign=\"top\"\u003e\n \u003cp\u003e1.002-1.045\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"38.44765342960289%\" colspan=\"2\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eNumber of positive lymph nodes\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"9.386281588447654%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"10.469314079422382%\" valign=\"top\"\u003e\n \u003cp\u003e46.954\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"5.956678700361011%\" valign=\"top\"\u003e\n \u003cp\u003e3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"10.469314079422382%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026lt;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"9.386281588447654%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"15.884476534296029%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"29.18918918918919%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003e0 Vs 1-2\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"9.36936936936937%\" valign=\"top\"\u003e\n \u003cp\u003e0.740\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"9.36936936936937%\" valign=\"top\"\u003e\n \u003cp\u003e0.302\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"10.45045045045045%\" valign=\"top\"\u003e\n \u003cp\u003e6.015\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"5.945945945945946%\" valign=\"top\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"10.45045045045045%\" valign=\"top\"\u003e\n \u003cp\u003e0.014\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"9.36936936936937%\" valign=\"top\"\u003e\n \u003cp\u003e2.097\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"15.855855855855856%\" valign=\"top\"\u003e\n \u003cp\u003e1.160-3.788\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"29.18918918918919%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003e0 Vs 3-6\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"9.36936936936937%\" valign=\"top\"\u003e\n \u003cp\u003e0.699\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"9.36936936936937%\" valign=\"top\"\u003e\n \u003cp\u003e0.381\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"10.45045045045045%\" valign=\"top\"\u003e\n \u003cp\u003e3.375\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"5.945945945945946%\" valign=\"top\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"10.45045045045045%\" valign=\"top\"\u003e\n \u003cp\u003e0.066\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"9.36936936936937%\" valign=\"top\"\u003e\n \u003cp\u003e2.013\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"15.855855855855856%\" valign=\"top\"\u003e\n \u003cp\u003e0.954-4.245\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"29.18918918918919%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003e0 VS \u0026ge;7\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"9.36936936936937%\" valign=\"top\"\u003e\n \u003cp\u003e1.817\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"9.36936936936937%\" valign=\"top\"\u003e\n \u003cp\u003e0.292\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"10.45045045045045%\" valign=\"top\"\u003e\n \u003cp\u003e38.648\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"5.945945945945946%\" valign=\"top\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"10.45045045045045%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026lt;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"9.36936936936937%\" valign=\"top\"\u003e\n \u003cp\u003e6.153\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"15.855855855855856%\" valign=\"top\"\u003e\n \u003cp\u003e3.470-10.911\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"29.18918918918919%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eNeural invasion\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"9.36936936936937%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"9.36936936936937%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"10.45045045045045%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"5.945945945945946%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"10.45045045045045%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"9.36936936936937%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"15.855855855855856%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"29.18918918918919%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eNegative Vs Positive\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"9.36936936936937%\" valign=\"top\"\u003e\n \u003cp\u003e0.570\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"9.36936936936937%\" valign=\"top\"\u003e\n \u003cp\u003e0.198\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"10.45045045045045%\" valign=\"top\"\u003e\n \u003cp\u003e8.262\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"5.945945945945946%\" valign=\"top\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"10.45045045045045%\" valign=\"top\"\u003e\n \u003cp\u003e0.004\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"9.36936936936937%\" valign=\"top\"\u003e\n \u003cp\u003e1.769\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"15.855855855855856%\" valign=\"top\"\u003e\n \u003cp\u003e1.199-2.610\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"29.18918918918919%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eKi67\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"9.36936936936937%\" valign=\"top\"\u003e\n \u003cp\u003e0.014\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"9.36936936936937%\" valign=\"top\"\u003e\n \u003cp\u003e0.004\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"10.45045045045045%\" valign=\"top\"\u003e\n \u003cp\u003e9.438\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"5.945945945945946%\" valign=\"top\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"10.45045045045045%\" valign=\"top\"\u003e\n \u003cp\u003e0.002\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"9.36936936936937%\" valign=\"top\"\u003e\n \u003cp\u003e1.014\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"15.855855855855856%\" valign=\"top\"\u003e\n \u003cp\u003e1.005-1.023\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003eAbbreviations: B, regression coefficient; SE, standard error;\u003cem\u003e\u0026nbsp;df,\u003c/em\u003e degree of freedom; HR, hazard ratio; CI, confidence interval\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;\u003c/p\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":true,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"bmc-cancer","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"bcan","sideBox":"Learn more about [BMC Cancer](http://bmccancer.biomedcentral.com/)","snPcode":"","submissionUrl":"https://www.editorialmanager.com/bcan/default.aspx","title":"BMC Cancer","twitterHandle":"BMC_series","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"em","reportingPortfolio":"BMC Series","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"gastric carcinoma, mismatch repair, progression-free survival, machine learning, nomogram","lastPublishedDoi":"10.21203/rs.3.rs-4639290/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-4639290/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003ch2\u003eObjective\u003c/h2\u003e \u003cp\u003eTo assess the effectiveness of a machine learning framework and nomogram in predicting progression-free survival (PFS) post radical gastrectomy in patients with dMMR.\u003c/p\u003e\u003ch2\u003eMethod\u003c/h2\u003e \u003cp\u003eAn observational study conducted at Shanxi Cancer Hospital from 2002 to 2020 focused on developing and evaluating three machine learning models and nomogram to forecast PFS in patients undergoing radical gastrectomy for nonmetastatic gastric cancer with dMMR. Independent risk factors were identified using Cox regression analysis to develop the nomogram. The performance of the models was assessed through C-index, time receiver operating characteristic (T-ROC) curves, calibration curves, and decision curve analysis (DCA) curves in both training and validation cohorts. Subsequently, patients were categorized into high-risk and low-risk groups based on the nomogram's risk scores.\u003c/p\u003e\u003ch2\u003eResults\u003c/h2\u003e \u003cp\u003eAmong the 582 patients studied, machine learning models exhibited higher c-index values compared to the nomogram. RSF demonstrated the highest c-index (0.968), followed by XG boosting (0.945), DST (0.924), the nomogram (0.808), and 8th TNM staging (0.757). Age, positive lymph nodes, neural invasion, and Ki67 were identified as key factors and integrated into the prognostic nomogram. Calibration and DCA curves provided evidence of the accuracy and clinical benefits of both machine learning and nomogram models.\u003c/p\u003e\u003ch2\u003eConclusion\u003c/h2\u003e \u003cp\u003eOur study first successfully developed and validated machine learning and nomogram model based on clinical parameters for predicting 3-, 5-year PFS among dMMR gastric patients following gastrectomy. The nomogram exhibited a remarkable capability in identifying high-risk patients, furnishing clinicians with invaluable insights for postoperative surveillance and tailored therapeutic interventions.\u003c/p\u003e","manuscriptTitle":"Construction and validation the first prognostic models of progression-free survival in gastric cancer patients after gastrectomy with deficient mismatch repair: Nomogram and three machine learning models approaches","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2024-07-19 19:51:07","doi":"10.21203/rs.3.rs-4639290/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Revision requested","date":"2024-06-27T08:56:49+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2024-06-27T00:54:12+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2024-06-27T00:52:55+00:00","index":"","fulltext":""},{"type":"submitted","content":"BMC Cancer","date":"2024-06-26T02:13:37+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"bmc-cancer","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"bcan","sideBox":"Learn more about [BMC Cancer](http://bmccancer.biomedcentral.com/)","snPcode":"","submissionUrl":"https://www.editorialmanager.com/bcan/default.aspx","title":"BMC Cancer","twitterHandle":"BMC_series","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"em","reportingPortfolio":"BMC Series","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"1b00424c-c23d-4ab1-a836-66de544a9565","owner":[],"postedDate":"July 19th, 2024","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"published-in-journal","subjectAreas":[],"tags":[],"updatedAt":"2025-01-27T16:00:20+00:00","versionOfRecord":{"articleIdentity":"rs-4639290","link":"https://doi.org/10.1186/s12885-025-13542-0","journal":{"identity":"bmc-cancer","isVorOnly":false,"title":"BMC Cancer"},"publishedOn":"2025-01-24 15:57:12","publishedOnDateReadable":"January 24th, 2025"},"versionCreatedAt":"2024-07-19 19:51:07","video":"","vorDoi":"10.1186/s12885-025-13542-0","vorDoiUrl":"https://doi.org/10.1186/s12885-025-13542-0","workflowStages":[]},"version":"v1","identity":"rs-4639290","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-4639290","identity":"rs-4639290","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}