Random Survival Forest Model Research Articles

Avelumab first-line maintenance in advanced urothelial carcinoma: Complete screening for prognostic and predictive factors using machine learning in the JAVELIN Bladder 100 phase 3 trial.

Avelumab first-line (1 L) maintenance is a standard of care for advanced urothelial carcinoma (aUC) based on the JAVELIN Bladder 100 phase 3 trial, which showed that avelumab 1 L maintenance + best supportive care (BSC) significantly prolonged overall survival (OS) and progression-free survival (PFS) vs BSC alone in patients who were progression free after receiving 1 L platinum-containing chemotherapy. Here, we comprehensively screened JAVELIN Bladder 100 trial datasets to identify prognostic factors that define subpopulations of patients with longer or shorter OS irrespective of treatment, and predictive factors that select patients who could obtain a greater OS benefit from avelumab 1 L maintenance treatment. We performed machine learning analyses to screen a large set of baseline covariates, including patient demographics, disease characteristics, laboratory values, molecular biomarkers, and patient-reported outcomes. Covariates were identified from previously reported analyses and established prognostic and predictive markers. Variables selected from random survival forest models were processed further in univariate Cox models with treatment interaction and visually inspected using correlation analysis and Kaplan-Meier curves. Results were summarized in a multivariable Cox model. Prognostic baseline covariates associated with OS included in the final model were assignment to avelumab 1 L maintenance treatment, Eastern Cooperative Oncology Group performance status, site of metastasis, sum of longest target lesion diameters, levels of C-reactive protein and alkaline phosphatase in blood, lymphocyte proportion in intratumoral stroma, tumor mutational burden, and tumor CD8+ T-cell infiltration. Potential predictive factors included site of metastasis, tumor mutation burden, and tumor CD8+ T-cell infiltration. An analysis in patients with PD-L1+ tumors had similar findings to those in the overall population. Machine learning analyses of data from the JAVELIN Bladder 100 trial identified potential prognostic and predictive factors for avelumab 1 L maintenance treatment in patients with aUC, which warrant further evaluation in other clinical datasets.

Mortality prediction and influencing factors for intensive care unit patients with acute tubular necrosis: random survival forest and cox regression analysis

Background: Patients with acute tubular necrosis (ATN) not only have severe renal failure, but also have many comorbidities, which can be life-threatening and require timely treatment. Identifying the influencing factors of ATN and taking appropriate interventions can effectively shorten the duration of the disease to reduce mortality and improve patient prognosis.Methods: Mortality prediction models were constructed by using the random survival forest (RSF) algorithm and the Cox regression. Next, the performance of both models was assessed by the out-of-bag (OOB) error rate, the integrated brier score, the prediction error curve, and area under the curve (AUC) at 30, 60 and 90 days. Finally, the optimal prediction model was selected and the decision curve analysis and nomogram were established.Results: RSF model was constructed under the optimal combination of parameters (mtry = 10, nodesize = 88). Vasopressors, international normalized ratio (INR)_min, chloride_max, base excess_min, bicarbonate_max, anion gap_min, and metastatic solid tumor were identified as risk factors that had strong influence on mortality in ATN patients. Uni-variate and multivariate regression analyses were used to establish the Cox regression model. Nor-epinephrine, vasopressors, INR_min, severe liver disease, and metastatic solid tumor were identified as important risk factors. The discrimination and calibration ability of both predictive models were demonstrated by the OOB error rate and the integrated brier score. However, the prediction error curve of Cox regression model was consistently lower than that of RSF model, indicating that Cox regression model was more stable and reliable. Then, Cox regression model was also more accurate in predicting mortality of ATN patients based on the AUC at different time points (30, 60 and 90 days). The analysis of decision curve analysis shows that the net benefit range of Cox regression model at different time points is large, indicating that the model has good clinical effectiveness. Finally, a nomogram predicting the risk of death was created based on Cox model.Conclusion: The Cox regression model is superior to the RSF algorithm model in predicting mortality of patients with ATN. Moreover, the model has certain clinical utility, which can provide clinicians with some reference basis in the treatment of ATN and contribute to improve patient prognosis.

Activated partial thromboplastin time predicts mortality in patients with severe fever with thrombocytopenia syndrome: A multicenter study in north China

BackgroundSevere fever with thrombocytopenia syndrome (SFTS) is an emerging infectious disease with high lethality. This study aimed to determine whether prolonged activated partial thromboplastin time (APTT) predicted SFTS mortality. MethodsSFTS patients were enrolled from 6 hospitals in the north China. Subjects were divided into training cohort and 5 externally validation cohorts. The least absolute shrinkage and selection operator Cox regression model was performed to screen potential prognostic factors. Risk factors were analyzed using multivariable regression models. Prognostic models were established by Cox regression and random survival forest (RSF) methods, and evaluated regarding discrimination, validity and clinical benefit. Time-dependent receiver operating characteristic (ROC) curve was used to evaluate the predictive effectiveness of variables. Results1332 SFTS cases were included, in which 211 patients died. Six potential prognostic factors were screened, and pulse, breath, APTT and aspartic transaminase (AST) were independently associated with mortality in both training cohort (Yantai, N = 791) and external validation cohort (N = 541). APTT was steadily correlated with the fatality (HR: 1.039–1.144; all P < 0.01) in each five sub-validation cohorts (Dandong, Dalian, Tai'an, Qingdao and Beijing). RSF model with variables of APTT, AST, pulse and breath had considerable prognostic effectiveness, which APTT showed the highest prognostic ability with the area under the curve of 0.848 and 0.787 for 7-day and 14-day survival, respectively. Survival differences were found between high and low levels of APTT for mortality using 50s as the optimal cut-off. ConclusionsSFTS patients have prolonged APTT, which is an independent risk factor for fatality. APTT≥50s was recommended as a biomarker to remind physicians to monitor and treat patients more aggressively to improve clinical prognosis.

Effect of chemotherapy on prognosis in patients with primary pancreatic signet ring cell carcinoma: A large real-world study based on machine learning.

Primary pancreatic signet ring cell carcinoma (PSRCC), an extremely rare histologic variant of pancreatic cancer, has a poor prognosis. This study aimed to investigate the prognostic value of chemotherapy in PSRCC. Patients with PSRCC between 2000 and 2019 were identified using the Surveillance Epidemiology and End Results (SEER) database. The main outcomes in this study were cancer-specific survival (CSS) and overall survival (OS). The baseline characteristics of patients were compared using Pearson's Chi-square test. Kaplan-Meier analysis was used to generate the survival curves. Least absolute shrinkage and selection operator (LASSO), univariate and multivariate Cox regression models, and Random Survival Forest model were used to analyze the prognostic variables for OS and CSS. The variance inflation factors (VIFs) were used to analyze whether there was an overfitting problem. A total of 588 patients were identified. Chemotherapy was an independent prognostic factor for OS and CSS, and significantly associated with OS (HR = 0.33, 95% CI = 0.27-0.40, P <0.001) and CSS (HR = 0.32, 95% CI = 0.26-0.39, P <0.001). Chemotherapy showed beneficial effects on OS and CSS in patients with PSRCC and should be recommended in clinical practice.

Analysis of risk characteristics for early progression and late progression in locally advanced rectal cancer patients: a large population-based and validated study.

The current study aimed to explore the factors influencing early progression (EP) and late progression (LP) in locally advanced rectal cancer (LARC) patients. The patients were classified into EP and LP groups using one year as a cutoff. The random survival forest model was utilized to calculate the probability of time-to-progression. Besides, inverse probability of treatment weighting (IPTW) analysis and the Surveillance, Epidemiology, and End Results (SEER) were conducted to validate our results. Our study revealed that PNI, CEA level, and pathological stage were independent prognostic factors for PFS both in EP group and LP group. For EP group patients, Group 1 had the highest probability of progression at the 9th month of follow-up, while Group 2 exhibited the highest probability at the 6th month. Group 3, on the other hand, showed two peaks of progression at the 4th and 8th months of follow-up. As for LP group patients, Groups 4, 5, and 6 all exhibited peaks of progression between the 18th and 24th months of follow-up. Furthermore, our results suggested that PNI was also an independent prognostic factor affecting OS in both EP group and LP group. Finally, the analysis of IPTW and SEER database further confirmed our findings. Our results indicated a significant correlation between immune and nutritional status with PFS and OS in both EP and LP groups. These insights can aid healthcare professionals in effectively identifying and evaluating patients' nutritional status, enabling them to develop tailored nutrition plans and interventions.

Development and Validation of a Novel Machine Learning Model to Predict the Survival of Patients with Gastrointestinal Neuroendocrine Neoplasms

Introduction: Well-calibrated models for personalized prognostication of patients with gastrointestinal neuroendocrine neoplasms (GINENs) are limited. This study aimed to develop and validate a machine-learning model to predict the survival of patients with GINENs. Methods: Oblique random survival forest (ORSF) model, Cox proportional hazard risk model, Cox model with least absolute shrinkage and selection operator penalization, CoxBoost, Survival Gradient Boosting Machine, Extreme Gradient Boosting survival regression, DeepHit, DeepSurv, DNNSurv, logistic-hazard model, and PC-hazard model were compared. We further tuned hyperparameters and selected variables for the best-performing ORSF. Then, the final ORSF model was validated. Results: A total of 43,444 patients with GINENs were included. The median (interquartile range) survival time was 53 (19–102) months. The ORSF model performed best, in which age, histology, M stage, tumor size, primary tumor site, sex, tumor number, surgery, lymph nodes removed, N stage, race, and grade were ranked as important variables. However, chemotherapy and radiotherapy were not necessary for the ORSF model. The ORSF model had an overall C index of 0.86 (95% confidence interval, 0.85–0.87). The area under the receiver operation curves at 1, 3, 5, and 10 years were 0.91, 0.89, 0.87, and 0.80, respectively. The decision curve analysis showed superior clinical usefulness of the ORSF model than the American Joint Committee on Cancer Stage. A nomogram and an online tool were given. Conclusion: The machine learning ORSF model could precisely predict the survival of patients with GINENs, with the ability to identify patients at high risk for death and probably guide clinical practice.

Abstract PO2-01-14: Development of Artificial Intelligence-based Machine Learning Models for Predicting Survival In Hormone-Receptor-Positive/HER2-Negative Early Breast Cancer undergoing Neoadjuvant Chemotherapy

Abstract Background. In hormone-receptor positive (HoR)/HER2 negative early breast cancer (BC) multiple efforts have been made to predict disease recurrence and survival. Machine learning techniques have been used but few studies have looked into their applicability in predicting survival on the basis of clinico-pathological characteristics. The aim of this study was to evaluate a random survival forest model to predict the prognosis of this specific BC subtype. Methods. In this multicenter, retrospective study, patients who fulfilled the following inclusion criteria were included: (1) diagnosis of pathologically confirmed HoR-positive/HER2-negative invasive BC; (2) early or locally advanced, stage I-II-III (3) patients receiving neoadjuvant anthracycline and/or taxane-based chemotherapy, concurrently or sequentially; (4) patients undergoing surgery for primary BC. Survival endpoints were disease-free survival (DFS) and overall survival (OS). A random survival forest algorithm was used to develop the predictive model. Ten-fold cross-validation was performed. The C-index and the continuous rank probability score (CRPS) were used to evaluate the discrimination of the predictive model, and a ROC curve was used to evaluate model precision. A cut point analysis based on maximally selected rank statistics was conducted to evaluate the best cut-off in the out-of-bag (OOB) mortality that could maximize DFS and OS prediction. Variable importance was assessed using Breiman-Cutler permutation importance. Results. Overall, 572 patients with HoR-positive/HER2-negative early BC were included. At univariate analysis age, T stage, N stage, grading, ER, PR and Ki67 were found to be significantly associated with DFS and ER, PgR, HER2, pCR ypT and ypN retained statistical significance at multivariate analysis. ER, pCR, pathological T and N stage were found to be significantly associated with OS at univariate and multivariate analysis. The following variables were included in the final model: menopausal status, age, histology, grade, clinical T/N, ER/PgR, Ki67 and HER2, pathological complete response (pCR), ypT and ypN. For DFS, the cross-validated C-index was 0.68 (95% CI 0.63-0.73) and the OOB CRPS was 0.15, with a OOB performance error of 0.31. The AUC calculated at 60 months was 0.91 (95% CI 0.88-0.94). Out-of bag RSF-based risk scores for individual patients were calculated and an optimal cut-off of 22.58 was identified. The HR between high risk and low risk group was 3.08 (95% CI 2.16-4.39, p&amp;lt; 0.001). For OS, the cross-validated C-index was 0.66 (95% CI 0.60-0.71) and the OOB CRPS was 0.17, with a OOB performance error of 0.35. The AUC calculated at 60 months 0.91 (95% CI 0.88-0.95). An optimal cut-off of 21.54 was identified and the HR between low risk and high risk group was 2.50 (95% CI 1.73-3.61, p&amp;lt; 0.001). For DFS, the most important variables were cN (9.00), ypN (7.35), ER (7.21), ypT (5.38), PgR (4.11), pCR (4.01) and age (2.82) while for OS the most important variables were cN (7.88), ER (5.37), PgR (3.61), age (3.37), pCR (2.80), ypT (2.64) and ypN (2.60). Discussion. We analyzed the performance of RSF in the prediction of DFS and OS based on the contribution of clinico-pathological features commonly available at the baseline of the NCT and post-surgery. By selecting patients with HoR-positive HER2-negative disease, we were able to show which clinico-pathological features within a ML model have greater predictive importance. The model could be integrated with image-based tumor and radiology features to improve its predictive accuracy. This retrospective multicenter study suggests that the combination of easily accessible clinico-pathological features within a ML model may reliably predict DFS and OS in the context of HoR-positive/HER2-negative BCs. Citation Format: Luca Mastrantoni, Giovanna Garufi, Noemi Maliziola, Elena Di Monte, Giorgia Arcuri, Valentina Frescura, Angelachiara Rotondi, Giulia Giordano, Luisa Carbognin, Alessandra Fabi, Ida Paris, Gianluca Franceschini, Armando Orlandi, Antonella Palazzo, Giovanni Scambia, Giampaolo Tortora, Emilio Bria. Development of Artificial Intelligence-based Machine Learning Models for Predicting Survival In Hormone-Receptor-Positive/HER2-Negative Early Breast Cancer undergoing Neoadjuvant Chemotherapy [abstract]. In: Proceedings of the 2023 San Antonio Breast Cancer Symposium; 2023 Dec 5-9; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2024;84(9 Suppl):Abstract nr PO2-01-14.

Preoperatively predicting survival outcome for clinical stage IA pure-solid non–small cell lung cancer by radiomics-based machine learning

ObjectiveClinical stage IA non–small cell lung cancer (NSCLC) showing a pure-solid appearance on computed tomography is associated with a worse prognosis. This study aimed to develop and validate machine-learning models using preoperative clinical and radiomic features to predict overall survival (OS) in clinical stage IA pure-solid NSCLC. MethodsPatients who underwent lung resection for NSCLC between January 2012 and December 2020 were reviewed. The radiomic features were extracted from the intratumoral and peritumoral regions on computed tomography. The machine-learning models were developed using random survival forest and eXtreme Gradient Boosting (XGBoost) algorithms, whereas the Cox regression model was set as a benchmark. Model performance was assessed using the integrated time-dependent area under the curve (iAUC) and validated by 5-fold cross-validation. ResultsIn total, 642 patients with clinical stage IA pure-solid NSCLC were included. Among 3748 radiomic and 34 preoperative clinical features, 42 features were selected. Both machine-learning models outperformed the Cox regression model (iAUC, 0.753; 95% confidence interval [CI], 0.629-0.829). The XGBoost model showed a better performance (iAUC, 0.832; 95% CI, 0.779-0.880) than the random survival forest model (iAUC, 0.795; 95% CI, 0.734-0.856). The XGBoost model showed an excellent survival stratification performance with a significant OS difference among the low-risk (5-year OS, 100.0%), moderate low-risk (5-year OS, 88.5%), moderate high-risk (5-year OS, 75.6%), and high-risk (5-year OS, 41.7%) groups (P < .0001). ConclusionsA radiomics-based machine-learning model can preoperatively and accurately predict OS and improve survival stratification in clinical stage IA pure-solid NSCLC.

Integration of multi-omics data for survival prediction of lung adenocarcinoma

Background and ObjectiveThe morbidity of lung adenocarcinoma (LUAD) has been increasing year by year and the prognosis is poor. This has prompted researchers to study the survival of LUAD patients to ensure that patients can be cured in time or survive after appropriate treatment. There is still no fully valid model that can be applied to clinical practice. MethodsWe introduced struc2vec-based multi-omics data integration (SBMOI), which could integrate gene expression, somatic mutations and clinical data to construct mutation gene vectors representing LUAD patient features. Based on the patient features, the random survival forest (RSF) model was used to predict the long- and short-term survival of LUAD patients. To further demonstrate the superiority of SBMOI, we simultaneously replaced scale-free gene co-expression network (FCN) with a protein-protein interaction (PPI) network and a significant co-expression network (SCN) to compare accuracy in predicting LUAD patient survival under the same conditions. ResultsOur results suggested that compared with SCN and PPI network, the FCN based SBMOI combined with RSF model had better performance in long- and short-term survival prediction tasks for LUAD patients. The AUC of 1-year, 5-year, and 10-year survival in the validation dataset were 0.791, 0.825, and 0.917, respectively. ConclusionsThis study provided a powerful network-based method to multi-omics data integration. SBMOI combined with RSF successfully predicted long- and short-term survival of LUAD patients, especially with high accuracy on long-term survival. Besides, SBMOI algorithm has the potential to combine with other machine learning models to complete clustering or stratificational tasks, and being applied to other diseases.

The prognostic value of sialylation-related long non-coding RNAs in lung adenocarcinoma

There has been increasing interest in the role of epigenetic modification in cancers recently. Among the various modifications, sialylation has emerged as a dominant subtype implicated in tumor progression, metastasis, immune evasion, and chemoresistance. The prognostic significance of sialylation-related molecules has been demonstrated in colorectal cancer. However, the potential roles and regulatory mechanisms of sialylation in lung adenocarcinoma (LUAD) have not been thoroughly investigated. Through Pearson correlation, univariate Cox hazards proportional regression, and random survival forest model analyses, we identified several prognostic long non-coding RNAs (lncRNAs) associated with aberrant sialylation and tumor progression, including LINC00857, LINC00968, LINC00663, and ITGA9-AS1. Based on the signatures of four lncRNAs, we classified patients into two clusters with different landscapes using a non-negative matrix factorization approach. Collectively, patients in Cluster 1 (C1) exhibited worse prognoses than those in Cluster 2 (C2), as well as heavier tumor mutation burden. Functional enrichment analysis showed the enrichment of several pro-tumor pathways in C1, differing from the upregulated Longevity and programmed cell death pathways in C2. Moreover, we profiled immune infiltration levels of important immune cell lineages in two subgroups using MCPcounter scores and single sample gene set enrichment analysis scores, revealing a relatively immunosuppressive microenvironment in C1. Risk analysis indicated that LINC00857 may serve as a pro-tumor regulator, while the other three lncRNAs may be protective contributors. Consistently, we observed upregulated LINC00857 in C1, whereas increased expressive levels of LINC00968, LINC00663, and ITGA9-AS1 were observed in C2. Finally, drug sensitivity analysis suggested that patients in the two groups may benefit from different therapeutic strategies, contributing to precise treatment in LUAD. By integrating multi-omics data, we identified four core sialylation-related lncRNAs and successfully established a prognostic model to distinguish patients with different characterizations. These findings may provide some insights into the underlying mechanism of sialylation, and offer a new stratification way as well as clinical guidance in LUAD.

Explainable and visualizable machine learning models to predict biochemical recurrence of prostate cancer.

Machine learning (ML) models presented an excellent performance in the prognosis prediction. However, the black box characteristic of ML models limited the clinical applications. Here, we aimed to establish explainable and visualizable ML models to predict biochemical recurrence (BCR) of prostate cancer (PCa). A total of 647 PCa patients were retrospectively evaluated. Clinical parameters were identified using LASSO regression. Then, cohort was split into training and validation datasets with a ratio of 0.75:0.25 andBCR-related features were included in Cox regression and five ML algorithm to construct BCR prediction models. The clinical utility of each model wasevaluated by concordance index (C-index) values and decision curve analyses (DCA). Besides, Shapley Additive Explanation (SHAP) values were used to explain the features in the models. We identified 11 BCR-related features using LASSO regression, then establishing five ML-based models, including random survival forest (RSF), survival support vector machine (SSVM), survival Tree (sTree), gradient boosting decision tree (GBDT), extreme gradient boosting (XGBoost), and a Cox regression model, C-index were 0.846 (95%CI 0.796-0.894), 0.774 (95%CI 0.712-0.834), 0.757 (95%CI 0.694-0.818), 0.820 (95%CI 0.765-0.869), 0.793 (95%CI 0.735-0.852), and 0.807 (95%CI 0.753-0.858), respectively. The DCA showed that RSF model had significant advantages over all models. In interpretability of ML models, the SHAP value demonstrated the tangible contribution of each feature in RSF model. Our score system provide reference for the identification for BCR, and the crafting of a framework for making therapeutic decisions for PCa on a personalized basis.

Abstract LB103: Cell-free DNA methylation as a noninvasive prognostic biomarker in hepatocellular carcinoma

Abstract The current noninvasive prognostic evaluation of hepatocellular carcinoma (HCC) using markers such as imaging features and serum biomarkers like alpha-fetoprotein (AFP) is limited due to low sensitivity. Pathologic tumor features can enhance prognosis beyond imaging and AFP, yet pre-surgical percutaneous biopsies are invasive, not routinely pursued, and may not be representative of the entire tumor. There is a critical unmet need to develop noninvasive blood-based HCC biomarkers to guide therapeutic decisions. Epigenetic changes in circulating cell-free DNA (cfDNA) have shown promise in early cancer diagnosis and prognosis. Therefore, we aim to evaluate the potential of cfDNA methylation as a noninvasive predictor for pre-treatment prognostication in HCC patients. Using Illumina HumanMethylation450 array data of 377 HCC tumors and 50 adjacent normal tissues obtained from The Cancer Genome Atlas (TCGA), we systematically identified 158 HCC-related DNA methylation markers associated with overall survival (OS). Functional analyses of these markers indicated their roles in regulating gene expression and revealed biological pathways associated with HCC patient survival. We predicted an OS risk score utilizing the DNA methylation panel, referred to as “methRisk”, and validated its predictive power using the Reduced Representation Bisulfite Sequencing data of HCC tumor tissue samples from 29 surgically resected patients. To assess the noninvasive prognostic value of these markers, we collected plasma cfDNA samples from an independent validation cohort of 52 HCC patients, and profiled the methylation patterns of these samples using our newly developed cost-effective cfDNA methylome sequencing assay, cfMethyl-Seq. We utilized random survival forest models combined with 10-fold cross-validation to predict the methRisk using the circulating markers. The cfDNA-based methRisk effectively stratifies patients into clinically relevant risk groups and predicts patient survival with high accuracy (3-year AUC = 0.81, 95% CI: 0.68-0.94). Integrating the cfDNA-based methRisk with existing risk indices like the Barcelona clinic liver cancer (BCLC) staging greatly improves the noninvasive prognostic assessments (3-year AUC = 0.91, 95% CI: 0.80-0.99). In summary, these findings indicate that cfDNA methylation is a promising noninvasive prognostic biomarker for HCC, providing a proof-of-concept for its potential clinical utility. Incorporating cfDNA methylation analysis into the existing noninvasive prognostic framework for HCC patients holds great potential for enhancing the effectiveness of current staging systems and optimizing patient selection for curative therapies like liver transplantation. Citation Format: Ran Hu, Benjamin Tran, Shuo Li, Mary L. Stackpole, Weihua Zeng, Yonggang Zhou, Andrew Melehy, Xianghong Zhou, Wenyuan Li, Vatche Agopian. Cell-free DNA methylation as a noninvasive prognostic biomarker in hepatocellular carcinoma [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2024; Part 2 (Late-Breaking, Clinical Trial, and Invited Abstracts); 2024 Apr 5-10; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2024;84(7_Suppl):Abstract nr LB103.

A prediction model based on random survival forest analysis of the overall survival of elderly female papillary thyroid carcinoma patients: a SEER-based study.

Papillary thyroid carcinoma (PTC) is a common malignancy whose incidence is three times greater in females than in males. The prognosis of ageing patients is poor. This research was designed to construct models to predict the overall survival of elderly female patients with PTC. We developed prediction models based on the random survival forest (RSF) algorithm and traditional Cox regression. The data of 4539 patients were extracted from the Surveillance, Epidemiology, and End Results (SEER) database. Twelve variables were analysed to establish the models. The C-index and the Brier score were selected to evaluate the discriminatory ability of the models. Time-dependent receiver operating characteristic (ROC) curves were also drawn to evaluate the accuracy of the models. The clinical benefits of the two models were compared on the basis of the DCA curve. In addition, the Shapley Additive Explanations (SHAP) plot was used to visualize the contribution of the variables in the RSF model. The C-index of the RSF model was 0.811, which was greater than that of the Cox model (0.781). According to the Brier score and the area under the ROC curve (AUC), the RSF model performed better than the Cox model. On the basis of the DCA curve, the RSF model demonstrated fair clinical benefit. The SHAP plot showed that age was the most important variable contributing to the outcome of PTC in elderly female patients. The RSF model we developed performed better than the Cox model and might be valuable for clinical practice.

Machine Learning Model Based on the Neutrophil-to-Eosinophil Ratio Predicts the Recurrence of Hepatocellular Carcinoma After Surgery.

Circulating eosinophils are associated with tumor development. An eosinophil-related index, the neutrophil to eosinophil ratio (NER), can be used to predict the prognosis of patients with tumors. However, there is still a lack of efficient prognostic biomarkers for HCC. In this study, we aimed to investigate the predictive value of the NER and develop an optimal machine learning model for the recurrence of HCC patients. Patients and methods: A retrospective collection of 562 patients who underwent hepatectomy with a pathologic diagnosis of HCC was performed. The relationship between NER and progression-free survival (PFS) was investigated. We developed a new machine learning framework with 10 machine learning algorithms and their 101 combinations to select the best model for predicting recurrence after hepatectomy. The performance of the model was assessed by the area under the curve (AUC) of characteristics and calibration curves, and clinical utility was evaluated by decision curve analysis (DCA). Kaplan‒Meier curves showed that the PFS in the low NER group was significantly better than that in the high NER group. Multivariate Cox regression analysis showed that NER was an independent risk factor for recurrence after surgery. The random survival forests (RSF) model was selected as the best model that had good predictive efficacy and outperformed the TNM, BCLC, and CNLC staging systems. The NER has good predictive value for postoperative recurrence in patients with hepatocellular carcinoma. Machine learning model based on NER can be used for accurate predictions.

Abstract 6123: Validating cancer modulated allostatic load as a composite biomarker for mortality in patients with cancer

Abstract Allostatic load (AL) is the cumulative burden of chronic stress and life events typically measured by lab and vital values routinely collected during standard care. AL has been associated with adverse socioenvironmental stressors and increased mortality rates including risk of cancer death. Despite modifications of AL as a result of progressing cancer and anti-cancer treatments, AL may serve as a valuable biomarker for cancer outcomes, and highlight similar disparities across sociodemographic groups agnostic of a history of cancer diagnosis. Recent investigations have shown associations of AL with overall mortality rates in patients with non-small cell lung cancer (NSCLC) and breast cancer, as well as various social determinants of health. The impact of AL on different malignancies, particularly in the context of genetic variations identified through genomic profiling, is yet to be thoroughly investigated. We identified a pan-cancer cohort of 29,433 patients sequenced with the MSK-IMPACT targeted panel sequencing assay, and collected clinical, genomic, and lab data. We first studied the association of AL with patient and tumor characteristics such as tumor type, stage, cancer genomics, comorbidities, social deprivation defined by the Yost index, and genetic ancestry. We derived AL using 10 standard biomarkers across cardiovascular, metabolic, renal, and immune systems, and created a composite AL summary score with a range between 0 and 10. Composite AL summary score into quartiles for comparisons. We found that AL was increased with higher cancer stage, older age, smoking, multiple comorbidities, African ancestry, and Yost index. Multivariate logistic regression analysis showed that AL was correlated with mutations e in KRAS and anti-correlated with EGFR mutations (FDR&amp;lt;0.02). Next, we investigated AL as a biomarker for mortality in patients with cancer and compared its prognostic value to other biomarkers. Using a Cox proportional hazard model we found that AL was significantly associated with shorter overall survival, controlling for stage and socioeconomic variables (HR=1.7, 2.2, 2.7 in AL Quartiles 2-4 respectively). Finally, we decomposed AL into its individual markers to quantify the most important features driving outcomes. Using a Random Survival Forest model with a focus on NSCLC patients, we found that albumin and alkaline phosphatase were the strongest features of shorter survival in our AL composite model. Our findings suggest that elevated AL is associated with a poorer prognosis across multiple cancer types, a trend observable across different stages of cancer. The association between AL and clinically actionable mutations highlights the importance of understanding the relationship between somatic alterations and social/environmental factors. The study underscores the utility of AL as a powerful prognostic tool that can be routinely collected in a clinical setting. Citation Format: Christopher J. Fong, Kaicheng U, Cheryl Phua, Xuechun Bai, Michele Waters, Tom Fu, Kanika Arora, Tomin Perea-Chamblee, Devika Jutagir, Michael Berger, Nikolaus Schultz, Adam Schoenfeld, Francesca Gany, Justin Jee, Jian Carrot-Zhang. Validating cancer modulated allostatic load as a composite biomarker for mortality in patients with cancer [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2024; Part 1 (Regular Abstracts); 2024 Apr 5-10; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2024;84(6_Suppl):Abstract nr 6123.

A prognostic model for Schistosoma japonicum infection-associated liver hepatocellular carcinoma: strengthening the connection through initial biological experiments

BackgroundNumerous studies have shown that Schistosoma japonicum infection correlates with an increased risk of liver hepatocellular carcinoma (LIHC). However, data regarding the role of this infection in LIHC oncogenesis are scarce. This study aimed to investigate the potential mechanisms of hepatocarcinogenesis associated with Schistosoma japonicum infection.MethodsBy examining chronic liver disease as a mediator, we identified the genes contributing to Schistosoma japonicum infection and LIHC. We selected 15 key differentially expressed genes (DEGs) using weighted gene co-expression network analysis (WGCNA) and random survival forest models. Consensus clustering revealed two subgroups with distinct prognoses. Least Absolute Shrinkage and Selection Operator (LASSO) and Cox regression identified six prognostic DEGs, forming an Schistosoma japonicum infection-associated signature for strong prognosis prediction. This signature, which is an independent LIHC risk factor, was significantly correlated with clinical variables. Four DEGs, including BMI1, were selected based on their protein expression levels in cancerous and normal tissues. We confirmed BMI1's role in LIHC using Schistosoma japonicum-infected mouse models and molecular experiments.ResultsWe identified a series of DEGs that mediate schistosomiasis, the parasitic disease caused by Schistosoma japonicum infection, and hepatocarcinogenesis, and constructed a suitable prognostic model. We analyzed the mechanisms by which these DEGs regulate disease and present the differences in prognosis between the different genotypes. Finally, we verified our findings using molecular biology experiments.ConclusionBioinformatics and molecular biology analyses confirmed a relationship between schistosomiasis and liver hepatocellular cancer. Furthermore, we validated the role of a potential oncoprotein factor that may be associated with infection and carcinogenesis. These findings enhance our understanding of Schistosoma japonicum infection's role in LIHC carcinogenesis.

Machine learning-based survival analysis approaches for predicting the risk of pneumonia post-stroke discharge

BackgroundPost-stroke pneumonia (PSP) is common among stroke patients. PSP occurring after hospital discharge continues to increase the risk of poor functional outcomes and death among stroke survivors. Currently, there is no prediction model specifically designed to predict the occurrence of PSP beyond the acute stage of stroke. This study aimed to explore the use of machine learning (ML) methods in predicting the risk of PSP after hospital discharge. MethodsThis study analyzed data from 5,754 hospitalized stroke patients. The dataset was randomly divided into a training set and a holdout test set, with a ratio of 80:20. Several clinical and laboratory variables were utilized as predictors and different ML algorithms were employed to model time-to-event data. The ML model's predictive performance was compared to existing risk-scoring systems. A model-agnostic method based on Shapley additive explanations was utilized to interpret the ML model. ResultsThe study found that 5.7% of the study patients experienced pneumonia within one year after discharge. Based on repeated 5-fold cross-validation on the training set, the random survival forest (RSF) model had the highest C-index among the various ML algorithms and traditional Cox regression analysis. The final RSF model achieved a C-index of 0.787 (95% confidence interval: 0.737–0.840) on the holdout test set, outperforming five existing risk-scoring systems. The top three important predictors were the Glasgow Coma Scale score, age, and length of hospital stay. ConclusionsThe RSF model demonstrated superior discriminative ability compared to other ML algorithms and traditional Cox regression analysis, suggesting a non-linear relationship between predictors and outcomes. The developed ML model can be integrated into the hospital information system to provide personalized risk assessments.

Development and validation of a machine learning model to predict prognosis in liver failure patients treated with non-bioartificial liver support system.

The prognosis of liver failure treated with non-bioartificial liver support systems is poor. Detecting its risk factors and developing relevant prognostic models still represent the top priority to lower its death risk. All 215 patients with liver failure treated with non-bioartificial liver support system were retrospectively analyzed. Potential prognostic factors were investigated, and the Nomogram and the Random Survival Forests (RSF) models were constructed, respectively. Notably, we evaluated the performance of models and calculated the risk scores to divide patients into low-risk and high-risk groups. In the training set, multifactorial Cox regression analysis showed that etiology, hepatic encephalopathy, total bilirubin, serum alkaline phosphatase, platelets, and MELD score were independent factors of short-term prognosis. The RSF model (AUC: 0.863, 0.792) performed better in prediction than the Nomogram model (AUC: 0.816, 0.756) and MELD (AUC: 0.658, 0.700) in the training and validation groups. On top of that, patients in the low-risk group had a significantly better prognosis than those in the high-risk group. We constructed the RSF model with etiology, hepatic encephalopathy, total bilirubin, serum alkaline phosphatase, platelets, and MELD score, which showed better prognostic power than the Nomogram model and MELD score and could help physicians make optimal treatment decisions.

Machine Learning Did Not Outperform Conventional Competing Risk Modeling to Predict Revision Arthroplasty.

Estimating the risk of revision after arthroplasty could inform patient and surgeon decision-making. However, there is a lack of well-performing prediction models assisting in this task, which may be due to current conventional modeling approaches such as traditional survivorship estimators (such as Kaplan-Meier) or competing risk estimators. Recent advances in machine learning survival analysis might improve decision support tools in this setting. Therefore, this study aimed to assess the performance of machine learning compared with that of conventional modeling to predict revision after arthroplasty. Does machine learning perform better than traditional regression models for estimating the risk of revision for patients undergoing hip or knee arthroplasty? Eleven datasets from published studies from the Dutch Arthroplasty Register reporting on factors associated with revision or survival after partial or total knee and hip arthroplasty between 2018 and 2022 were included in our study. The 11 datasets were observational registry studies, with a sample size ranging from 3038 to 218,214 procedures. We developed a set of time-to-event models for each dataset, leading to 11 comparisons. A set of predictors (factors associated with revision surgery) was identified based on the variables that were selected in the included studies. We assessed the predictive performance of two state-of-the-art statistical time-to-event models for 1-, 2-, and 3-year follow-up: a Fine and Gray model (which models the cumulative incidence of revision) and a cause-specific Cox model (which models the hazard of revision). These were compared with a machine-learning approach (a random survival forest model, which is a decision tree-based machine-learning algorithm for time-to-event analysis). Performance was assessed according to discriminative ability (time-dependent area under the receiver operating curve), calibration (slope and intercept), and overall prediction error (scaled Brier score). Discrimination, known as the area under the receiver operating characteristic curve, measures the model's ability to distinguish patients who achieved the outcomes from those who did not and ranges from 0.5 to 1.0, with 1.0 indicating the highest discrimination score and 0.50 the lowest. Calibration plots the predicted versus the observed probabilities; a perfect plot has an intercept of 0 and a slope of 1. The Brier score calculates a composite of discrimination and calibration, with 0 indicating perfect prediction and 1 the poorest. A scaled version of the Brier score, 1 - (model Brier score/null model Brier score), can be interpreted as the amount of overall prediction error. Using machine learning survivorship analysis, we found no differences between the competing risks estimator and traditional regression models for patients undergoing arthroplasty in terms of discriminative ability (patients who received a revision compared with those who did not). We found no consistent differences between the validated performance (time-dependent area under the receiver operating characteristic curve) of different modeling approaches because these values ranged between -0.04 and 0.03 across the 11 datasets (the time-dependent area under the receiver operating characteristic curve of the models across 11 datasets ranged between 0.52 to 0.68). In addition, the calibration metrics and scaled Brier scores produced comparable estimates, showing no advantage of machine learning over traditional regression models. Machine learning did not outperform traditional regression models. Neither machine learning modeling nor traditional regression methods were sufficiently accurate in order to offer prognostic information when predicting revision arthroplasty. The benefit of these modeling approaches may be limited in this context.

Altered Motor Activity Patterns within 10-Minute Timescale Predict Incident Clinical Alzheimer's Disease.

Fractal motor activity regulation (FMAR), characterized by self-similar temporal patterns in motor activity across timescales, is robust in healthy young humans but degrades with aging and in Alzheimer's disease (AD). To determine the timescales where alterations of FMAR can best predict the clinical onset of AD. FMAR was assessed from actigraphy at baseline in 1,077 participants who had annual follow-up clinical assessments for up to 15 years. Survival analysis combined with deep learning (DeepSurv) was used to examine how baseline FMAR at different timescales from 3 minutes up to 6 hours contributed differently to the risk for incident clinical AD. Clinical AD occurred in 270 participants during the follow-up. DeepSurv identified three potential regions of timescales in which FMAR alterations were significantly linked to the risk for clinical AD: <10, 20-40, and 100-200 minutes. Confirmed by the Cox and random survival forest models, the effect of FMAR alterations in the timescale of <10 minutes was the strongest, after adjusting for covariates. Subtle changes in motor activity fluctuations predicted the clinical onset of AD, with the strongest association observed in activity fluctuations at timescales <10 minutes. These findings suggest that short actigraphy recordings may be used to assess the risk of AD.