Survival Forest Models Research Articles

SPINK5 is a key regulator of eosinophil extracellular traps in head and neck squamous cell carcinoma

Enhanced infiltration of eosinophils is observed surrounding solid tumors. Some studies indicate that Eosinophil extracellular traps (EETs) play a crucial role in tumor progression and metastasis. However, its specific role in head and neck squamous cell carcinoma (HNSCC) remains unclear. This study established a gene set associated with eosinophil differentiation, chemotaxis, and EETs release from previous research. Employing bioinformatics techniques, the expression and biological significance of these genes in HNSCC were analyzed. Briefly, unsupervised clustering based on expression patterns of 133 EETs-related genes to classify TCGA-HNSCC patients. Immune cell infiltration patterns were assessed using “ImmuCellAI” package. A prognostic model was constructed using ten algorithms, with EETs-related gene sets as input features. Here, unsupervised clustering of samples into two types revealed worse prognosis for Cluster 1 (C1) patients after the first year. Cluster 2 (C2) exhibited higher ImmuneScore, but with a distinct immune cell infiltration pattern from the C1. Additionally, high eosinophil abundance only in the C2 had a positive prognostic impact. Serine peptidase inhibitor kazal type 5 (SPINK5) emerged as a potential key gene mediating the formation of EETs in HNSCC. EETs not only exhibit a positive correlation with diverse anti-cancer pathways but also demonstrate positive associations with processes such as proliferation, migration, and other critical pathways. The random survival forest (RSF) model was identified as the optimal eosinophil-related prognostic model. Collectively, this study elucidates the potential impact and mediating pathways of EETs on tumors, providing a reference for targeted therapy based on EETs-related genes.

Random survival forest algorithm for risk stratification and survival prediction in gastric neuroendocrine neoplasms.

This study aimed to construct and assess a machine-learning algorithm designed to forecast survival rates and risk stratification for patients with gastric neuroendocrine neoplasms (gNENs) after diagnosis. Data on patients with gNENs were extracted and randomly divided into training and validation sets using the Surveillance, Epidemiology, and End Results database. We developed a prediction model using 10 machine learning algorithms across 101 combinations to forecast cancer-related mortality in patients with gNENs, selecting the best model using the highest mean over a sequence of time-dependent area under the receiver operating characteristic (ROC) curve (AUC). The performance of the final model was assessed through time-dependent ROC curves for discrimination and calibration curves for calibration. The maximum selection rank method was used to determine the best prognostic risk score threshold for classifying patients into high- and low-risk groups. Afterward, Kaplan-Meier analysis and log-rank test were used to compare survival rates among these groups. Our study examined 775 patients with gNENs, dividing them into training and validation sets. A training set comprised 543 patients, with a median follow-up of 42 months and cumulative mortality rates of 40.0% at 1 year, 48.6% at 3 years, and 54.0% at 5 years. A validation set comprised 232 patients, with cumulative mortality rates of 29.1% at 1 year, 43.5% at 3 years, and 53.2% at 5 years. The optimal random survival forest (RSF) model (mtry = 4, node size = 5) achieved an AUC of 0.839 for survival prediction in the training set. Comprising 11 variables such as demographics, treatment details, tumor characteristics, T staging, N staging, and M staging, the RSF model revealed high predictive accuracy with AUCs of 0.92, 0.96, and 0.96 for 1-, 3-, and 5-year survival, respectively, which was consistently reflected in the validation set with AUCs of 0.88, 0.92, and 0.89, respectively. Moreover, patients were risk-stratified. Although our RSF model effectively stratified patients into different prognostic groups, it needs external validation to confirm its utility for noninvasive prognostic prediction and risk stratification in gNENs. Further research is required to verify its broader clinical applicability.

Predicting survival benefits of immune checkpoint inhibitor therapy in lung cancer patients: a machine learning approach using real-world data.

Due to the heterogeneity in the effectiveness of immunotherapy for lung cancer, identifying predictors is crucial. This study aimed to develop a machine learning model to identify predictors of overall survival in lung cancer patients treated with immune checkpoint inhibitors (ICIs). A retrospective analysis was performed on data from 1314 lung cancer patients at the Chongqing University Cancer Hospital from September 2018 to September 2022. We used the random survival forest (RSF) model to identify survival-influencing factors, using backward elimination for variable selection. A Cox proportional hazards (CPH) model was constructed using the most significant predictors. We assessed model performance and generalizability using time-dependent receiver operating characteristics (ROC) and predictive error curves. The RSF model demonstrated better predictive accuracy than the CPH (IBS 0.17 vs. 0.17; C-index 0.91 vs. 0.68), with better discrimination and prediction performance. The influential variables identified included D-dimer, Karnofsky performance status, albumin, surgery, TNM stage, platelet count, and age. The RSF model, which incorporated these variables, achieved area under the curve (AUC) scores of 0.95, 0.94, and 0.98 for 1-, 3-, and 5-year survival predictions, respectively, in the training set. The validation set showed AUCs of 0.94, 0.90, and 0.95, respectively, exceeding the performance of the CPH model. The study successfully developed a machine learning model that accurately predicted the survival benefits of ICI therapy in lung cancer patients, supporting clinical decision-making in lung cancer treatment.

A retrospective study on the prognosis of endoscopic surgery for 385 early glottic cancer patients

Objective: To investigate the prognosis and influencing factors of endoscopic surgery for early glottic carcinoma. Methods: In this retrospective study, we applied the Cox proportional hazards regression model and the random survival forest model to analyze the clinical characteristics of 385 patients [362 males, 23 females, age ranging from 33 to 91 years (62.0±9.6)] who visited the Sixth Medical Center of the General Hospital of the People's Liberation Army from January 2009, to December 2022 and diagnosed with early glottic carcinoma, encompassing variables such as age, gender, T stage, surgical approach, pathological typing, etc. The primary evaluation indicators were overall survival(OS) and disease-free survival rates (DFS). The follow-up duration ranged from 30 to 5,557 days (with a median follow-up time of 1,596 days). Results: After a three-year follow-up, the OS rate for the 385 patients was 95.83%, while the DSF rate was 82.98%. The Cox proportional hazards regression analysis revealed age (HR=2.35, 95%CI: 1.75 to 3.15, P<0.001) and T staging (HR=1.59, 95%CI: 1.13 to 2.23, P=0.019) as predominant factors affecting the OS and DFS. The random survival forest model identified poor tumor differentiation, and high expression of P53 and Ki-67 as predictors of inferior prognosis. Conclusion: Endoscopic surgery for early glottic carcinoma yields favorable short-term OS and reduces short-term recurrence rates, with T-stage emerging as a pivotal factor influencing recurrence. Tumors with poor differentiation and elevated expression of P53 may be indicative of an increased risk of recurrence.

Creating an interactive database for nasopharyngeal carcinoma management: applying machine learning to evaluate metastasis and survival.

Nasopharyngeal carcinoma (NPC) patients frequently present with distant metastasis (DM), which is typically associated with poor prognosis. This study aims to develop and apply machine learning models to predict DM, overall survival (OS), and cancer-specific survival (CSS) in NPC patients to provide optimal tools for improved predictive accuracy and performance. We retrieved over 8,000 NPC patient samples with associated clinical information from the Surveillance, Epidemiology, and End Results (SEER) database. Utilizing two methods for handling missing values-imputation or deletion-we created various cohorts: DM-all, DM-slim, OS-all, OS-slim, CSS-all, and CSS-slim. Five machine learning models were deployed for the binary classification task of DM, and their performance was evaluated using the area under the curve (AUC). For the survival prediction tasks of OS and CSS, we constructed 45 combinations using nine survival machine learning algorithms. The Concordance Index (C-index), 5-year AUC, and Brier score assessed model accuracy. Patients were stratified into two risk groups for survival analysis, and the survival curves were presented. This study examines the relationships between clinical factors and survival in NPC patients. The analysis, visualized through forest plots, indicates that demographic and clinical variables like gender, marital status, tumor grade, and stage significantly affect metastatic risks and survival. Specifically, factors such as advanced stages increase metastasis and survival risks, while enhanced treatments improve survival rates. In the cohort for DM prediction, results revealed that the random forest model was the most effective, with an AUC of 0.687. In contrast, when predicting overall survival (OS), the random survival forest (RSF) model consistently showed superior performance with the highest mean C-index of 0.802, a 5-year AUC of 0.857, and a Brier score of 0.167. Similarly, for cancer-specific survival (CSS) prediction, the RSF model demonstrated a mean C-index of 0.822, a 5-year AUC of 0.884, and a Brier score of 0.165. An online Shiny server was developed to allow the models to be used freely and efficiently via http://npcml.shinyapps.io/NPCpre. This study successfully established an online tool by machine learning models for NPC metastasis and survival prediction, providing valuable references for clinicians.

A random survival forest-based pathomics signature classifies immunotherapy prognosis and profiles TIME and genomics in ES-SCLC patients

BackgroundSmall cell lung cancer (SCLC) is a highly aggressive neuroendocrine tumor with high mortality, and only a limited subset of extensive-stage SCLC (ES-SCLC) patients demonstrate prolonged survival under chemoimmunotherapy, which warrants the exploration of reliable biomarkers. Herein, we built a machine learning-based model using pathomics features extracted from hematoxylin and eosin (H&E)-stained images to classify prognosis and explore its potential association with genomics and TIME.MethodsWe retrospectively recruited ES-SCLC patients receiving first-line chemoimmunotherapy at Nanjing Jinling Hospital between April 2020 and August 2023. Digital H&E-stained whole-slide images were acquired, and targeted next-generation sequencing, programmed death ligand-1 staining, and multiplex immunohistochemical staining for immune cells were performed on a subset of patients. A random survival forest (RSF) model encompassing clinical and pathomics features was established to predict overall survival. The function of putative genes was assessed via single-cell RNA sequencing.Results and conclusionDuring the median follow-up period of 12.12 months, 118 ES-SCLC patients receiving first-line immunotherapy were recruited. The RSF model utilizing three pathomics features and liver metastases, bone metastases, smoking status, and lactate dehydrogenase, could predict the survival of first-line chemoimmunotherapy in patients with ES-SCLC with favorable discrimination and calibration. Underlyingly, the higher RSF-Score potentially indicated more infiltration of CD8+ T cells in the stroma as well as a greater probability of MCL-1 amplification and EP300 mutation. At the single-cell level, MCL-1 was associated with TNFA-NFKB signaling and apoptosis-related processes. Hopefully, this noninvasive model could act as a biomarker for immunotherapy, potentially facilitating precision medicine in the management of ES-SCLC.

A Random Survival Forest Model for Predicting Residual and Recurrent High-Grade Cervical Intraepithelial Neoplasia in Premenopausal Women.

Loop electrosurgical excision procedure (LEEP) for high-grade cervical intraepithelial neoplasia (CIN) carries significant risks of recurrence and persistence. This study compares the efficacy of a random survival forest (RSF) model with that of a conventional Cox regression model for predicting residual and recurrent high-grade CIN in premenopausal women after LEEP. Data from 458 premenopausal women treated for CIN2/3 at our hospital between 2016 and 2020 were analyzed. The RSF model incorporated demographic, pathological, and treatment-related variables. Feature selection utilizing LASSO and three other algorithms was performed to enhance the RSF model, which was further compared to a Cox regression model. Model performance was assessed using area under the curve (AUC), out-of-bag (OOB) error rates, and SHAP values to interpret predictor importance. The RSF model showed superior performance compared to the Cox regression model, with AUC values of 0.767-0.901 and peak predictive performance at 36 months post-LEEP. In contrast, the highest AUC achieved by Cox regression was 0.880. The RSF model also exhibited relatively lower OOB error rates, indicating better generalizability. Moreover, SHAP value analysis identified margin status and CIN severity as the most prominent predictors that directly affected risk predictions. Lastly, an online tool providing real-time predictions in clinical settings was successfully implemented using the RSF model. The RSF model outperformed the traditional Cox regression model in predicting residual and recurrent high-grade CIN risks post-LEEP. This model may be a more accurate clinical tool that facilitates improved personalized care and early interventions in gynecological oncology.

Association between systemic inflammatory markers and all-cause mortality in patients with stroke: A prospective study using data from the UK Biobank

Background and aimsThe systemic immune-inflammation index (SII), neutrophil-to-lymphocyte ratio (NLR), platelet-to-lymphocyte ratio (PLR), and lymphocyte-to-monocyte ratio (LMR) are novel inflammatory biomarkers used to determine various disease prognoses. However, the effects of these systemic inflammatory markers on all-cause mortality in patients with stroke remain unclear. MethodsWe used data from the UK Biobank for this prospective analysis. Overall, 6,020 eligible individuals were included. Over a median follow-up of 13.4 years, 1,233 participants died. We examined the effects of systemic inflammatory markers on all-cause mortality using random survival forest (RSF) and Cox proportional hazards models. Covariate adjustments in the Cox model, selected by RSF, included age, sex, body mass index (BMI), Townsend deprivation index, smoking status, alcohol intake frequency, sleep duration, diabetes, and malignant neoplasms. ResultsIn the marginal effect plots and restricted cubic spline analysis for systemic inflammatory markers, LMR exhibited a linear negative correlation, NLR showed a linear positive correlation, and SII and PLR demonstrated a U-shaped association. After covariates were adjusted, the all-cause mortality risk increased by 14 % for LMR <4 (hazards ratio [HR]: 1.14, 95 % confidence interval [CI]: 1.01–1.29; p= 0.03), by 26 % for NLR ≥2 (HR: 1.26; 95 % CI: 1.11–1.43; p < 0.001),by 26 % for PLR ≥175 (HR: 1.26; 95 % CI: 1.07–1.47; p < 0.001), and by 31 % for SII ≥526 (HR: 1.31; 95 % CI, 1.16–1.47; p= 0.014). In addition, sensitivity analyses, excluding participants who had been followed-up for <2 years and those with malignant neoplasms, yielded results consistent with those of previous research. ConclusionSII, NLR, PLR, and LMR significantly correlate with all-cause mortality in stroke patients. Thresholds established by the RSF model could potentially refine prognostic decision-making in stroke care.

Genomic cluster formation among invasive group A streptococcal infections in the USA: a whole-genome sequencing and population-based surveillance study

Clusters of invasive group Astreptococcal (iGAS) infection, linked to genomically closely related group Astreptococcal (GAS) isolates (referred to as genomic clusters), pose public health threats, and are increasingly identified through whole-genome sequencing (WGS) analysis. In this study, we aimed to assess the risk of genomic cluster formation among iGAS cases not already part of existing genomic clusters. In this WGS and population-based surveillance study, we analysed iGAS case isolates from the Active Bacterial Core surveillance (ABCs), which is part of the US Centers for Disease Control and Prevention's Emerging Infections Program, in ten US states from Jan 1, 2015, to Dec 31, 2019. We included all residents in ABCs sites with iGAS infections meeting the case definition and excluded non-conforming GAS infections and cases with whole-genome assemblies of the isolate containing fewer than 1·5million total bases or more than 150contigs. For iGAS cases we collected basic demographics, underlying conditions, and risk factors for infection from medical records, and for isolates we included emm types, antimicrobial resistance, and presence of virulence-related genes. Two iGAS cases were defined as genomically clustered if their isolates differed by three or less single-nucleotide variants. An iGAS case not clustered with any previous cases at the time of detection, with a minimum trace-back time of 1year, was defined as being at risk of cluster formation. We monitored each iGAS case at risk for a minimum of 1year to identify any cluster formation event, defined as the detection of a subsequent iGAS case clustered with the case at risk. We used the Kaplan-Meier method to estimate the cumulative incidence of cluster formation events over time. We used Cox regression to assess associations between features of cases at risk upon detection and subsequent cluster formation. We developed a random survival forest machine-learning model based on a derivation cohort (random selection of 50% of cases at risk) to predict cluster formation risk. This model was validated using a validation cohort consisting of the remaining 50%of cases at risk. We identified 2764iGAS cases at risk from 2016to 2018, of which 656 (24%) formed genomic clusters by the end of 2019. Overall, the cumulative incidence of cluster formation was 0·057 (95% CI 0·048-0·066) at 30days after detection, 0·12 (0·11-0·13) at 90days after detection, and 0·16 (0·15-0·18) at 180days after detection. Ahigher risk of cluster formation was associated with emm type (adjusted hazard ratio as compared with emm89 was 2·37 [95% CI 1·71-3·30] for emm1, 2·72 [1·82-4·06] for emm3, 2·28 [1·49-3·51] for emm6, 1·47 [1·05-2·06] for emm12, and 2·21 [1·38-3·56] for emm92), homelessness (1·42 [1·01-1·99]), injection drug use (2·08 [1·59-2·72]), residence in a long-term care facility (1·78[1·29-2·45]), and the autumn-winter season (1·34 [1·14-1·57]) in multivariable Cox regression analysis. The machine-learning model stratified the validation cohort (n=1382) into groups at low (n=370), moderate (n=738), and high (n=274) risk. The 90-day risk of cluster formation was 0·03 (95% CI 0·01-0·05) for the group at low risk, 0·10(0·08-0·13) for the group at moderate risk, and 0·21 (0·17-0·25) for the group at high risk. These results were consistent with the cross-validation outcomes in the derivation cohort. Using population-based surveillance data, we found that pathogen, host, and environment factors of iGAS cases were associated with increased likelihood of subsequent genomic cluster formation. Groups at high risk were consistently identified by a predictive model which could inform prevention strategies, although future work to refine the model, incorporating other potential risk factors such as host contact patterns and immunity to GAS, is needed to improve its predictive performance. Centers for Disease Control and Prevention.

Exploring the prognostic value of T follicular helper cell levels in chronic lymphocytic leukemia

Chronic lymphocytic leukemia (CLL) presents with heterogeneous clinical outcomes, suggesting varied underlying pathogenic mechanisms. This study aims to elucidate the impact of T follicular helper (Tfh) cells on CLL progression and prognosis. Gene expression profile data for CLL were collected from GSE22762 and GSE39671 datasets. Patients were divided into high and low groups using Tfh levels using the optimal cutoff value based on overall survival (OS) and time-to-first treatment (TTFT). Differential expression analysis was performed between these groups, followed by co-expression network analysis and single-sample Gene Set Enrichment Analysis (ssGSEA). Marker genes of Tfh cells were used to construct prognostic models. Additionally, 40 CLL patients were recruited and categorized based on median Tfh levels. Marker gene expression was assessed using RT-qPCR and Western Blot, and immune cell levels were determined through flow cytometry. The high group showed better prognosis compared to the low group. Among the 1121 differentially expressed genes identified, five co-expression networks were constructed, with the turquoise module showing the highest correlation with Tfh cells. Genes within this module significantly participate in cytokine-cytokine receptor interaction, PI3K-Akt signaling pathway, and natural killer cell mediated cytotoxicity. Tfh cells were significantly negatively correlated with activated B cells and positively correlated with Tregs. The Random Survival Forest (RSF) model identified 10 marker genes, and further analysis using Lasso regression and nomogram selected CLEC4A, RAE1, CD84, and PRDX1 as prognostic markers. In the high group, levels of CLEC4A and RAE1 were higher than in the low group, whereas CD84 and PRDX1 were lower. Flow cytometry revealed that the level of activated B cells in the high Tfh group was significantly lower than in the low Tfh group, while the level of Tregs is significantly higher in the high Tfh group. This study seeks to contribute to a more detailed understanding of the pathogenesis of CLL, delving into the prognostic significance of Tfh.

Developing machine learning models for personalized treatment strategies in early breast cancer patients undergoing neoadjuvant systemic therapy based on SEER database

This study aimed to compare the long-term outcomes of breast-conserving surgery plus radiotherapy (BCS + RT) and mastectomy in early breast cancer (EBC) patients who received neoadjuvant systemic therapy (NST), and sought to construct and authenticate a machine learning algorithm that could assist healthcare professionals in formulating personalized treatment strategies for this patient population. We analyzed data from the Surveillance, Epidemiology, and End Results database on EBC patients undergoing BCS + RT or mastectomy post-NST (2010–2018). Employing propensity score matching (PSM) to minimize potential biases, we compared breast cancer-specific survival (BCSS) and overall survival (OS) between the two surgical groups. Additionally, we trained and validated six machine learning survival models and developed a cloud-based recommendation system for surgical treatment based on the optimal model. Among the 13,958 patients, 9028 (64.7%) underwent BCS + RT and 4930 (35.3%) underwent mastectomy. After PSM, there were 3715 patients in each group. Compared to mastectomy, BCS + RT significantly improved BCSS (p < 0.001) and OS (p < 0.001). Prognostic variables associated with BCSS were utilized to develop machine learning models. In both the training and validation cohorts, the random survival forest (RSF) model demonstrated superior predictive performance (0.847 and 0.795), not only outperforming other machine learning models, including Rpart (0.725 and 0.707), Xgboost (0.762 and 0.727), Glmboost (0.748 and 0.788), Survctree (0.764 and 0.766), and Survsvm (0.777 and 0.790), but also outperforming the classical COX model (0.749 and 0.782). Lastly, a web-based prediction tool was built to facilitate clinical application [https://jhren.shinyapps.io/shinyapp1]. After adjusting other confounders, BCS + RT was associated with improved outcomes in patients with EBC after NST, compared to those who underwent mastectomy. Moreover, the RSF model, a reliable tool, can predict long-term outcomes for patients, providing valuable guidance for operative methods and postoperative follow-up.

Associations between inflammatory burden index, prostate cancer, and mortality among middle-aged and elderly individuals.

Inflammation plays a crucial role in prostate cancer (PCa) progression and mortality. This study aimed to investigate the predictive value of the inflammatory burden index (IBI) and its components for mortality risk among men aged 40 years and older. A total of 7,344 participants from the NHANES 2001-2010 were included. High PCa risk was defined as a %fPSA greater than 25% and a tPSA level less than 4.0 ng/mL. Cox regression and logistic regression analyses were conducted to assess the association between IBI, PCa risk, and mortality. Receiver operating characteristic (ROC) curve analysis and random survival forest (RSF) model were utilized to evaluate the predictive value of IBI and its components for mortality. Elevated IBI levels were significantly associated with an increased risk of all-cause mortality (HR = 1.08 [1.05-1.10]) and cancer mortality (HR = 1.11 [1.07-1.15]). High-risk PCa cases also exhibited elevated mortality risk (all-cause: HR = 1.35 [1.19-1.54]; cancer: HR = 1.65 [1.27-2.14]). Additionally, the combined effect of elevated IBI levels and high PCa risk showed a synergistic impact on mortality outcomes (all-cause: HR = 1.49 [1.27-1.74]; cancer: HR = 1.76 [1.29-2.40]). ROC curve analysis revealed that IBI had the highest AUC for predicting all-cause mortality (AUC = 0.690 at 3 years, 0.622 at 5 years, 0.634 at 10 years, and 0.632 at 15 years) compared to its individual components (CRP, NEU, LYM). RSF analysis highlighted IBI as the most significant predictor of all-cause and cancer mortality. The combined effect of elevated IBI levels and high PCa risk demonstrated a synergistic impact on increased mortality risk among men aged 40 years and older. IBI demonstrated superior predictive performance for mortality outcomes compared to individual inflammatory markers. These findings underscore the potential utility of IBI as a prognostic biomarker for mortality risk assessment in individual with high PCa risk.

A Machine Learning Approach for Predicting Biochemical Outcome After PSMA-PET-Guided Salvage Radiotherapy in Recurrent Prostate Cancer After Radical Prostatectomy: Retrospective Study.

Salvage radiation therapy (sRT) is often the sole curative option in patients with biochemical recurrence after radical prostatectomy. After sRT, we developed and validated a nomogram to predict freedom from biochemical failure. This study aims to evaluate prostate-specific membrane antigen-positron emission tomography (PSMA-PET)-based sRT efficacy for postprostatectomy prostate-specific antigen (PSA) persistence or recurrence. Objectives include developing a random survival forest (RSF) model for predicting biochemical failure, comparing it with a Cox model, and assessing predictive accuracy over time. Multinational cohort data will validate the model's performance, aiming to improve clinical management of recurrent prostate cancer. This multicenter retrospective study collected data from 13 medical facilities across 5 countries: Germany, Cyprus, Australia, Italy, and Switzerland. A total of 1029 patients who underwent sRT following PSMA-PET-based assessment for PSA persistence or recurrence were included. Patients were treated between July 2013 and June 2020, with clinical decisions guided by PSMA-PET results and contemporary standards. The primary end point was freedom from biochemical failure, defined as 2 consecutive PSA rises >0.2 ng/mL after treatment. Data were divided into training (708 patients), testing (271 patients), and external validation (50 patients) sets for machine learning algorithm development and validation. RSF models were used, with 1000 trees per model, optimizing predictive performance using the Harrell concordance index and Brier score. Statistical analysis used R Statistical Software (R Foundation for Statistical Computing), and ethical approval was obtained from participating institutions. Baseline characteristics of 1029 patients undergoing sRT PSMA-PET-based assessment were analyzed. The median age at sRT was 70 (IQR 64-74) years. PSMA-PET scans revealed local recurrences in 43.9% (430/979) and nodal recurrences in 27.2% (266/979) of patients. Treatment included dose-escalated sRT to pelvic lymphatics in 35.6% (349/979) of cases. The external outlier validation set showed distinct features, including higher rates of positive lymph nodes (47/50, 94% vs 266/979, 27.2% in the learning cohort) and lower delivered sRT doses (<66 Gy in 57/979, 5.8% vs 46/50, 92% of patients; P<.001). The RSF model, validated internally and externally, demonstrated robust predictive performance (Harrell C-index range: 0.54-0.91) across training and validation datasets, outperforming a previously published nomogram. The developed RSF model demonstrates enhanced predictive accuracy, potentially improving patient outcomes and assisting clinicians in making treatment decisions.

Nuclear magnetic resonance-based metabolomics with machine learning for predicting progression from prediabetes to diabetes.

Identification of individuals with prediabetes who are at high risk of developing diabetes allows for precise interventions. We aimed to determine the role of nuclear magnetic resonance (NMR)-based metabolomic signature in predicting the progression from prediabetes to diabetes. This prospective study included 13,489 participants with prediabetes who had metabolomic data from the UK Biobank. Circulating metabolites were quantified via NMR spectroscopy. Cox proportional hazard (CPH) models were performed to estimate the associations between metabolites and diabetes risk. Supporting vector machine, random forest, and extreme gradient boosting were used to select the optimal metabolite panel for prediction. CPH and random survival forest (RSF) models were utilized to validate the predictive ability of the metabolites. During a median follow-up of 13.6 years, 2525 participants developed diabetes. After adjusting for covariates, 94 of 168 metabolites were associated with risk of progression to diabetes. A panel of nine metabolites, selected by all three machine-learning algorithms, was found to significantly improve diabetes risk prediction beyond conventional risk factors in the CPH model (area under the receiver-operating characteristic curve, 1 year: 0.823 for risk factors + metabolites vs 0.759 for risk factors, 5 years: 0.830 vs 0.798, 10 years: 0.801 vs 0.776, all p < 0.05). Similar results were observed from the RSF model. Categorization of participants according to the predicted value thresholds revealed distinct cumulative risk of diabetes. Our study lends support for use of the metabolite markers to help determine individuals with prediabetes who are at high risk of progressing to diabetes and inform targeted and efficient interventions. Shanghai Municipal Health Commission (2022XD017). Innovative Research Team of High-level Local Universities in Shanghai (SHSMU-ZDCX20212501). Shanghai Municipal Human Resources and Social Security Bureau (2020074). Clinical Research Plan of Shanghai Hospital Development Center (SHDC2020CR4006). Science and Technology Commission of Shanghai Municipality (22015810500).

Construction of an interpretable model for predicting survival outcomes in patients with middle to advanced hepatocellular carcinoma (≥5cm) using lasso-cox regression.

In this retrospective study, we aimed to identify key risk factors and establish an interpretable model for HCC with a diameter ≥ 5cm using Lasso regression for effective risk stratification and clinical decision-making. In this study, 843 patients with advanced hepatocellular carcinoma (HCC) and tumor diameter ≥ 5cm were included. Using Lasso regression to screen multiple characteristic variables, cox proportional hazard regression and random survival forest models (RSF) were established. By comparing the area under the curve (AUC), the optimal model was selected. The model was visualized, and the order of interpretable importance was determined. Finally, risk stratification was established to identify patients at high risk. Lasso regression identified 8 factors as characteristic risk factors. Subsequent analysis revealed that the lasso-cox model had AUC values of 0.773, 0.758, and 0.799, while the lasso-RSF model had AUC values of 0.734, 0.695, and 0.741, respectively. Based on these results, the lasso-cox model was chosen as the superior model. Interpretability assessments using SHAP values indicated that the most significant characteristic risk factors, in descending order of importance, were tumor number, BCLC stage, alkaline phosphatase (ALP), ascites, albumin (ALB), and aspartate aminotransferase (AST). Additionally, through risk score stratification and subgroup analysis, it was observed that the median OS of the low-risk group was significantly better than that of the middle- and high-risk groups. We have developed an interpretable predictive model for middle and late HCC with tumor diameter ≥ 5cm using lasso-cox regression analysis. This model demonstrates excellent prediction performance and can be utilized for risk stratification.

A large population-based and validated study on the follow-up management and supportive strategy of locally advanced rectal cancer patients.

Our objective was to evaluate the predictive factors and metastatic time for liver and lung metastasis in locally advanced rectal cancer (RC) patients. Univariate and multivariate analysis were performed to identify risk factors and prognostic factors for liver metastasis and lung metastasis in RC. Survival probabilities were calculated using the Kaplan-Meier model and compared using the log-rank test between groups. The probability of time-to-event occurrence was calculated using the random survival forest model. Finally, the SEER database was used to verify our findings. Our results indicated that pathological T stage and pathological N stage were independent predictive factors for liver metastasis. Furthermore, CEA level, pathological T stage, and tumor deposit were independent predictive factors for lung metastasis. Based on the results of a multivariate Cox analysis, we categorized patients with liver and lung metastasis into three groups based on their scores. The results revealed that patients with higher scores had a higher probability of experiencing metastasis. For liver metastasis, Groups 1, 2, and 3 all exhibited higher occurrence rates within the first 24months. However, for lung metastasis, Group 4 showed the highest occurrence rate at the 12th month, while Groups 5 and 6 exhibited the highest occurrence rates at the 15th month. In summary, we developed predictive models to determine the likelihood of liver and lung metastasis in RC patients. It is crucial to implement a more intensive surveillance program for patients with unfavorable risk profiles in order to facilitate early detection of metastasis.

Analysis of Highway Vehicle Lane Change Duration Based on Survival Model

To investigate highway vehicle lane-changing behavior, we utilized the publicly available naturalistic driving dataset, HighD, to extract the movement data of vehicles involved in lane changes and their proximate counterparts. We employed univariate and multivariate Cox proportional hazards models alongside random survival forest models to analyze the influence of various factors on lane change duration, assess their statistical significance, and compare the performance of multiple random survival forest models. Our findings indicate that several variables significantly impact lane change duration, including the standard deviation of lane-changing vehicles, lane-changing vehicle speed, distance to the following vehicle in the target lane, lane-changing vehicle length, and distance to the following vehicle in the current lane. Notably, the standard deviation and vehicle length act as protective factors, with increases in these variables correlating with longer lane change durations. Conversely, higher lane-changing vehicle speeds and shorter distances to following vehicles in both the current and target lanes are associated with shorter lane change durations, indicating their role as risk factors. Feature variable selection did not substantially improve the training performance of the random survival forest model based on our findings. However, validation set evaluation showed that careful feature variable selection can enhance model accuracy, leading to improved AUC values. These insights lay the groundwork for advancing research in predicting lane-changing behaviors, understanding lane-changing intentions, and developing pre-emptive safety measures against hazardous lane changes.

Elevated inflammatory burden index increases mortality in adults with chronic inflammatory airway diseases: a nationwide cohort study

ObjectiveThe objective of this study was to investigate the potential association between the inflammatory burden index (IBI) and the prevalence of chronic inflammatory airway diseases (CIAD), as well as mortality rates among individuals diagnosed with CIAD.MethodsParticipants were sourced from the National Health and Nutrition Examination Survey (NHANES) conducted between 1999 and 2010. The IBI was calculated using the formula: IBI = C-reactive protein * neutrophils / lymphocytes. CIAD comprised self-reported asthma, chronic bronchitis, and chronic obstructive pulmonary disease (COPD). Mortality outcomes, including all-cause and respiratory disease mortality, were determined through linked data from the National Death Index (NDI) up to December 2019.ResultsA total of 27,495 adults were included. IBI was divided into quartiles, with the lowest quartile as the reference group. After adjusting for confounding variables, a positive correlation was observed between higher IBI and increased prevalence of total CIAD (OR = 1.383 [1.215–1.575]), asthma (OR = 1.267 [1.096–1.465]), chronic bronchitis (OR = 1.568 [1.263–1.946]), and COPD (OR = 1.907 [1.311–2.774]). Over a median follow-up of 12.33 [9.92-16.00] years, there were 1221 deaths from all causes and 220 deaths from respiratory disease among 4499 patients with CIAD. Following multivariate adjustments, the fourth quartile was significantly associated with increased risk of all-cause mortality (HR = 2.227 [1.714–2.893]) and respiratory disease mortality (HR = 2.748 [1.383–5.459]) compared to the first quartile of IBI in CIAD participants. Moreover, variable importance analysis using a random survival forest model demonstrated the significance of IBI in predicting mortality from both all-cause and respiratory diseases.ConclusionIBI exhibited an association with the prevalence of CIAD, with higher IBI levels correlating with elevated all-cause and respiratory disease mortality among individuals with CIAD.

Comparison of Extremely Randomized Survival Trees and Random Survival Forests: A Simulation Study

Abstract. Purpose: This simulation study investigates the Extremely Randomized Survival Trees (EST) model, a machine learning technique expected to handle survival analysis, particularly in large survival datasets, effectively. The study compares the performance of the EST model with that of the Random Survival Forest (RSF) model, focusing on the C-index value to determine which model performs better. Methods: The analysis begins with the generation of 540 simulated datasets, created by combining three levels of sample sizes, two levels of censoring proportions, three types of hazard functions, and 30 repetitions for each scenario. The simulation data were split into 80% training and 20% testing data. The training data were used to build the EST and RSF models, while the test data were used to evaluate their performance. The model with the highest C-index value was deemed the best performer, as a higher C-index indicates superior model performance. Result: The results indicate that the sample size, type of hazard function, and the method used influence that model performance. The EST model significantly outperformed the RSF model when the sample size was large, though no significant difference was observed when the sample size was small or medium. Additionally, the EST model consistently demonstrated faster computation times across all simulation scenarios. Novelty: This study provides a pioneering exploration into applying decision tree algorithms, specifically EST and RSF, in survival analysis. While these methods have been extensively studied in regression and classification contexts, their application in survival analysis remains relatively unexplored.

Development and internal validation of individualized prediction models of overall survival and 6-month mortality among patients with synchronous early-onset colorectal liver metastases

BackgroundEarly-onset colorectal cancer with synchronous liver metastasis (EO-CRLM) is a growing concern with a grim prognosis. MethodsEO-CRLM patients were identified from the National Cancer Database. Random survival forest model and random forest (RF) model were developed for the prediction of overall survival (OS) and 6-month mortality, respectively. ResultsThe variables with top contributions for random survival forest model of OS included primary tumor resection, chemotherapy and bone metastases. The AUCs of 1-, 3- and 5-year OS were 0.787, 0.763 and 0.761, respectively. The individualized risk profile predicted by the models closely aligned with the actual survival outcomes observed for the patients. The variables with top contributions for RF model for 6-month mortality included chemotherapy, Charlson-Deyo comorbidity score and presence of tumor deposits. RF model for 6-month mortality resulted in an AUC of 0.821 in training set, 0.828 in cross-validation and 0.852 in testing cohort. RF models for OS and 6-month mortality exhibited great net benefit with favorable clinical utility. ConclusionThe models generated in this study accurately identified EO-CRLM patients at risk of worse OS and short-term mortality, which may complement standard clinical assessment and aid in creation of advanced care planning.