Modeling time-to-event (survival) data using classification tree analysis

Development and validation of a hypoxia- and mitochondrial dysfunction- related prognostic model based on integrated single-cell and bulk RNA sequencing analyses in gastric cancer

Introduction

Gastric cancer (GC) remains a major global health threat ranking as the fifth most prevalent cancer. Hypoxia, a characteristic feature of solid tumors, significantly contributes to the malignant progression of GC. Mitochondria are the major target of hypoxic injury that promotes mitochondrial dysfunction during the development of cancers including GC. However, the gene signature and prognostic model based on hypoxia- and mitochondrial dysfunction-related genes (HMDRGs) in the prediction of GC prognosis have not yet been established.

Methods

The gene expression profile datasets of stomach cancer patients were retrieved from The Cancer Genome Atlas and the Gene Expression Omnibus databases. Prognostic genes were selected using Least Absolute Shrinkage and Selection Operator Cox (LASSO-Cox) regression analysis to construct a prognostic model. Immune infiltration was evaluated through ESTIMATE, CIBERSORT, and ssGSEA analyses. Tumor immune dysfunction and exclusion (TIDE) and immunophenoscore (IPS) were utilized to explore implications for immunotherapy. Furthermore, in vitro experiments were conducted to validate the functional roles of HMDRGs in GC cell malignancy.

Results

In this study, five HMDRGs (ZFP36, SERPINE1, DUSP1, CAV1, and AKAP12) were identified for developing a prognostic model in GC. This model stratifies GC patients into high- and low-risk groups based on median risk scores. A nomogram predicting overall survival (OS) was constructed and showed consistent results with observed OS. Immune infiltration analysis indicated that individuals in the high-risk group tend to exhibit increased immune cell infiltration. Additionally, analysis of cancer immunotherapy responses revealed that high-risk group patients exhibit poorer responses to cancer immunotherapy compared to the low-risk group. Immunohistochemistry (IHC) staining indicated that the expression levels of HMDRGs were remarkably correlated with GC, of which, SERPINE1 displayed the most pronounced up-regulation, while ZFP36 exhibited the most notable down-regulation in GC patients. Furthermore, in vitro investigation validated that SERPINE1 and ZFP36 contribute to the malignant processes of GC cells correlated with mitochondrial dysfunction.

Conclusions

This study presents a novel and efficient approach to evaluate GC prognosis and immunotherapy efficacy, and also provides insights into understanding the pathogenesis of GC.

Individual and contextual factors associated with the survival of patients with severe acute respiratory syndrome by COVID-19 in Brazil

OBJECTIVE

To analyze the influence of individual and contextual factors of the hospital and the municipality of care on the survival of patients with Severe Acute Respiratory Syndrome due to COVID-19.

METHODS

Hospital cohort study with data from 159,948 adults and elderly with Severe Acute Respiratory Syndrome due to COVID-19 hospitalized from January 1 to December 31, 2022 and reported in the Influenza Epidemiological Surveillance Information System. The contextual variables were related to the structure, professionals and equipment of the hospital establishments and socioeconomic and health indicators of the municipalities. The outcome was hospital survival up to 90 days. Survival tree and Kaplan-Meier curves were used for survival analysis.

RESULTS

Hospital lethality was 30.4%. Elderly patients who underwent invasive mechanical ventilation and were hospitalized in cities with low tax collection rates had lower survival rates compared to other groups identified in the survival tree (p<0.001).

CONCLUSION

The study indicated the interaction of contextual factors with the individual ones, and it shows that hospital and municipal characteristics increase the risk of death, highlighting the attention to the organization, operation, and performance of the hospital network.

A Novel Decision Tree Model for Predicting the Cancer-Specific Survival of Patients with Bladder Cancer Treated with Radical Cystectomy

Background/Objectives: The aim was to develop a decision tree and a new prognostic tool to predict cancer-specific survival in patients with urothelial bladder cancer treated with radical cystectomy. Methods: A total of 11,834 patients with bladder cancer treated with radical cystectomy between 2004 and 2019 from the SEER database were randomly split into the derivation (n = 7889) and validation cohorts (n = 3945). Survival curves were estimated using conditional decision tree analysis. We used Multiple Imputation by Chained Equations for the treatment of missing values and the pec package to compare the predictive performance. We extracted data from our model following CHARMS and assessed the risk of bias and applicability with PROBAST. Results: A total of 4824 (41%) patients died during the follow-up period due to bladder cancer. A decision tree was made and 12 groups were obtained. Patients with a higher AJCC stage and older age have a worse prognosis. The risk groups were summarized into high, intermediate and low risk. The integrated Brier scores between 0 and 191 months for the bootstrap estimates of the prediction error are the lowest for our conditional survival tree (0.189). The model showed a low risk of bias and low concern about applicability. The results must be externally validated. Conclusions: Decision tree analysis is a useful tool with significant discrimination. With this tool, we were able to stratify patients into 12 subgroups and 3 risk groups with a low risk of bias and low concern about applicability.

Development of a prognostic signature for immune-associated genes in bladder cancer and exploring potential drug findings

BACKGROUND

Bladder cancer, predominantly affecting men, is a prevalent malignancy of the urinary system. Although platinum-based chemotherapy has demonstrated certain enhancements in overall survival when compared to surgery alone, the efficacy of treatments is impeded by the unfavorable side effects of conventional chemotherapy medications. Nonetheless, immunotherapy exhibits potential in the treatment of bladder cancer.

METHODS

To create an immune-associated prognostic signature for bladder cancer, bioinformatics analyses were performed utilizing The Cancer Genome Atlas (TCGA) database in this study. By identifying differential gene expressions between the high-risk and low-risk groups, a potential therapeutic drug was predicted using the Connectivity Map database. Subsequently, the impact of this drug on the growth of T24 cells was validated through MTT assay and 3D cell culture techniques.

RESULTS

The signature included 1 immune-associated LncRNA (NR2F1-AS1) and 16 immune-associated mRNAs (DEFB133, RBP7, PDGFRA, CGB3, PDGFD, SCG2, ADCYAP1R1, OPRL1, PGR, PSMD1, TANK, PRDX1, ADIPOR2, S100A8, AHNAK, EGFR). Based on the assessment of risk scores, the patients were classified into cohorts of low-risk and high-risk individuals. The cohort with low risk demonstrated a considerably higher likelihood of survival in comparison to the group with high risk. Furthermore, variations in immune infiltration were noted among the two categories. Cephaeline, a possible medication, was discovered by analyzing variations in gene expression. It exhibited promise in suppressing the viability and growth of T24 bladder cancer cells.

CONCLUSION

The novel predictive pattern allows for efficient categorization of patients with bladder cancer, enabling focused and rigorous treatment for those expected to have a worse prognosis. The discovery of a possible curative medication establishes a basis for forthcoming immunotherapy trials in bladder cancer.

Late Mortality After Myocardial Injury in Critical Care Non-Cardiac Surgery Patients Using Machine Learning Analysis

Myocardial injury after noncardiac surgery (MINS) increases mortality within 30 days. We aimed to evaluate the long-term impact of myocardial injury in a large cohort of patients admitted to intensive care after noncardiac surgery. All patients who stayed, at least, overnight with measurement of high-sensitive cardiac troponin were included. Clinical characteristics and occurrence of MINS were assessed between patients who died and survivors using chi-square test and Student t test. Variables with p <0.01 in the univariate model were included in the Cox regression model to identify predictor variables. Survival decision tree (SDT), a machine learning model, was also used to find the predictors and their correlations. We included 2,230 patients with mean age of 63.8±16.3 years, with most (55.6%) being women. The prevalence of MINS was 9.4% (209 patients) and there were 556 deaths (24.9%) in a median follow-up of 6.7 years. Univariate analysis showed variables associated with late mortality, namely: MINS, arterial hypertension, previous myocardial infarction, atrial fibrillation, dementia, urgent surgery, peripheral artery disease (PAD), chronic health status, and age. These variables were included in the Cox regression model and SDT. The predictor variables of all-cause death were MINS (hazard ratio [HR] 2.21; 95% confidence interval [CI] 1.77 to 2.76), previous myocardial infarction (HR 1.47; 95% CI 1.14 to 1.89); urgent surgery (HR 1.24; 95% CI 1.01 to 1.52), PAD (HR 1.83; 95% CI 1.23 to 2.73), dementia (HR 2.54; 95% CI 1.86 to 3.46) and age (HR 1.05; 95% CI 1.04 to 1.06). SDT had the same predictors, except PAD. In conclusion, increased high-sensitive troponin levels in patients who underwent noncardiac surgery raised the risk of short and late mortality.

Development and validation of questionnaire-based machine learning models for predicting all-cause mortality in a representative population of China

Background

Considering that the previously developed mortality prediction models have limited applications to the Chinese population, a questionnaire-based prediction model is of great importance for its accuracy and convenience in clinical practice.

Methods

Two national cohort, namely, the China Health and Nutrition Survey (8,355 individual older than 18) and the China Health and Retirement Longitudinal Study (12,711 individuals older than 45) were used for model development and validation. One hundred and fifty-nine variables were compiled to generate predictions. The Cox regression model and six machine learning (ML) models were used to predict all-cause mortality. Finally, a simple questionnaire-based ML prediction model was developed using the best algorithm and validated.

Results

In the internal validation set, all the ML models performed better than the traditional Cox model in predicting 6-year mortality and the random survival forest (RSF) model performed best. The questionnaire-based ML model, which only included 20 variables, achieved a C-index of 0.86 (95%CI: 0.80-0.92). On external validation, the simple questionnaire-based model achieved a C-index of 0.82 (95%CI: 0.77-0.87), 0.77 (95%CI: 0.75-0.79), and 0.79 (95%CI: 0.77-0.81), respectively, in predicting 2-, 9-, and 11-year mortality.

Conclusions

In this prospective population-based study, a model based on the RSF analysis performed best among all models. Furthermore, there was no significant difference between the prediction performance of the questionnaire-based ML model, which only included 20 variables, and that of the model with all variables (including laboratory variables). The simple questionnaire-based ML prediction model, which needs to be further explored, is of great importance for its accuracy and suitability to the Chinese general population.

Machine learning in sudden cardiac death risk prediction: a systematic review

AIMS

Most patients who receive implantable cardioverter defibrillators (ICDs) for primary prevention do not receive therapy during the lifespan of the ICD, whilst up to 50% of sudden cardiac death (SCD) occur in individuals who are considered low risk by conventional criteria. Machine learning offers a novel approach to risk stratification for ICD assignment.

METHODS AND RESULTS

Systematic search was performed in MEDLINE, Embase, Emcare, CINAHL, Cochrane Library, OpenGrey, MedrXiv, arXiv, Scopus, and Web of Science. Studies modelling SCD risk prediction within days to years using machine learning were eligible for inclusion. Transparency and quality of reporting (TRIPOD) and risk of bias (PROBAST) were assessed. A total of 4356 studies were screened with 11 meeting the inclusion criteria with heterogeneous populations, methods, and outcome measures preventing meta-analysis. The study size ranged from 122 to 124 097 participants. Input data sources included demographic, clinical, electrocardiogram, electrophysiological, imaging, and genetic data ranging from 4 to 72 variables per model. The most common outcome metric reported was the area under the receiver operator characteristic (n = 7) ranging between 0.71 and 0.96. In six studies comparing machine learning models and regression, machine learning improved performance in five. No studies adhered to a reporting standard. Five of the papers were at high risk of bias.

CONCLUSION

Machine learning for SCD prediction has been under-applied and incorrectly implemented but is ripe for future investigation. It may have some incremental utility in predicting SCD over traditional models. The development of reporting standards for machine learning is required to improve the quality of evidence reporting in the field.

Cox model and decision trees: an application to breast cancer data

Objective.

To evaluate, using semiparametric methodologies of survival analysis, the relationship between covariates and time to death of patients with breast cancer, as well as the determination discriminatory power in the conditional inference tree of patients who had cancer.

Methods.

A retrospective cohort study was conducted using data collected from medical records of women who had breast cancer and underwent treatment between 2005 and 2015 at the Hospital da Fundação de Assistencial da Paraíba in Campina Grande, State of Paraiba, Brazil. Survival curves were estimated using the Kaplan–Meier method, Cox regression, and conditional decision tree.

Results.

Women with triple-negative molecular subtypes had a shorter survival time compared to women with positive hormone receptors. The addition of hormone therapy reduced the risk of a patient dying by 5.5%, and the risk of a HER2-positive patient dying was 34.5% lower compared to those who were negative for this gene. Patients undergoing hormone therapy had a median survival time of 4 753 days.

Conclusions.

This paper shows a favorable scenario for the use of immunotherapy for patients with HER2 overexpression. Further studies could assess the effectiveness of immunotherapy in patients with other conditions, to favor the prognosis and better quality of life for the patient.

Machine-learning algorithm as a prognostic tool in non-obstructive acute-on-chronic kidney disease in the cat

OBJECTIVES

The aim of this study was to develop an algorithm capable of predicting short- and medium-term survival in cases of intrinsic acute-on-chronic kidney disease (ACKD) in cats.

METHODS

The medical record database was searched to identify cats hospitalised for acute clinical signs and azotaemia of at least 48 h duration and diagnosed to have underlying chronic kidney disease based on ultrasonographic renal abnormalities or previously documented azotaemia. Cases with postrenal azotaemia, exposure to nephrotoxicants, feline infectious peritonitis or neoplasia were excluded. Clinical variables were combined in a clinical severity score (CSS). Clinicopathological and ultrasonographic variables were also collected. The following variables were tested as inputs in a machine learning system: age, body weight (BW), CSS, identification of small kidneys or nephroliths by ultrasonography, serum creatinine at 48 h (Crea48), spontaneous feeding at 48 h (SpF48) and aetiology. Outputs were outcomes at 7, 30, 90 and 180 days. The machine-learning system was trained to develop decision tree algorithms capable of predicting outputs from inputs. Finally, the diagnostic performance of the algorithms was calculated.

RESULTS

Crea48 was the best predictor of survival at 7 days (threshold 1043 µmol/l, sensitivity 0.96, specificity 0.53), 30 days (threshold 566 µmol/l, sensitivity 0.70, specificity 0.89) and 90 days (threshold 566 µmol/l, sensitivity 0.76, specificity 0.80), with fewer cats still alive when their Crea48 was above these thresholds. A short decision tree, including age and Crea48, predicted the 180-day outcome best. When Crea48 was excluded from the analysis, the generated decision trees included CSS, age, BW, SpF48 and identification of small kidneys with an overall diagnostic performance similar to that using Crea48.

CONCLUSIONS AND RELEVANCE

Crea48 helps predict short- and medium-term survival in cats with ACKD. Secondary variables that helped predict outcomes were age, CSS, BW, SpF48 and identification of small kidneys.

A Novel Six Autophagy-Related Genes Signature Associated With Outcomes and Immune Microenvironment in Lower-Grade Glioma

Since autophagy and the immune microenvironment are deeply involved in the tumor development and progression of Lower-grade gliomas (LGG), our study aimed to construct an autophagy-related risk model for prognosis prediction and investigate the relationship between the immune microenvironment and risk signature in LGG. Therefore, we identified six autophagy-related genes (BAG1, PTK6, EEF2, PEA15, ITGA6, and MAP1LC3C) to build in the training cohort (n = 305 patients) and verify the prognostic model in the validation cohort (n = 128) and the whole cohort (n = 433), based on the data from The Cancer Genome Atlas (TCGA). The six-gene risk signature could divide LGG patients into high- and low-risk groups with distinct overall survival in multiple cohorts (all p < 0.001). The prognostic effect was assessed by area under the time-dependent ROC (t-ROC) analysis in the training, validation, and whole cohorts, in which the AUC value at the survival time of 5 years was 0.837, 0.755, and 0.803, respectively. Cox regression analysis demonstrated that the risk model was an independent risk predictor of OS (HR > 1, p < 0.05). A nomogram including the traditional clinical parameters and risk signature was constructed, and t-ROC, C-index, and calibration curves confirmed its robust predictive capacity. KM analysis revealed a significant difference in the subgroup analyses' survival. Functional enrichment analysis revealed that these autophagy-related signatures were mainly involved in the phagosome and immune-related pathways. Besides, we also found significant differences in immune cell infiltration and immunotherapy targets between risk groups. In conclusion, we built a powerful predictive signature and explored immune components (including immune cells and emerging immunotherapy targets) in LGG.

A Novel S100 Family-Based Signature Associated with Prognosis and Immune Microenvironment in Glioma

Background

Glioma is the most common central nervous system (CNS) cancer with a short survival period and a poor prognosis. The S100 family gene, comprising 25 members, relates to diverse biological processes of human malignancies. Nonetheless, the significance of S100 genes in predicting the prognosis of glioma remains largely unclear. We aimed to build an S100 family-based signature for glioma prognosis.

Methods

We downloaded 665 and 313 glioma patients, respectively, from The Cancer Genome Atlas (TCGA) and Chinese Glioma Genome Atlas (CGGA) database with RNAseq data and clinical information. This study established a prognostic signature based on the S100 family genes through multivariate COX and LASSO regression. The Kaplan-Meier curve was plotted to compare overall survival (OS) among groups, whereas Receiver Operating Characteristic (ROC) analysis was performed to evaluate model accuracy. A representative gene S100B was further verified by in vitro experiments.

Results

An S100 family-based signature comprising 5 genes was constructed to predict the glioma that stratified TCGA-derived cases as a low- or high-risk group, whereas the significance of prognosis was verified based on CGGA-derived cases. Kaplan-Meier analysis revealed that the high-risk group was associated with the dismal prognosis. Furthermore, the S100 family-based signature was proved to be closely related to immune microenvironment. In vitro analysis showed S100B gene in the signature promoted glioblastoma (GBM) cell proliferation and migration.

Conclusions

We constructed and verified a novel S100 family-based signature associated with tumor immune microenvironment (TIME), which may shed novel light on the glioma diagnosis and treatment.

Risk Prediction of Cardiovascular Events by Exploration of Molecular Data with Explainable Artificial Intelligence

Cardiovascular diseases (CVD) annually take almost 18 million lives worldwide. Most lethal events occur months or years after the initial presentation. Indeed, many patients experience repeated complications or require multiple interventions (recurrent events). Apart from affecting the individual, this leads to high medical costs for society. Personalized treatment strategies aiming at prediction and prevention of recurrent events rely on early diagnosis and precise prognosis. Complementing the traditional environmental and clinical risk factors, multi-omics data provide a holistic view of the patient and disease progression, enabling studies to probe novel angles in risk stratification. Specifically, predictive molecular markers allow insights into regulatory networks, pathways, and mechanisms underlying disease. Moreover, artificial intelligence (AI) represents a powerful, yet adaptive, framework able to recognize complex patterns in large-scale clinical and molecular data with the potential to improve risk prediction. Here, we review the most recent advances in risk prediction of recurrent cardiovascular events, and discuss the value of molecular data and biomarkers for understanding patient risk in a systems biology context. Finally, we introduce explainable AI which may improve clinical decision systems by making predictions transparent to the medical practitioner.

Can machine learning improve mortality prediction following cardiac surgery?

OBJECTIVES

Interest in the clinical usefulness of machine learning for risk prediction has bloomed recently. Cardiac surgery patients are at high risk of complications and therefore presurgical risk assessment is of crucial relevance. We aimed to compare the performance of machine learning algorithms over traditional logistic regression (LR) model to predict in-hospital mortality following cardiac surgery.

METHODS

A single-centre data set of prospectively collected information from patients undergoing adult cardiac surgery from 1996 to 2017 was split into 70% training set and 30% testing set. Prediction models were developed using neural network, random forest, naive Bayes and retrained LR based on features included in the EuroSCORE. Discrimination was assessed using area under the receiver operating characteristic curve, and calibration analysis was undertaken using the calibration belt method. Model calibration drift was assessed by comparing Goodness of fit χ2 statistics observed in 2 equal bins from the testing sample ordered by procedure date.

RESULTS

A total of 28 761 cardiac procedures were performed during the study period. The in-hospital mortality rate was 2.7%. Retrained LR [area under the receiver operating characteristic curve 0.80; 95% confidence interval (CI) 0.77-0.83] and random forest model (0.80; 95% CI 0.76-0.83) showed the best discrimination. All models showed significant miscalibration. Retrained LR proved to have the weakest calibration drift.

CONCLUSIONS

Our findings do not support the hypothesis that machine learning methods provide advantage over LR model in predicting operative mortality after cardiac surgery.

Comprehensive In Silico Analysis of a Novel Serum Exosome-Derived Competitive Endogenous RNA Network for Constructing a Prognostic Model for Glioblastoma

Purpose

Glioblastoma (GBM) is one of the most aggressive brain tumors with high mortality, and tumor-derived exosomes provide new insight into the mechanisms of GBM tumorigenesis, metastasis and therapeutic resistance. We aimed to establish an exosome-derived competitive endogenous RNA (ceRNA) network for constructing a prognostic model for GBM.

Methods

We obtained the expression profiles of long noncoding RNAs (lncRNAs), miRNAs, and mRNAs from the GEO and TCGA databases and identified differentially expressed RNAs in GBM to construct a ceRNA network. By performing lasso and multivariate Cox regression analyses, we identified optimal prognosis-related differentially expressed lncRNAs (DElncRNAs) and generated a risk score model termed the exosomal lncRNA (exo-lncRNA) signature. The exo-lncRNA signature was subsequently validated in the CGGA GBM cohort. Finally, a novel prognostic nomogram was constructed based on the exo-lncRNA signature and clinicopathological parameters and validated in the CGGA external cohort. Based on the ceRNA hypothesis, oncocers were identified based on highly positive correlations between lncRNAs and mRNAs mediated by the same miRNAs. Furthermore, regression analyses were performed to assess correlations between the expression abundances of lncRNAs in tumors and exosomes.

Results

A total of 45 DElncRNAs, six DEmiRNAs, and 38 DEmRNAs were identified, and an exosome-derived ceRNA network was built. Three optimal prognostic-related DElncRNAs, HOTAIR (HR=0.341, P<0.001), SOX21-AS1 (HR=0.30, P<0.001), and STEAP3-AS1 (HR=2.47, P<0.001), were included to construct the exo-lncRNA signature, which was further proven to be an independent prognostic factor. The novel prognostic nomogram was constructed based on the exo-lncRNA signature, patient age, pharmacotherapy, radiotherapy, IDH mutation status, and MGMT promoter status, with a concordance index of 0.878. ROC and calibration plots both suggested that the nomogram had beneficial discrimination and predictive abilities. A total of 11 pairs of prognostic oncocers were identified. Regression analysis suggested excellent consistency of the expression abundance of the three exosomal lncRNAs between exosomes and tumor tissues.

Conclusions

Exosomal lncRNAs may serve as promising prognostic predictors and therapeutic targets. The prognostic nomogram based on the exo-lncRNA signature might provide an intuitive method for individualized survival prediction and facilitate better treatment strategies.

Identification of mortality-risk-related missense variant for renal clear cell carcinoma using deep learning

Introduction:

Kidney renal clear cell carcinoma (KIRCC) is a highly heterogeneous and lethal cancer that can arise in patients with renal disease. DeepSurv combines a deep feed-forward neural network with a Cox proportional hazards function and could provide optimized survival results compared with convenient survival analysis.

Methods:

This study used an improved DeepSurv algorithm to identify the candidate genes to be targeted for treatment on the basis of the overall mortality status of KIRCC subjects. All the somatic mutation missense variants of KIRCC subjects were abstracted from TCGA-KIRC database.

Results:

The improved DeepSurv model (95.1%) achieved greater balanced accuracy compared with the DeepSurv model (75%), and identified 610 high-risk variants associated with overall mortality. The results of gene differential expression analysis also indicated nine KIRCC mortality-risk-related pathways, namely the tRNA charging pathway, the D-myo-inositol-5-phosphate metabolism pathway, the DNA double-strand break repair by nonhomologous end-joining pathway, the superpathway of inositol phosphate compounds, the 3-phosphoinositide degradation pathway, the production of nitric oxide and reactive oxygen species in macrophages pathway, the synaptic long-term depression pathway, the sperm motility pathway, and the role of JAK2 in hormone-like cytokine signaling pathway. The biological findings in this study indicate the KIRCC mortality-risk-related pathways were more likely to be associated with cancer cell growth, cancer cell differentiation, and immune response inhibition.

Conclusion:

The results proved that the improved DeepSurv model effectively classified mortality-related high-risk variants and identified the candidate genes. In the context of KIRCC overall mortality, the proposed model effectively recognized mortality-related high-risk variants for KIRCC.

An Autophagy-Related Gene Signature Associated With Clinical Prognosis and Immune Microenvironment in Gliomas

Glioma is one of the leading causes of death from cancer, and autophagy-related genes (ARGs) play an important role in glioma occurrence, progression, and treatment. In this study, the gene expression profiles and clinical data of glioma patients were obtained from The Cancer Genome Atlas (TCGA) and the Chinese Glioma Genome Atlas (CGGA), respectively. ARGs were obtained from the Human Autophagy Database. We analyzed the expression of the ARGs in glioma and found that 73 ARGs were differentially expressed in tumor and normal tissues. Univariate Cox regression analysis was used to identify prognostic differentially expressed ARGs (PDEARGs). Least absolute shrinkage and selection operator (LASSO) and multivariate Cox regression analyses were performed on the PDEARGs to determine the risk genes; and BRIC5, NFE2L2, GABARAP, IKBKE, BID, MAPK3, FKBP1B, MAPK8IP1, PRKCQ, CX3CL1, NPC1, HSP90AB1, DAPK2, SUPT20H, and PTEN were selected to establish a prognostic risk score model for TCGA and CGGA cohorts. This model accurately stratified patients with different survival outcomes, and the autophagy-related signature was also appraised as being an independent prognostic factor. We also constructed a prognostic nomogram using risk score, age, gender, WHO grade, isocitrate dehydrogenase (IDH) mutation status, and 1p/19q co-deletion status; and the calibration plots showed excellent prognostic performance. Finally, Pearson correlation analysis suggested that the ARG signature also played an essential role in the tumor immune microenvironment. In summary, we constructed and verified a novel autophagy-related signature that was tightly associated with the tumor immune microenvironment and could serve as an independent prognostic biomarker in gliomas.

Integrated analysis of serum lipid profile for predicting clinical outcomes of patients with malignant biliary tumor

Background

Serum lipids were reported to be the prognostic factors of various cancers, but their prognostic value in malignant biliary tumor (MBT) patients remains unclear. Thus we aim to assess and compare prognosis values of different serum lipids, and construct a novel prognostic nomogram based on serum lipids.

Methods

Patients with a confirmed diagnosis of MBT at our institute from 2003 to 2017 were retrospectively reviewed. Prognosis-related factors were identified via univariate and multivariate Cox regression analyses. Then the novel prognostic nomogram and a 3-tier staging system were constructed based on these factors and further compared to the TNM staging system.

Results

A total of 368 patients were included in this study. Seven optimal survival-related factors—TC/HDL >  10.08, apolipoprotein B >  0.9 g/L, lipoprotein> 72 mg/L, lymph node metastasis, radical cure, CA199 > 37 U/mL, and tumor differentiation —were included to construct the prognostic nomogram. The C-indexes in training and validation sets were 0.738 and 0.721, respectively. Besides, ROC curves, calibration plots, and decision curve analysis all suggested favorable discrimination and predictive ability. The nomogram also performed better predictive ability than the TNM system and nomogram without lipid parameters. And the staging system based on nomogram also presented better discriminative ability than TNM system (P < 0.001).

Conclusions

The promising prognostic nomogram based on lipid parameters provided an intuitive method for performing survival prediction and facilitating individualized treatment and was a great complement to the TNM staging system in predicting overall survival.

Prognostic Implication of a Metabolism-Associated Gene Signature in Lung Adenocarcinoma

Lung cancer is the most common cancer worldwide, leading to high mortality each year. Metabolic pathways play a vital role in the initiation and progression of lung cancer. We aimed to establish a prognostic prediction model for lung adenocarcinoma (LUAD) patients based on a metabolism-associated gene (MTG) signature. Differentially expressed (DE)-MTGs were screened from The Cancer Genome Atlas (TCGA) LUAD cohorts. Univariate Cox regression analysis was performed on these DE-MTGs to identify genes significantly correlated with prognosis. Least absolute shrinkage and selection operator (LASSO) regression was performed on the resulting genes to establish an optimal risk model. Survival analysis was used to assess the prognostic ability of the model. The prognostic value of the gene signature was further validated in independent Gene Expression Omnibus (GEO) datasets. A gene signature with 13 metabolic genes was identified as an independent prognostic factor. Kaplan-Meier survival analysis demonstrated the good performance of the risk model in both TCGA training and GEO validation cohorts. Finally, a nomogram incorporating clinical parameters and the metabolic gene signature was constructed to help individualize outcome predictions. The calibration curves showed excellent agreement between the actual and predicted survival.

Preoperative C-reactive protein-to-albumin ratio and clinical outcomes after resection of colorectal liver metastases

PURPOSE

Accumulating evidence suggests that the inflammatory tumor microenvironment can potentiate tumor progression and metastasis. The C-reactive protein-to-albumin ratio (CAR) is a novel inflammation-based prognostic score. This study was performed to examine the associations of the preoperative CAR with clinical outcomes in patients with colorectal liver metastases (CRLM) after curative resection.

METHODS

We retrospectively assessed the preoperative CAR in 184 patients who underwent curative resection for CRLM from November 2001 to January 2018 at Kumamoto University (Kumamoto, Japan). The optimal cutoff level of the preoperative CAR was determined by survival classification and regression tree (CART) analysis. We compared clinicopathological factors and prognoses between the high-CAR and low-CAR groups. A Cox proportional hazards model was used to calculate hazard ratios (HRs), controlling for potential confounders.

RESULTS

A higher preoperative CAR was associated with worse overall survival (OS) (p < 0.0001) and recurrence-free survival (RFS) (p = 0.003). Applying survival CART analysis, the high-CAR group comprised 33 patients (17.9%). In the multivariate analyses, a high CAR was independently associated with shorter OS (HR, 2.82; 95% confidence interval, 1.63-4.72; p = 0.0004) and RFS (HR, 1.62; 95% confidence interval, 1.02-2.49; p = 0.040). A high CAR was associated with a large tumor size, high serum carcinoembryonic antigen and carbohydrate antigen 19-9 levels, high intraoperative blood loss, and more postoperative complications.

CONCLUSION

A high preoperative CAR is associated with shorter OS and RFS and might serve as a prognostic marker for patients with CRLM after curative resection.

Use of Machine Learning for Prediction of Patient Risk of Postoperative Complications After Liver, Pancreatic, and Colorectal Surgery

BACKGROUND

Surgical resection is the only potentially curative treatment for patients with colorectal, liver, and pancreatic cancers. Although these procedures are performed with low mortality, rates of complications remain relatively high following hepatopancreatic and colorectal surgery.

METHODS

The American College of Surgeons (ACS) National Surgical Quality Improvement Program was utilized to identify patients undergoing liver, pancreatic and colorectal surgery from 2014 to 2016. Decision tree models were utilized to predict the occurrence of any complication, as well as specific complications. To assess the variability of the performance of the classification trees, bootstrapping was performed on 50% of the sample.

RESULTS

Algorithms were derived from a total of 15,657 patients who met inclusion criteria. The algorithm had a good predictive ability for the occurrence of any complication, with a C-statistic of 0.74, outperforming the ASA (C-statistic 0.58) and ACS-Surgical Risk Calculator (C-statistic 0.71). The algorithm was able to predict with high accuracy thirteen out of the seventeen complications analyzed. The best performance was in the prediction of stroke (C-statistic 0.98), followed by wound dehiscence, cardiac arrest, and progressive renal failure (all C-statistic 0.96). The algorithm had a good predictive ability for superficial SSI (C-statistic 0.76), organ space SSI (C-statistic 0.76), sepsis (C-statistic 0.79), and bleeding requiring transfusion (C-statistic 0.79).

CONCLUSION

Machine learning was used to develop an algorithm that accurately predicted patient risk of developing complications following liver, pancreatic, or colorectal surgery. The algorithm had very good predictive ability to predict specific complications and demonstrated superiority over other established methods.

The dangerous and contradictory prognostic significance of PVR<3WU when TAPSE<16mm in postcapillary pulmonary hypertension

Aims

In 2019, pulmonary vascular resistance (PVR) < 3WU was adopted to stratify patients at low risk in pulmonary hypertension due to left heart disease (PH‐LHD) as well those with isolated PH‐LHD. We sought to evaluate whether supervised machine learning with decision tree analysis, which provides more information than Cox Proportional analysis by forming a hierarchy of multiple covariates, confirms this risk stratification.

Methods and results

Two hundred two consecutive patients (mean age: 69 ± 11 years, female: 42%) with mean pulmonary artery pressure ≥ 20 mmHg and wedge pressure > 15 mmHg were recruited. Transpulmonary pressure gradient ⩾̸ 12 mmHg, PVR ⩾̸ 3WU, diastolic pressure gradient ⩾̸ 7 mmHg, pulmonary arterial capacitance < 1.1 mL/mmHg, tricuspid annular plane systolic excursion (TAPSE) < 16 mm, peak systolic tissue Doppler velocity < 10 cm/s, right ventricular end‐diastolic area ⩾̸ 25 cm² were the seven categorical values entered into the model due to their prognostic significance in PH. We used the chi‐squared automatic interaction detection method to predict mortality. Each node and branch were compared using survival analysis at 6‐year follow‐up. Mean pulmonary artery pressure, wedge pressure, cardiac index, and PVR were 40.3 ± 10.0 mmHg, 22.3 ± 7.1 mmHg, 2.9 ± 0.8 L/min/m², and 3.6 ± 2.1WU, respectively. Among the seven dichotomous, TAPSE was first selected following by PVR. Compared with patients with PVR < 3WU and TAPSE ⩾̸ 16 mm, patients with PVR ⩾̸ 3WU and TAPSE ⩾̸ 16 mm, or patients with PVR ⩾̸ 3WU and TAPSE<16 mm had significantly increased mortality, HR = 3.0, 95% CI = [1.4–6.4], P = 0.006 and HR = 3.3, 95% CI = [1.6–6.9], P = 0.002, respectively, while patients with PVR < 3WU and TAPSE < 16 mm exhibited the worst prognosis, HR = 7.2, 95% CI = [3.3–15.9], P = 0.0001.

Conclusions

Used for solving regression and classification problems, decision tree analysis confirms that PVR and TAPSE have to be analysed together in PH‐LHD and revealed the dangerous and contradictory prognostic significance of PVR < 3WU when TAPSE<16 mm.

Development of a Nomogram With Alternative Splicing Signatures for Predicting the Prognosis of Glioblastoma: A Study Based on Large-Scale Sequencing Data

Purpose: Alternative splicing (AS) was reported to play a vital role in development and progression of glioblastoma (GBM), the most common and fatal brain tumor. Systematic analysis of survival-associated AS event profiles and prognostic prediction model based on multiple AS events in GBM was needed. Methods: Genome-wide AS and RNA sequencing profiles were generated in 152 patients with GBM in the cancer genome atlas (TCGA). Prognosis-associated AS events were screened by integrated Cox regression analysis to construct the prognostic risk score model in the training cohort (n = 101). The AS-based signature and clinicopathologic parameters were applied to construct a prognostic nomogram for 0.5-, 1-, and 3-year OS prediction. Finally, the regulatory networks between prognostic AS events and splicing factors (SFs) were constructed. Results: A total of 1,598 prognosis-related AS events from 1,183 source genes were determined. Eight prognostic risk score model based on integrated AS events and 7 AS types were established, respectively. Concordance index (C-index) and receiver operating characteristic (ROC) curve analysis demonstrated powerful ability in distinguishing patients' outcomes. Only Alternate Donor site (AD) and Exon Skip (ES) signature out of the eight types of AS signature were identified as independent prognostic factors for GBM, which was validated in the internal validation cohort. The nomogram with age, new event, pharmaceutical therapy, radiation therapy, AD signature and ES signature were constructed, with C-index of 0.892 (95% CI, 0.853-0.931; P = 5.13 × 10^-15). Calibration plots, ROC, and decision curve analysis suggested excellent predictive performance for the nomogram in both TCGA training cohort and validation cohort. Splicing network indicated distinguished correlations between prognostic AS events and SFs in GBM patients. Conclusions: AS-based prediction model could serve as a promising prognostic predictor and potential therapeutic target for GBM, facilitating better treatment strategies in clinical practice.

Prognostic model and nomogram construction based on autophagy signatures in lower grade glioma

The median survival time of lower grade glioma (LGG) tumors spans a wide range of 2-10 years and is highly dependent on the molecular characteristics and tumor location. Currently, there is no prognostic predictor for these tumors based on autophagy-related (ATG) genes. A prognostic risk score model based on the most significant seven ATG genes was established for LGG. These seven genes, including GRID2, FOXO1, MYC, PTK6, IKBKE, BIRC5, and TP73, have been screened as potentially therapeutic targets. The Kaplan-Meier survival curve analyses validated that patients with high or low risk scores had significantly different overall survival. Following the multivariate Cox regression and area under the ROC curve (AUC) analysis, a final prognostic model based on age, World Health Organization grade, 1p19q-codeletion status, and ATG risk score was performed as an independent prognostic indicator (training set: p = 4.09E-05, AUC = 0.901; validation set-1: p = .00069, AUC = 0.808; validation set-2: p = .0376, AUC = 0.830). Subsequently, a prognostic nomogram was constructed for individualized survival prediction. The calibration plots showed excellent predict efficiency between probability and actual overall survival. In this study, we provided several potential biomarkers for further developing potentially therapeutic targets of LGG. We also established a prognostic model and nomogram to improve the clinical glioma management and assist individualized survival prediction.

Development and validation of a nomogram with an autophagy-related gene signature for predicting survival in patients with glioblastoma

Glioblastoma (GBM) is the most common brain tumor with significant morbidity and mortality. Autophagy plays a vital role in GBM development and progression. We aimed to establish an autophagy-related multigene expression signature for individualized prognosis prediction in patients with GBM. Differentially expressed autophagy-related genes (DE-ATGs) in GBM and normal samples were screened using TCGA. Univariate and multivariate Cox regression analyses were performed on DE-ATGs to identify the optimal prognosis-related genes. Consequently, NRG1 (HR=1.142, P=0.008), ITGA3 (HR=1.149, P=0.043), and MAP1LC3A (HR=1.308, P=0.014) were selected to establish the prognostic risk score model and validated in the CGGA validation cohort. GSEA revealed that these genes were mainly enriched in cancer- and autophagy-related KEGG pathways. Kaplan-Meier survival analysis demonstrated that patients with high risk scores had significantly poorer overall survival (OS, log-rank P= 6.955×10^-5). The autophagy signature was identified as an independent prognostic factor. Finally, a prognostic nomogram including the autophagy signature, age, pharmacotherapy, radiotherapy, and IDH mutation status was constructed, and TCGA/CGGA-based calibration plots indicated its excellent predictive performance. The autophagy-related three-gene risk score model could be a prognostic biomarker and suggest therapeutic targets for GBM. The prognostic nomogram could assist individualized survival prediction and improve treatment strategies.

Oblique Survival Trees in Discrete Event Time Analysis

One of the main objectives of survival analysis is to predict the failure time that is usually considered as a continuous variable. In longitudinal studies, the data are often collected at every certain period of time, for example, monthly, quarterly, or yearly. Such data require appropriate techniques to handle the discrete time values that often have incomplete information about the failure occurrence-so-called "censored cases." Tree-based models are common, assumption-free methods of survival prediction. In this paper, the author proposes three recursive partitioning techniques able to cope with discrete-time censored survival data, which, in contrast to already-existing models limited to univariate trees, allow splits to have a form of any hyperplane. The performance of proposed methods, expressed as a mean absolute error, was examined on the basis of both synthetic and real data sets available in the literature and compared with existing tree-based models. To demonstrate the applicability of the methods in identifying subgroups of patients with a similar survival experience and to assess the influence of covariates on the risk of failure, a Veteran's Administration lung cancer data set was used. The results confirm proposed models to be good prediction tools for discrete-time survival data.

Identifying causal mechanisms in health care interventions using classification tree analysis

RATIONALE, AIMS, AND OBJECTIVES

Mediation analysis identifies causal pathways by testing the relationships between the treatment, the outcome, and an intermediate variable that mediates the relationship between the treatment and outcome. This paper introduces classification tree analysis (CTA), a machine-learning procedure, as an alternative to conventional methods for analysing mediation effects.

METHOD

Using data from the JOBS II study, we compare CTA to structural equation models (SEMs) by assessing their consistency in revealing mediation effects on 2 outcomes; reemployment (a binary variable) and depressive symptoms (a continuous variable). Because study participants were not randomized sequentially to both treatment and mediator, an additional model was generated including baseline covariates to strengthen the validity of some key identifying assumptions required of all mediation analyses.

RESULTS

Using SEM, no statistically significant treatment or mediated effects were found for either outcome. In contrast, CTA found a significant treatment effect for reemployment (P = .047) and a mediated pathway for individuals in the treatment group (P = .014). No CTA model could be generated for the reemployment outcome. When covariates were added to the model, CTA found numerous interactions, while SEM found no effects.

CONCLUSIONS

CTA may uncover mediation effects where conventional approaches do not, because CTA does not require any assumptions about the distribution of variables nor of the functional form of the model, and CTA will systematically identify all statistically viable interactions. The versatility of CTA enables the investigator to explore the theorized underlying causal mechanism of an intervention in a much more comprehensive manner than conventional mediation analytic approaches.

Estimating causal effects for survival (time-to-event) outcomes by combining classification tree analysis and propensity score weighting

RATIONALE, AIMS AND OBJECTIVES

A common approach to assessing treatment effects in nonrandomized studies with time-to-event outcomes is to estimate propensity scores and compute weights using logistic regression, test for covariate balance, and then estimate treatment effects using Cox regression. A machine-learning alternative-classification tree analysis (CTA)-used to generate propensity scores and to estimate treatment effects in time-to-event data may identify complex relationships between covariates not found using conventional regression-based approaches.

METHOD

Using empirical data, we identify all statistically valid CTA propensity score models and then use them to compute strata-specific, observation-level propensity score weights that are subsequently applied in outcomes analyses. We compare findings obtained using this framework to the conventional method, by evaluating covariate balance and treatment effect estimates obtained using Cox regression and a weighted CTA outcomes model.

RESULTS

All models had some imbalanced covariates. Nevertheless, treatment effect estimates were generally consistent across all weighted models.

CONCLUSIONS

In the study sample, given that all approaches elicited similar results, using CTA increased confidence that bias could not be reduced any further. Because the CTA algorithm identifies all statistically valid propensity score models for a sample, it is most likely to identify a correctly specified propensity score model-and therefore should be used either to confirm results using traditional methods, or to reveal biases that may be missed by traditional methods. Moreover, given that the true treatment effect is never known in observational data, CTA should be considered for estimating outcomes because no statistical assumptions are required.

Minimizing imbalances on patient characteristics between treatment groups in randomized trials using classification tree analysis

RATIONALE, AIMS, AND OBJECTIVES

Randomization ensures that treatment groups do not differ systematically in their characteristics, thereby reducing threats to validity that may otherwise explain differences in outcomes. Large observed imbalances in patient characteristics may indicate that selection bias is being introduced into the treatment allocation process. We introduce classification tree analysis (CTA) as a novel algorithmic approach for identifying potential imbalances in characteristics and their interactions when provisionally assigning each new participant to one or the other treatment group. The participant is then permanently assigned to the treatment group that elicits either no or less imbalance than if assigned to the alternate group.

METHOD

Using data on participant characteristics from a clinical trial, we compare 3 different treatment allocation approaches: permuted block randomization (the original allocation method), minimization, and CTA. Treatment allocation performance is assessed by examining balance of all 17 patient characteristics between study groups for each of the allocation techniques.

RESULTS

While all 3 treatment allocation techniques achieved excellent balance on main effect variables, Classification tree analysis further identified imbalances on interactions and in the distributions of some of the continuous variables.

CONCLUSIONS

Classification tree analysis offers an algorithmic procedure that may be used with any randomization methodology to identify and then minimize linear, nonlinear, and interactive effects that induce covariate imbalance between groups. Investigators should consider using the CTA approach as a real-time complement to randomization for any clinical trial to safeguard the treatment allocation process against bias.