Survival Tree Research Articles

A Survival Tree of Advanced Melanoma Patients with Brain Metastases Treated with Immune Checkpoint Inhibitors.

The efficacy of immune checkpoint inhibitors (ICIs) in patients with advanced melanoma that develop brain metastases (BM) remains unpredictable. In this study, we aimed to identify prognostic factors in patients with melanoma BM who are treated with ICIs. Data from advanced melanoma patients with BM treated with ICIs in any line between 2013 and 2020 were obtained from the Dutch Melanoma Treatment Registry. Patients were included from the time of the treatment of BM with ICIs. Survival tree analysis was performed with clinicopathological parameters as potential classifiers and overall survival (OS) as the response variable. In total, 1278 patients were included. Most patients were treated with ipilimumab-nivolumab combination therapy (45%). The survival tree analysis resulted in 31 subgroups. The median OS ranged from 2.7 months to 35.7 months. The strongest clinical parameter associated with survival in advanced melanoma patients with BM was the serum lactate dehydrogenase (LDH) level. Patients with elevated LDH levels and symptomatic BM had the worst prognosis. The clinicopathological classifiers identified in this study can contribute to optimizing clinical studies and can aid doctors in giving an indication of the patients' survival based on their baseline and disease characteristics.

Integrated analysis of ovarian cancer patients from prospective transcription factor activity reveals subtypes of prognostic significance

Transcription factors are protein molecules that act as regulators of gene expression. Aberrant protein activity of transcription factors can have a significant impact on tumor progression and metastasis in tumor patients. In this study, 868 immune-related transcription factors were identified from the transcription factor activity profile of 1823 ovarian cancer patients. The prognosis-related transcription factors were identified through univariate Cox analysis and random survival tree analysis, and two distinct clustering subtypes were subsequently derived based on these transcription factors. We assessed the clinical significance and genomics landscape of the two clustering subtypes and found statistically significant differences in prognosis, response to immunotherapy, and chemotherapy among ovarian cancer patients with different subtypes. Multi-scale Embedded Gene Co-expression Network Analysis was used to identify differential gene modules between the two clustering subtypes, which allowed us to conduct further analysis of biological pathways that exhibited significant differences between them. Finally, a ceRNA network was constructed to analyze lncRNA-miRNA-mRNA regulatory pairs with differential expression levels between two clustering subtypes. We expected that our study may provide some useful references for stratifying and treating patients with ovarian cancer.

Machine-learning predicts time-series prognosis factors in metastatic prostate cancer patients treated with androgen deprivation therapy

Machine learning technology is expected to support diagnosis and prognosis prediction in medicine. We used machine learning to construct a new prognostic prediction model for prostate cancer patients based on longitudinal data obtained from age at diagnosis, peripheral blood and urine tests of 340 prostate cancer patients. Random survival forest (RSF) and survival tree were used for machine learning. In the time-series prognostic prediction model for metastatic prostate cancer patients, the RSF model showed better prediction accuracy than the conventional Cox proportional hazards model for almost all time periods of progression-free survival (PFS), overall survival (OS) and cancer-specific survival (CSS). Based on the RSF model, we created a clinically applicable prognostic prediction model using survival trees for OS and CSS by combining the values of lactate dehydrogenase (LDH) before starting treatment and alkaline phosphatase (ALP) at 120 days after treatment. Machine learning provides useful information for predicting the prognosis of metastatic prostate cancer prior to treatment intervention by considering the nonlinear and combined impacts of multiple features. The addition of data after the start of treatment would allow for more precise prognostic risk assessment of patients and would be beneficial for subsequent treatment selection.

MP08-03 PROGRESSION-FREE SURVIVAL AS SURROGATE ENDPOINT IN HIGH-RISK NON-MUSCLE INVASIVE BLADDER CANCER STUDIES: RESULTS FROM A MACHINE LEARNING-BASED ANALYSIS OF A LARGE MULTI-INSTITUTIONAL DATABASE

MP08-03 PROGRESSION-FREE SURVIVAL AS SURROGATE ENDPOINT IN HIGH-RISK NON-MUSCLE INVASIVE BLADDER CANCER STUDIES: RESULTS FROM A MACHINE LEARNING-BASED ANALYSIS OF A LARGE MULTI-INSTITUTIONAL DATABASE

Machine learning for predicting the survival in osteosarcoma patients: Analysis based on American and Hebei Province cohort.

Osteosarcoma, a rare malignant tumor, has a poor prognosis. This study aimed to find the best prognostic model for osteosarcoma. There were 2912 patients included from the SEER database and 225 patients from Hebei Province. Patients from the SEER database (2008-2015) were included in the development dataset. Patients from the SEER database (2004-2007) and Hebei Province cohort were included in the external test datasets. The Cox model and three tree-based machine learning algorithms (survival tree [ST], random survival forest [RSF] and gradient boosting machine [GBM]) were used to develop the prognostic models by 10-fold cross-validation with 200 iterations. Additionally, performance of models in the multivariable group was compared with the TNM group. The 3-year and 5-year cancer specific survival (CSS) were 72.71% and 65.92% in the development dataset, respectively. The predictive ability in the multivariable group was superior to that in the TNM group. The calibration curves and consistency in the multivariable group were superior to those in the TNM group. The Cox and RSF models performed better than the ST and GBM models. A nomogram was constructed to predict the 3-year and 5-year CSS of osteosarcoma patients. The RSF model can be used as a nonparametric alternative to the Cox model. The constructed nomogram based on the Cox model can provide reference for clinicians to formulate specific therapeutic decisions both in America and China.

The Application of Artificial Intelligence to Investigate Long-Term Outcomes and Assess Optimal Margin Width in Hepatectomy for Intrahepatic Cholangiocarcinoma.

Intrahepatic cholangiocarcinoma (ICC) is associated with poor long-term outcomes, and limited evidence exists on optimal resection margin width. This study used artificial intelligence to investigate long-term outcomes and optimal margin width in hepatectomy for ICC. The study enrolled patients who underwent curative-intent resection for ICC between 1990 and 2020. The optimal survival tree (OST) was used to investigate overall (OS) and recurrence-free survival (RFS). An optimal policy tree (OPT) assigned treatment recommendations based on random forest (RF) counterfactual survival probabilities associated with each possible margin width between 0 and 20 mm. Among 600 patients, the median resection margin was 4 mm (interquartile range [IQR], 2-10). Overall, 379 (63.2 %) patients experienced recurrence with a 5-year RFS of 28.3 % and a 5-year OS of 38.7 %. The OST identified five subgroups of patients with different OS rates based on tumor size, a carbohydrate antigen 19-9 [CA19-9] level higher than 200 U/mL, nodal status, margin width, and age (area under the curve [AUC]: training, 0.81; testing, 0.69). The patients with tumors smaller than 4.8 cm and a margin width of 2.5 mm or greater had a relative increase in 5-year OS of 37 % compared with the entire cohort. The OST for RFS estimated a 46 % improvement in the 5-year RFS for the patients younger than 60 years who had small (<4.8 cm) well- or moderately differentiated tumors without microvascular invasion. The OPT suggested five optimal margin widths to maximize the 5-year OS for the subgroups of patients based on age, tumor size, extent of hepatectomy, and CA19-9 levels. Artificial intelligence OST identified subgroups within ICC relative to long-term outcomes. Although tumor biology dictated prognosis, the OPT suggested that different margin widths based on patient and disease characteristics may optimize ICC long-term survival.

Long-term outcomes of phenoclusters in severe tricuspid regurgitation.

Severe tricuspid regurgitation (TR) exhibits high 1-year morbidity and mortality, yet long-term cardiovascular risk overall and by subgroups remains unknown. This study characterizes 5-year outcomes and identifies distinct clinical risk profiles of severe TR. Patients were included from a large US tertiary referral center with new severe TR by echocardiography based on four-category American Society of Echocardiography grading scale between 2007 and 2018. Patients were categorized by TR etiology (with lead present, primary, and secondary) and by supervised recursive partitioning (survival trees) for outcomes of death and the composite of death or heart failure hospitalization. The Kaplan-Meier estimates and Cox regression models were used to evaluate any association by (i) TR etiology and (ii) groups identified by survival trees and outcomes over 5 years. Among 2379 consecutive patients with new severe TR, median age was 70 years, 61% were female, and 40% were black. Event rates (95% confidence interval) were 30.9 (29.0-32.8) events/100 patient-years for death and 49.0 (45.9-52.2) events/100 patient-years for the composite endpoint, with no significant difference by TR etiology. After applying supervised survival tree modeling, two separate groups of four phenoclusters with distinct clinical prognoses were separately identified for death and the composite endpoint. Variables discriminating both outcomes were age, albumin, blood urea nitrogen, right ventricular function, and systolic blood pressure (all P < 0.05). Patients with newly identified severe TR have high 5-year risk for death and death or heart failure hospitalization. Partitioning patients using supervised survival tree models, but not TR etiology, discriminated clinical risk. These data aid in identifying relevant subgroups in clinical trials of TR and clinical risk/benefit analysis for TR therapies.

A comparison of machine learning models and Cox proportional hazards models regarding their ability to predict the risk of gastrointestinal cancer based on metabolic syndrome and its components.

Little is known about applying machine learning (ML) techniques to identify the important variables contributing to the occurrence of gastrointestinal (GI) cancer in epidemiological studies. We aimed to compare different ML models to a Cox proportional hazards (CPH) model regarding their ability to predict the risk of GI cancer based on metabolic syndrome (MetS) and its components. A total of 41,837 participants were included in a prospective cohort study. Incident cancer cases were identified by following up with participants until December 2019. We used CPH, random survival forest (RSF), survival trees (ST), gradient boosting (GB), survival support vector machine (SSVM), and extra survival trees (EST) models to explore the impact of MetS on GI cancer prediction. We used the C-index and integrated Brier score (IBS) to compare the models. In all, 540 incident GI cancer cases were identified. The GB and SSVM models exhibited comparable performance to the CPH model concerning the C-index (0.725). We also recorded a similar IBS for all models (0.017). Fasting glucose and waist circumference were considered important predictors. Our study found comparably good performance concerning the C-index for the ML models and CPH model. This finding suggests that ML models may be considered another method for survival analysis when the CPH model's conditions are not satisfied.

Abstract P4-06-03: Pathological complete response (ypCR) in early and locally advanced breast cancer (LABC) patients treated with neoadjuvant chemotherapy in a middle-income country. Results from a real-world historical cohort

Abstract Introduction Breast cancer (BC) is the most frequent neoplasm in Colombia, with mortality rate increasing in past decade. The expected 5-year overall survival (OS) is &amp;lt; 80%, being the Latin American country with the worst prognosis. 2 factors related to this are higher frequency of advanced stages and higher frequency of aggressive tumor subtypes.Neoadjuvant chemotherapy (NACT), and ypCR, have demonstrated impact on the risk of relapse and death. This trial describes the characteristics, treatment patterns, clinical outcomes, and ypCR after NACT in a cohort of Colombian BC patients. Methods We included BC adult patients treated with NACT in 3 institutions in Colombia. Clinical, sociodemographic, and outcome variables were retrieved retrospectively from clinical charts. Univariate, bivariate and time-to-event analyses were performed. A partition survival tree was performed to explore interactions between variables. Statistical analyses were done in R software.The protocol was approved by the IRC and EC of Fundación Cardioinfantil and Hospital Militar Central. Results We included 312 patients treated between 2013 and 2019. 50.9% were ER and/or PgR positive, 28.5% HER2 positive and 20.5% triple negative breast cancer (TNBC). 75.9% of the patients had a LABC. TNBC patients were younger (median age 46 years), premenopausal (58%), had higher Ki67 index (Ki67 ≥ 20%= 91%), and a higher tumor grade (Grade 3= 56.7%). Most of the patients (91.7%), received Doxorubicin + Cyclophosphamide, and Taxanes (97.8%), 76.6% of TNBC patients received Carboplatin. Breast conserving surgery was performed in 53.7% of the patients, and only 38.9% received immediate breast reconstruction. 88.5% of the patients were treated with adjuvant radiotherapy. The ypCR rate was 34.6% and was achieved more frequently in HER2 positive subtype. The ypCR rates in TBNC were higher in patients treated with carboplatin (55% vs 20%, p=0.02). Other factors related to ypCR were high Ki67 (p= 0.0005) and higher tumor grade (p= 0.0034). Most of the relapses were distant (14.1%), with only 1.9% of local relapse rate. Table No 1 summarizes clinical outcomes. With a median follow up of 4.9 years, the 5 year OS was 88.2%. Clinical stage at diagnosis (p= 0.01); node involvement (p= 0.0014); ER (p= 0.0032); PgR (p= 0.0069); HER2 (p= 0.0334); phenotype (p=0.0145); type of surgery (p&amp;lt; 0.0001); ypCR (p=0.003); and relapse (p= &amp;lt; 0.00001) were related with OS. We did not found differences in the 5 year OS between stage II and III patients (87.8% vs 88%). The 5 year OS did not differ between ypCR and ypCR-IS (94.2% vs 93%). The relapse rate was 3.7% in ypCR vs 22.1% in non ypCR patients (p&amp;lt; 0.00001). The 5 year OS rate was 94.3% for ypCR and 85% for non ypCR patients (p= 0.0059). When analyzed by phenotype, ypCR was only related to risk of death among TNBC (p= 0.011). We perform a partition survival tree, identifying 6 subgroups with different OS behaviors: TNBC and non TBNC; low proliferative and high proliferative; early stage and locally advanced stage. Discussion We found an increased number of LABC. The ypCR rate was higher than described elsewhere, explained by higher proportion of TNBC and HER2 enriched tumors and differences in treatment selection in subgroups. In the TNBC subgroup the use of carboplatin was associated with higher ypCR. ypCR was related with risk of death and relapse, but with statistical significance for OS only in the TNBC population. This study offers a baseline characteristic of NACT BC cohort, that will allow us to design and implement strategies to improve outcomes in early and LABC. Table 1: Treatment outcomes by phenotype. Citation Format: William Mantilla, Maria-Fernanda Gonzalez, Sebastian Rojas, Mariana Borras-Osorio, Nicolas Molano-Gonzalez, Joaquin Guerra, Isabel Munevar, Diego Moran. Pathological complete response (ypCR) in early and locally advanced breast cancer (LABC) patients treated with neoadjuvant chemotherapy in a middle-income country. Results from a real-world historical cohort. [abstract]. In: Proceedings of the 2022 San Antonio Breast Cancer Symposium; 2022 Dec 6-10; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2023;83(5 Suppl):Abstract nr P4-06-03.

A novel preoperative inflammation score system established for postoperative prognosis predicting of intrahepatic cholangiocarcinoma

BackgroundInflammation is implicated in tumorigenesis and has been reported as an important prognostic factor in cancers. In this study, we aimed to develop and validate a novel inflammation score (IFS) system based on 12 inflammatory markers and explore its impact on intrahepatic cholangiocarcinoma (ICC) survival after hepatectomy.MethodsClinical data of 446 ICC patients undergoing surgical treatment were collected from the Primary Liver Cancer Big Data, and then served as a training cohort to establish the IFS. Furthermore, an internal validation cohort including 175 patients was used as internal validation cohort of the IFS. A survival tree analysis was used to divide ICC patients into three groups (low-, median-, and high- IFS-score groups) according to different IFS values. Kaplan-Meier (KM) curves were used to compare the overall survival (OS) and recurrence-free survival (RFS) rates among three different groups. Cox regression analyses were applied to explore the independent risk factors influencing OS and RFS.ResultsIn the training cohort, 149 patients were in the low-IFS-score group, 187 in the median-IFS-score group, and 110 in the high-IFS-score group. KM curves showed that the high-IFS-score group had worse OS and RFS rates than those of the low- and median-IFS-score groups (P < 0.001) in both the training and validation cohorts. Moreover, multivariable Cox analyses identified high IFS as an independent risk factor for OS and RFS in the training cohort. The area under the curve values for OS prediction of IFS were 0.703 and 0.664 in the training and validation cohorts, respectively, which were higher than those of the American Joint Committee on Cancer (AJCC) 7th edition TNM stage, AJCC 8th edition TNM stage, and the Child-Pugh score.ConclusionOur results revealed the IFS was an independent risk factor for OS and RFS in patients with ICC after hepatectomy and could serve as an effective prognostic prediction system in daily clinical practice.

Practice of the new supervised machine learning predictive analytics for glioma patient survival after tumor resection: Experiences in a high-volume Chinese center.

This study aims to assess the effectiveness of the Gradient Boosting (GB) algorithm on glioma prognosis prediction and to explore new predictive models for glioma patient survival after tumor resection. A cohort of 776 glioma cases (WHO grades II-IV) between 2010 and 2017 was obtained. Clinical characteristics and biomarker information were reviewed. Subsequently, we constructed the conventional Cox survival model and three different supervised machine learning models, including support vector machine (SVM), random survival forest (RSF), Tree GB, and Component GB. Then, the model performance was compared with each other. At last, we also assessed the feature importance of models. The concordance indexes of the conventional survival model, SVM, RSF, Tree GB, and Component GB were 0.755, 0.787, 0.830, 0.837, and 0.840, respectively. All areas under the cumulative receiver operating characteristic curve of both GB models were above 0.800 at different survival times. Their calibration curves showed good calibration of survival prediction. Meanwhile, the analysis of feature importance revealed Karnofsky performance status, age, tumor subtype, extent of resection, and so on as crucial predictive factors. Gradient Boosting models performed better in predicting glioma patient survival after tumor resection than other models.

Quantifying the Historical Development of Abugadaf Natural Forest Using GIS-Remote Sensing Analytical Techniques/ Blue Nile State/Sudan

The main objective of this research is to provide current information about the growth and general status of Abugadaf natural forest reserve in Blue Nile State, Sudan, and to identify the possible agents and factors that contributed to the current forest status whether positive or negative. The secondary objective of the study is assess and demonstrate the capabilities of RS and GIS analytical tools in addressing the issue of monitoring and assessing natural forest areas in Sudan. Four satellite images covering the period1990, 2000, 2010 and 2018 were selected from cloud-free Land-Sat (TM, ETM and Land-Sat 8) covering the study area. All images were false color composite (FCC). Images 1990 and 2000 were thematic mapper (TM), while Image 2010 was enhanced Thematic Mapper (ETM), and image 2018 was Land-sat 8. Land-sat 8 was used as a base to assess the change detection, and GPS data used to extracts topographic information in the study areaThe tentative locations of the Land cover and land use sites were selected using image analysis aided by actual ground surveying. GPS (Garmin 62C) was used to navigate among check samples and to record the coordinates (X, Y and Z values) of all samples. The total number of samples covering the study area were 144 (800 x 400 m). The recorded data in each sample include tree species, regeneration, grass type, tree stumps, dead trees, density of survival trees, diameter at breast height, and total tree height. All training sites were utilized in the generation of signatures required for supervised maximum likelihood classification. Supervised maximum likelihood classifier used the Gaussian threshold stored in each class signature to determine if a given pixel fell within class or not. The study with the help of remote sensing facilities has succeeded to highlight the condition and development of the study area during the period 1990 – 2018. The study concluded that although there were some fluctuations in the amount of annual rainfall in the area, there are no enough evidences to support their direct effect on the growth of vegetation cover. The study also indicated that the availability of other vegetation cover, such as grasses, in the forest can reduce the pressure on mature forest tress and seedlings caused by domestic animal herders. In the last decade of the study period, the decrease of forest cover is mostly associated with decrease or absence of vegetation cover as in that case forest trees and young seedlings become the main source of animal feeding. If other forest destruction agents are added due to the absence of management control, the present forest will disappear within the coming five years.

Discussion.

Discussion.

Recipient Survival after Orthotopic Liver Transplantation: Interpretable Machine Learning Survival Tree Algorithm for Patient-Specific Outcomes.

Elucidating contributors affecting liver transplant survival is paramount. Current methods offer crude global group outcomes. To refine patient-specific mortality probability estimation and to determine covariate interaction using recipient and donor data, we generated a survival tree algorithm, Recipient Survival After Orthotopic Liver Transplantation (ReSOLT), using United Network Organ Sharing (UNOS) transplant data. The UNOS database was queried for liver transplants in patients ≥18 years old between 2000 and 2021. Preoperative factors were evaluated with stepwise logistic regression; 43 significant factors were used in survival tree modeling. Graft survival of <7 days was excluded. The data were split into training and testing sets and further validated with 10-fold cross-validation. Survival tree pruning and model selection was achieved based on Akaike information criterion and log-likelihood values. Log-rank pairwise comparisons between subgroups and estimated survival probabilities were calculated. A total of 122,134 liver transplant patients were included for modeling. Multivariable logistic regression (area under the curve = 0.742, F1 = 0.822) and survival tree modeling returned 8 significant recipient survival factors: recipient age, donor age, recipient primary payment, recipient hepatitis C status, recipient diabetes, recipient functional status at registration and at transplantation, and deceased donor pulmonary infection. Twenty subgroups consisting of combinations of these factors were identified with distinct Kaplan-Meier survival curves (p < 0.001 among all by log rank test) with 5- and 10-year survival probabilities. Survival trees are a flexible and effective approach to understand the effects and interactions of covariates on survival. Individualized survival probability following liver transplant is possible with ReSOLT, allowing for more coherent patient and family counseling and prediction of patient outcome using both recipient and donor factors.

Evaluation of race/ethnicity-specific survival machine learning models for Hispanic and Black patients with breast cancer

ObjectivesSurvival machine learning (ML) has been suggested as a useful approach for forecasting future events, but a growing concern exists that ML models have the potential to cause racial disparities...

Development of an Electronic Frailty Index for Predicting Mortality in Patients Undergoing Transcatheter Aortic Valve Replacement using Machine Learning

Background: Electronic frailty indices can be useful surrogate measures of frailty. Objective: This study is to develop an electronic frailty index that incorporates patient demographics, baseline comorbidities, health-care utilization characteristics, electrocardiographic measurements, and laboratory examinations for predicting all-cause mortality in patients undergoing transcatheter aortic valve replacement (TAVR). Methods: This was a multicenter retrospective observational study of patients undergoing for TAVR. Significant univariate and multivariate predictors of all-cause mortality were identified using Cox regression. Importance ranking of variables was obtained with a gradient boosting survival tree (GBST) model, a supervised sequential ensemble learning algorithm, and used to build the frailty models. Comparisons were made between multivariate Cox, GBST, and random survival forest models. Results: A total of 450 patients (49% of females; median age at procedure, 82.3 [interquartile range, 79.0–86.0]) were included, of which 22 died during follow-up. A machine learning survival analysis model found that the most important predictors of mortality were activated partial thromboplastin time, followed by INR, severity of tricuspid regurgitation, cumulative hospital stays, cumulative number of readmissions, creatinine, urate, alkaline phosphatase, and QTc/QT intervals. GBST significantly outperformed random survival forests and multivariate Cox regression (precision: 0.91, recall: 0.89, AUC: 0.93, C-index: 0.96, and KS-index: 0.50) for mortality prediction. Conclusions: An electronic frailty index incorporating multidomain data can efficiently predict all-cause mortality in patients undergoing TAVR. A machine learning survival learning model significantly improves the risk prediction performance of the frailty models.

Clinical-epidemiological characteristics and survival of cases of severe acute respiratory syndrome (SARS) due to COVID-19, according to the COVID-19 vaccination schedule in Brazil, 2021-2022: a prospective study.

To analyze the clinical and sociodemographic characteristics and survival of individuals with severe acute respiratory syndrome due to COVID-19 according to the COVID-19 vaccination schedule, Brazil, 2021-2022. This was a cohort study based on data from the Influenza Epidemiological Surveillance Information System; the Kaplan-Meier and Survival Tree methods were used to analyze survival. Among the 559,866 hospitalized cases, a higher proportion of vaccinated individuals was found among female (15.0%), elderly people aged ≥ 80 (34.5%), people from the Southeast region (15.7%), those who did not undergo respiratory support (21.2%) and those who did progress to death (15.2%); the survival curve showed that risk of death for unvaccinated individuals was higher in all age groups (p-value < 0.001); elderly people aged ≥ 80, who did not undergo mechanical ventilation and who had a booster dose had lower risk when compared to their peers who had two doses or were unvaccinated (hazard ratio = 0.64; 95%CI 0.62;0.67). Lowest risk of death was found in vaccinated individuals, especially those who had two doses or a booster dose as well. Prevalence was found to be high among unvaccinated individuals. Risk of death was lower among those vaccinated with a booster dose, compared to those not vaccinated, in all age groups analyzed. The number of hospitalizations of unvaccinated individuals with severe acute respiratory syndrome was high, which increases the demand for health services to care for these individuals. It is necessary to promote widespread vaccination of the entire population of Brazil, in addition to the regular provision of booster doses for the different population groups.

Prognostic stratification of oropharyngeal cancer patients in a betel nut chewing and low HPV area

BackgroundThis study aimed to establish a simple predictive model for oropharyngeal cancer (OPC) in an area with a relatively low prevalence of human papillomavirus (HPV) and frequent betel nut chewing.MethodsA total of 116 patients with OPC were recruited from the clinical research database of a referral cancer center between 2013 and 2018. Patient characteristics—including age, gender, tumor stage, differentiation, and treatment modality—were extracted from the database. Patients diagnosed after 2018 were staged using the 7th AJCC staging system to explore the impact of extra-nodal tumor extension (ENE) on survival. Immunohistochemical analysis was performed for p16, epidermal growth factor receptor (EGFR), p53, and programmed cell death ligand 1 (PD-L1). ENE status was evaluated by pathological analysis or radiological features. Primary outcome was disease-specific overall survival (OS). Univariate and multivariate Cox regression were used to establish a predictive model.ResultsMean age was 57.3 ± 9.9 years; 103 patients (88.8%) were male. P16 positive OPC was positively associated with higher PD-L1 and a tonsillar sub-site and negatively associated with betel nut chewing and cigarette smoking. In Cox regression, age, p16 status, EGFR, cT4, ENE, and cigarette smoking were significantly associated with OS. In survival tree analysis, cT stage was the most important risk stratification parameter, followed by EGFR expression and p16 status. Patients with cT4 stage or high EGFR were classified as the high-risk group and had poorest OS.ConclusionsDue to the low prevalence of HPV and popularity of betel nut chewing in Asia, the relative importance of prognostic predictors for OPC are not identical to Western countries. Identification of significant prognostic biomarkers may improve treatment.Trial registration This study was registered and approved by the Institutional Review Board (IRB) of Kaohsiung Veterans General Hospital (VGHKS19-CT9-07; date of approval: Aug 9, 2019).Graphic

A Fuzzy Random Survival Forest for Predicting Lapses in Insurance Portfolios Containing Imprecise Data

We propose a fuzzy random survival forest (FRSF) to model lapse rates in a life insurance portfolio containing imprecise or incomplete data such as missing, outlier, or noisy values. Following the random forest methodology, the FRSF is proposed as a new machine learning technique for solving time-to-event data using an ensemble of multiple fuzzy survival trees. In the learning process, the combination of methods such as the c-index, fuzzy sets theory, and the ensemble of multiple trees enable the automatic handling of imprecise data. We analyse the results of several experiments and test them statistically; they show the FRSF’s robustness, verifying that its generalisation capacity is not reduced when modelling imprecise data. Furthermore, the results obtained using a real portfolio of a life insurance company demonstrate that the FRSF has a better performance in comparison with other state-of-the-art algorithms such as the traditional Cox model and other tree-based machine learning techniques such as the random survival forest.

Interpretable prognostic modeling of endometrial cancer

Endometrial carcinoma (EC) is one of the most common gynecological cancers in the world. In this work we apply Cox proportional hazards (CPH) and optimal survival tree (OST) algorithms to the retrospective prognostic modeling of disease-specific survival in 842 EC patients. We demonstrate that linear CPH models are preferred for the EC risk assessment based on clinical features alone, while interpretable, non-linear OST models are favored when patient profiles can be supplemented with additional biomarker data. We show how visually interpretable tree models can help generate and explore novel research hypotheses by studying the OST decision path structure, in which L1 cell adhesion molecule expression and estrogen receptor status are correctly indicated as important risk factors in the p53 abnormal EC subgroup. To aid further clinical adoption of advanced machine learning techniques, we stress the importance of quantifying model discrimination and calibration performance in the development of explainable clinical prediction models.