Analyzing oropharyngeal cancer survival outcomes: a decision tree approach

Abstract

Objectives: To analyze survival outcomes in patients with oropharygeal cancer treated with primary intensity modulated radiotherapy (IMRT) using decision tree algorithms. Methods: A total of 273 patients with newly diagnosed oropharyngeal cancer were identified between March 2010 and December 2016. The data set contained nine predictor variables and a dependent variable (overall survival (OS) status). The open-source R software was used. Survival outcomes were estimated by Kaplan–Meier method. Important explanatory variables were selected using the random forest approach. A classification tree that optimally partitioned patients with different OS rates was then built. Results: The 5 year OS for the entire population was 78.1%. The top three important variables identified were HPV status, N stage and early complete response to treatment. Patients were partitioned in five groups on the basis of these explanatory variables. Conclusion: The proposed classification tree could help to guide future research in oropharyngeal cancer field. Advances in knowledge: Decision tree method seems to be an appropriate tool to partition oropharyngeal cancer patients.

Validation of NRG Oncology's prognostic nomograms for oropharyngeal cancer in the Veterans Affairs database

“…Despite these potential concerns, in this analysis, we found that NRG Oncology nomograms for predicting PFS and OS performed well on a large, population‐based data set with higher rates of comorbidities and treatment variation than the data sets used to create them. Numerous models have been developed to account for the array of potentially relevant risk factors, to predict outcomes for patients with OPSCC (and non‐OPSCC) 1,2,3,4,5,7,8,9,20 . Generally, the objectives of creating such models are: 1) to give patients, providers, and researchers reliable estimates of long‐term outcome probabilities associated with a particular individual's (or set of individuals') disease and its treatment, and 2) to define risk strata for the purposes of guiding appropriate treatment for these individuals, particularly with respect to treatment (de)intensification.…”

“…[1][2][3][4][5][6] Various risk modeling approaches have been used to determine the relative prognostic value of risk factors, including recursive partitioning analysis (RPA), random forests, machine learning, and risk scores based on multivariable regression. 1,2,5,7,8,9 The latter approach in particular lends itself to the generation of nomograms and other graphical tools that can map a large set (vector) of risk factors to a single (scalar) estimate, such as the probability of overall survival (OS) at 5 years. Such estimates may be useful in aiding provider-patient discussions and defining eligibility for clinical trials.…”

“…For example, association with human papillomavirus (HPV), T and N category, pathologic and/or radiographic extranodal extension, age, tobacco use, performance status, anemia, weight loss, body mass index (BMI), education, comorbidity, marital status, and socioeconomic status (SES) are among the known risk factors that can affect prognosis in this disease 1‐6 . Various risk modeling approaches have been used to determine the relative prognostic value of risk factors, including recursive partitioning analysis (RPA), random forests, machine learning, and risk scores based on multivariable regression 1,2,5,7,8,9 . The latter approach in particular lends itself to the generation of nomograms and other graphical tools that can map a large set (vector) of risk factors to a single (scalar) estimate, such as the probability of overall survival (OS) at 5 years.…”

Identification and validation of an immune-related lncRNAs signature to predict the overall survival of ovarian cancer

“…After then, the “rpart” Package ( ) was used to construct decision tree and split patients as different from each other as possible. It was implemented to decide which of these variables to split and the splitting value in each step of the tree’s construction ( 43 ). Moreover, a nomogram model, which is an individualized risk prediction model to predict the 1, 3, 5-year survival probability, was constructed using the “RMS” package.…”

DataSheet_1.doc

“…Overall, 12 of the included studies considered oncological outcomes following curative-intent treatment as their target of prediction. In details, six studies (40,42,(78)(79)(80)(81) aimed at predicting OS, while five (40, 82-85) considered loco-regional control (LRC) and one (86) distant metastasis-free survival (DMFS). Only two works focused on more than one oncological outcomes (40,87).…”