Bayesian variable selection and survival modeling: Assessing the most important comorbidities that impact lung and colorectal cancer survival in Spain

Cancer survival represents one of the main indicators of interest in cancer epidemiology. However, the survival of cancer patients can be affected by a number of factors, such as comorbidities, that may interact with the cancer tumour. Moreover, it is of interest to understand how different cancer sites and tumour stages are affected by different comorbidities. Identifying the comorbidities that affect cancer survival is thus of interest as it can be used to identify factors driving the survival of cancer patients. This information can also be used to identify vulnerable groups of patients with comorbidities that may lead to a worst prognosis of cancer. We address these questions and propose a principled selection and evaluation of the effect of comorbidities in the overall survival in cancer patients. In the first step, we apply a Bayesian variable selection method that can be used to identify the comorbidities that predict overall survival. In the second step, we build a general Bayesian survival model that accounts for time-varying effects. In the third step, we derive several posterior predictive measures to quantify the effect of individual comorbidities on the population overall survival. We present applications to data on lung and colorectal cancers from two Spanish population-based cancer registries. The proposed methodology is implemented with a combination of the R-packages mombf and rstan. We provide the code for reproducibility at https://github.com/migariane/BayesVarImpComorbiCancer.

Assessment of Deterioration of Highway Pavement using Bayesian Survival Model

The size and level of complexity of highway pavement data and its associated covariates have led to the application of different approaches in the analysis of the highway pavement data for deterioration modeling. With the goal of predicting the survival of highway pavement with interpretable and reproducible models that are robust to uncertainties, errors, and overfitting, the Bayesian survival model (BSM) is proposed in this paper as a good method of estimating parameters for survival functions. Deterioration patterns in relation to the failure time distribution are treated as random quantities sampled from some stochastic prior processes. The specified priors are combined with the data sampled to obtain the distribution of the survival function using Bayes theorem and the Markov chain Monte Carlo method. A posteriori distribution of the survival function is obtained from the pavement information and compared with the classical product limit survival (Kaplan-Meier) estimate and the univariate parametric survival function. This paper reports experimental results of the three candidate models and their efficiency in describing the survival of highway pavement in the presence of deterioration. It is observed from the BSM outcomes that the posterior estimates are accurate in estimating the survival times of roadway segments at 95% credible interval. The outputs also show the robustness of the BSM in describing the uncertainties associated with the survival of highway pavement compared with the Kaplan-Meier and the univariate parametric survival models in the event of limited data and misspecification of underlying distribution.

Development and validation of a Bayesian survival model for inclusion body myositis

Background Associations between disease characteristics and payer-relevant outcomes can be difficult to establish for rare and progressive chronic diseases with sparse available data. We developed an exploratory bridging model to predict premature mortality from disease characteristics, and using inclusion body myositis (IBM) as a representative case study. Methods Candidate variables that may be potentially associated with premature mortality were identified by disease experts and from the IBM literature. Interdependency between candidate variables in IBM patients were assessed using existing patient-level data. A Bayesian survival model for the IBM population was developed with identified variables as predictors for premature mortality in the model. For model selection and external validation, model predictions were compared to published mortality data in IBM patient cohorts. After validation, the final model was used to simulate the increased risk of premature death in IBM patients. Baseline survival was based on age- and gender-specific survival curves for the general population in Western countries as reported by the World Health Organisation. Results Presence of dysphagia, aspiration pneumonia, falls, being wheelchair-bound and 6-min walking distance (6MWD in meters) were identified as candidate variables to be used as predictors for premature mortality based on inputs received from disease experts and literature. There was limited correlation between these functional performance measures, which were therefore treated as independent variables in the model. Based on the Bayesian survival model, among all candidate variables, presence of dysphagia and decrease in 6MWD [m] were associated with poorer survival with contributing hazard ratios (HR) 1.61 (95% credible interval [CrI]: 0.84–3.50) and 2.48 (95% CrI: 1.27–5.00) respectively. Excess mortality simulated in an IBM cohort vs. an age- and gender matched general-population cohort was 4.03 (95% prediction interval 1.37–10.61). Conclusions For IBM patients, results suggest an increased risk of premature death compared with the general population of the same age and gender. In the absence of hard data, bridging modelling generated survival predictions by combining relevant information. The methodological principle would be applicable to the analysis of associations between disease characteristics and payer-relevant outcomes in progressive chronic and rare diseases. Studies with lifetime follow-up would be needed to confirm the modelling results.