The Biostatistics and Epidemiology Unit develops new tools for research and analysis. We attempt to diffuse these original methodological tools which are centered around the modeling of patients with a chronic pathology. More precisely, the tools which we propose are as follows:
This is a collection of simple R functions that were used for computing time-dependent ROC curve using Kaplan-Meier (KM) estimator or the k-nearest neighbor's (KNN) adaptation. Both approaches are developed for traditional survival analysis (all-cause analysis) and for the the additive relative survival analysis (excess of mortality). It executes tasks on a single, multiprocessor machine. Read more.
This is a collection of simple R functions that were used for computing the time-dependant ROC curve for a prognostic marker from aggregated data (survival probabilities in strata of the marker) and from several studies. Read more.
The separation between two survival curves represents the magnitude of the association between the intervention and the time-to-event. A statistical test can determine the statistical significance of the difference but does not quantify its magnitude. The purpose of the prognostic ROC curve is to represent this uncertainty: the AUC is the probability that the time-to-event is improved in one arm compared to the other. This package is designed for computing such prognostic ROC curve. Read more.
Microarray data can be used to identify prognostic signatures based on time-to-event data. The analysis of microarrays is often associated with overfitting and many papers have dealt with this issue. However, little attention has been paid to incomplete time-to-event data (truncated and censored follow-up). We have adapted the 0.632+ bootstrap estimator for the evaluation of time-dependent ROC curves. The interpration of ROC-based results is well-established among the scientific and medical comunity. Moreover, the results do not depend on the incidence of the event, as opposed to many other prognostic statistics. Here, we have validated this methodology by similutions. We have illustrated its utility by analyzing a data set of diffuse large-B-cell lymphoma patients. Our results demonstrate the well-adapted properties of the 0.632+ ROC-based approach to evaluate the true prognostic capacity of a microarray-based signature. This method has been implemented in the R package ROCt632. Read more.
In epidemiology, when the clinical outcomes are time-dependent (for instance death or disease progression), survival analyses are used to assess the cumulated risk over time. A survival curve is more informative than the risk of event assessed at a single point in time. In meta-analyses of studies reporting a survival curve, the finding should be a summary survival curve. With this package, we aim to compute a summary survival curve. Read more.
This is a collection of simple R functions that were used for computing a multiplicative-regression model for relative survival. The purpose is to study the heterogeneity of risk factors between two groups of patients: a reference and a relative population. This package is designed for enable such accurate comparisons. Read more.
Medical researchers are often interested in investigating the relationship between explicative variables and times-to-events such as disease progression or death. Multistate models allows multiple times-to-event to be studied simultaneously. Time-inhomogeneous Markov models consist of modelling the probabilities of transitions according to the chronological times (times since the baseline of the study). Semi-Markov (SM) models consist of modelling the probabilities of transitions according to the times spent in states. These SM models are becoming increasingly popular to deal with the complex evolution of chronic diseases. In this package, we propose functions implementing usual 3-state and 4-state multistate models (SM models and time-inhomogeneous Markov models). We also propose to take into account the mortality of the general population (relative survival approach). Read more.
In observational studies, the presence of confounding factors is common and the comparison of different groups of subjects requires adjustment. In the presence of survival data, this adjustment can be achieved with a multivariate model. A recent alternative solution is the use of adjusted survival curves and log-rank test based on inverse probability weighting (IPW). By using the approach proposed by Xie and Liu (2005), we illustrate the usefulness of such methodology by studying the patient and graft survival of kidney transplant recipients according to the expanded donor criteria (ECD). Read more.