CollapseAll image

Generating a Survival Analysis

Dataset Explorer allows you to generate a time-to-event analysis – that is, a relationship between one or more predictive variables and the survival times of individuals in two groups of cohorts.

You may run multiple visualizations in the background while continuing to use Dataset Explorer. For more information on running visualizations in the background, see Asynchronous Operations.

To generate a survival analysis, you must introduce the following datasets into each group:

■    The observed survival times of the individuals in each group.

■    At least one variable (such as different medications or different genetic attributes) that distinguishes the two groups.

■    The specific event (death) being tracked for the individuals in the study, and optionally, any censoring factors that occurred before the event took place. 

A censoring factor might be the withdrawal of an individual from the study, or the conclusion of the study before the event occurred for a given individual.

Note:  In Dataset Explorer survival analysis datasets, the event is always death.  Subjects who are censored are not included in the death count.

Dataset Explorer’s survival analysis functionality is based on the Cox regression model.

To generate a survival analysis:

1.  Run tranSMART, then click the Dataset Explorer tab.

2.  Select the study of interest.

3.  Define the cohorts for the survival analysis, being sure to include the required datasets indicated above.

4.  Click Advanced > Survival Analysis.

Example

 

1.  In Dataset Explorer, open public study Wang_TransBIG_BreastCancer_GSE7390.

2.  Open the node ClinicalData.

3.  Open the following nested nodes in the following order:

a.  Clinical Data

b.  Overall Survival in Days

4.  Drag the time-to-event dataset Overall Survival in days (OS) into Subset 1.

Don’t drag the folder into the subset box – just the dataset inside the folder.

5.  In the Set Value dialog, select No Value, then click OK:

Specifying a value limits the values in the dataset.  By not specifying a value, the entire time-to-event dataset is used.

6.  Repeat step 4 and step 5 for Subset 2.

7.  In the same Clinical Data node, drag the Censored (OSCENS) folder into empty boxes in Subset 1 and Subset 2.

The contents of the Censored (OSCENS) folder are No and Yes.  This concept introduces the Event and Censored datasets into the analysis.

8.  Open the following nested nodes in the following order:

a.  Subjects

b.  Medical History

c.  Estrogen Receptor Status

9.  Drag Estrogen Receptor Negative into an empty box in Subset 1.

10. Drag Estrogen Receptor Positive into an empty box in Subset 2.

The subset boxes are now defined as follows:

11. Click the Advanced tab, then click Survival Analysis.

The analysis includes the following Kaplan-Meier curves:

In the figure, the x-axis represents survival time in days, and the y-axis represents the percentage of subjects who were still alive at a given point in time during the study.

Note the hashmarks in the plot lines.  These represent censored data – for example, subjects who dropped out of the study before the event (death) occurred.


Related Topics

Hazard Ratio and Relative Risk