2

Runnning and interpreting a survival analysis

  • updated 1 mth ago

The survival analysis module allows to study any quantity of interest defined as time-to-event variable. An example of such a variable can be the time at which a clinical progression happens, namely Time To Progression (TTP). The graphical representation of such variables  is the survival curve, usually estimated by the Kaplan-Meier method. 

Analysis setup

To perform a survival analysis on simulation outputs, you first need a Trial Visualization. The fourth tab corresponds to the Survival Analysis.

To visualize your results, you can:

  • Select outputs: select your time-to-event outputs of interest.

  • Select arms: by default, all trial arms are selected. You have the ability to select specific arms in the “Arms” selector card.

  • Filter results: you have the possibility using the Filtering options on the left-side panel to apply filters on baseline descriptors, patients descriptors and measures.

  • Group results: by default, all results will be grouped by arm. You can also apply various groupings, using distinct values, span, buckets, and quantiles.

  • Restrict to an observation window: the analysis can be restricted to a shorter period than the trial duration by changing the “Period end” field.

 

Analysis Methodology

In the following example, we are using a Lung cancer model and a virtual population of 50 patients. The in silico trial was run to simulate 3 years of follow-up.

There are two protocol arms in the simulation:

  • Treatment-non-mutated: one dose of treatment, and K-Ras protein is not mutated

  • Treatment-mutated: one dose of treatment, and K-Ras protein is mutated

The K-Ras protein is a protein that triggers resistance to the treatment.
The time to the event of interest is the time from the start of the study to the start of tumor growth, i.e. the Time To Clinical Progression (TTCP).

The questions we might want to answer with this simulation using the survival analysis are:

  • what is the impact of the mutational status of KRAS on TTCP?
  • Is the impact of the mutational status statistically significant? 

Plots interpretation

Three plots are displayed:

  • Cumulative survival rates: evolution of the survival rates and confidence intervals for each group of patients estimated using the Kaplan-Meier estimator. 

  • Numbers of patients at risk: evolution of the number of patients for whom the event of interest did not occur. 

  • Cumulative hazard rates: evolution of the cumulative hazard rate estimated using the Nelson-Aalen estimator.

Looking at the cumulative survival rate plot, and the summary table of medians, we can conclude that the mutational status does have an impact on the time to clinical progression. Indeed,  the yellow curve falls faster than the blue one, indicating that patients without the mutation respond better to the treatment than the others (respectively, 19 days [CI 95% 15 - 30] vs. 273 days [CI 95% 175 - 301]).

Statistics table interpretation

Two statistics tables are displayed:

  • Survival rate statistics: computes for each arm the first, second and third survival quartiles with confidence intervals

  • Log rank test : Computes the LogRank statistics between two groups. This statistical test is used to compare two survival curves and assess if there is a statistically significant difference between them. The test is based on the computation of the difference between the number of expected and observed events for each group.

 

Looking at the log rank test table, one can see that the survival curves for arm “Treatment-non-mutated” and arm “Treatment-mutated” are statistically different (p-value < 0.05).

Reply Oldest first
  • Oldest first
  • Newest first
  • Active threads
  • Popular