Simulate Patient Dropout for Single Endpoint 2 Arm Trials

Shubham Lahoti

September 26, 2024

Introduction :

For all examples, we assume the trial design consist of control and an experimental treatment. Patients may dropout of a trial for a variety of reasons such as safety issues, treatment burden or other non-trial related issues. The dropout rate can be as high as 30% in some trials if the drug has adverse side effects. The introduction of dropout probabilities or dropout hazard rate plays a significant role during data generation that can be further utilized during the analysis.

In the RCode directory of this example you will find the following R files:

1. GenerateCensoringUsingBinomialProportion.R

The R function named “GenerateCensoringUsingBinomialProportion” in the file generates the censor ID using the same dropout probability for both treatment using rbinom(). In this case, the Dropout probability is a common value across both treatments.

Steps : 1) Let pd = Dropout probability 2) Draw a random sample from Bernoulli distribution with p = 1- pd i.e. Binomial(1, pd) of a size n = NumSub 3) The sample generated in step (2) is a censoring indicator where 1 is a patient that does NOT drop out, eg completer, and 0 for a patient that drops out, eg non-completer.

2. GenerateDropoutTimeForSurvivalEp.R

The function named “GenDropoutTimeForSurvival.R” generates dropout time for 2- Arm survival design. In this cases, there is an option to provide a dropout information for each arm which then is utilized to generate dropout times for each arm from Exponential distribution. We fix Number of periods = 1 and Dropout distribution = Exponential