Linked In Icon Twitter Icon Facebook Icon

"interesting to hear views on the subject"

Organon

PhUSE Wiki

Calculating Events Needed For Event Driven Trials 

Wiki content first added by: Mat Soukup

  

Tool Name 

Section content first added by: Mat Soukup

EventsNeeded

Tool Description 

Section content first added by: Mat Soukup

Provides sample size calculations for the number of events needed to adequately power an event driven trial.

Engine Type (intended to run the script) 

Section content first added by: Mat Soukup

R statistical computing environment

Engine Version 

Section content first added by: Mat Soukup

Written using R version 2.10.1

Operating Environment 

Section content first added by: Mat Soukup

To date, this function has only been evaluated in Microsoft Windows.

Ontology/Categorization and Keywords (tags for searching) 

Section content first added by: Mat Soukup

sample size calculation, time to event, survival analysis

Author Name and Email 

Section content first added by: Mat Soukup

Written by Xiao Ding and adapted by Mat Soukup (Mat.Soukup@fda.hhs.gov)

Create Date 

Section content first added by: Mat Soukup

October 2011

Last Modified Date 

Section content first added by: Mat Soukup

No modifications to date

Required Arguments 

Section content first added by: Mat Soukup

  • power: a scalar between 0 and 1 corresponding to the power of the study (default is 0.90)
  • error: a scalar between 0 and 1 corresponding to the two-sided Type I error rate which the hypothesis will be tested (default is 0.05)
  • HR: a scalar of the assumed true hazard ratio of the two treatment groups
  • alt: a scalar or vector of the value of the alternative hypothesis defining the degree of risk (or benefit) to rule out

Optional Arguments 

Section content first added by: Mat Soukup

None

Output / Value Returned 

Section content first added by: Mat Soukup

Returns a list containing the number of events needed (NEN) for the specified arguments as well as the value for the maximum value the point estimate can take and still meet the alternative hypothesis margin (PE.threshold). The value of the alternative hypothesis (margin) is also returned.

Error Return Codes 

Section content first added by: Mat Soukup

Errors are returned if the power or error arguments are not between 0 and 1.

Example Data Source 

Section content first added by: Mat Soukup

No data source is required to use the function

Protocol / Trial Source 

Section content first added by: Mat Soukup

This was developed to provide a simple way of calculating the number of events needed for powering an event driven safety trial designed to rule out some degree of risk.

Basic Usage Example (Call) 

Section content first added by: Mat Soukup

EventsNeeded(power=0.9, error=0.05, HR, alt)

Sample Inputs 

Section content first added by: Mat Soukup

#Call 1 - Power of a 90% assuming a background HR=1 to rule out a two-fold risk
EventsNeeded(power=0.9, error=0.05, HR=1, alt=2.0)

#Call 2 - Power of 90% assuming a background HR = 0.75 to show superiority
EventsNeeded(power=0.9, error=0.05, HR=0.75, alt=1.0)

#Call 3 - Power of 90% assuming a background HR = 1 to rule out several levels of risk
risk <- seq(1.25, 2.0, .25)
EventsNeeded(power=0.9, error=0.05, HR=1, alt=risk)

Sample Results 

Section content first added by: Mat Soukup

#Call 1
$NEN
[1] 87.4793
$margin
[1] 2
$PE.threshold
[1] 1.315267

#Call 2
$NEN
[1] 507.8443
$margin
[1] 1
$PE.threshold
[1] 0.8403428

#Call 3
$NEN
[1] 844.0876 255.6520 134.2073 87.4793
$margin
[1] 1.25 1.50 1.75 2.00
$PE.threshold
[1] 1.092230 1.173866 1.247632 1.315267

References (e.g. scientific papers) 

Section content first added by: Mat Soukup

None

See Also 

Section content first added by: Mat Soukup

None

Contibute to this Wiki

Register You need to register to contribute to this Wiki.

Login If you are already regsitered please login.