A new "Go/No Go" decision rule for single-arm phase II oncology trials


This page computes Testing Confidence Value (TCV) and provides suggested "Go/No Go" deicsion based on a new decision rule using both the p-value and TCV.

Please refer to the paper “Improve transitional decisions for single-arm phase II oncology trials: a new statistical decison rule” for the concept and computation formula of TCV and the associated decision rule.


Setting Confidence Levels

Confidence Level:

Please input your desired confidence level. Notice that the TCV is computed based on two-sided test. If you intend to use that for one-sided test, please modify your confidence level accordingly. For example, if you are conducting one-sided test with 95% confidence level, please input “90” into the space for “Confidence Level”.

% Area in Lower Tail:


Computing TCV

The TCV is a statistical parameter that evaluates the confidence on the hypothesis testing result. It takes into consideration the uncertainty associated with the specified null hypothesis (π0).

Please refer to the paper “Improve transitional decisions for single-arm phase II oncology trials: a new statistical decison rule” for the definition and computation formula of TCV. The confidence interval around proportion is computed using exact method.

Enter the following values, then click the Compute TCV button:

nE :

Sample size of current single-arm phase II trial. Must be greater than 1.

yE :

Number of observed success (for example, tumor response) in the current single-arm trial. Must be greater or equal to 1, and less or equal to ne.

Sample size of hypothetical data representing investigators belief on the standard treatment. Must be greater than 1. To compute n0 using confidence interval approach, please enter π0 , lq, and uq. Then, click Calculate below:

π0 :

Specified null hypothesis. For example, the expected response rate of the standard treatment. Must be greater than 0, higher than lq and lower than uq

lq : 25% percentile of π0:

lq must be lower than π0

uq : 75% percentile of π0:

uq must be higher than π0

n0 :

Equvalence effective sample size

CDF with low and upper point (please use firefox if IE8 does not display the distribution)

Mouse hovers at (0, 0).

Estimated πE :

Expected rate of treatment success for the experimental treatment

lower confidence limit for estimated πE :

TCV :

Testing Confidence Value

TCV based decision rule

The TCV is a positive value bounded by 1. It reflects the confidence on the hypothesis testing result (reject or do not reject the null hypothesis).

The new decision rule based on both p-value and TCV we proposed is as follows:

  • "Go" to a phase III trial if ALL of the following conditions are satisfied
    • πE > π0
    • p-value < α
    • TCV > L
  • "No Go" to a phase III trial otherwise

Please specify L for your trial, then click the TCV Decision button:

L:

L represents the lowest acceptable confidence on the "significant" result to proceed to a phase III trial. We suggest to use L=0.5 which achieves desirable operating characteristics in most cases.

Result :

Suggested "Go/No Go" decision

Reference: John C. Pezzullo, "Exact Binomial and Poisson Confidence Intervals." http://statpages.org/confint.html