Replies to post #28113 on Cel Sci Corporation (CVM)

[Fosco1]

@lightrock

From CT :
"Primary Outcome Measures Overall Survival (OS) in LI + CIZ + SOC vs. SOC [ Time Frame: 3 year ]
OS will be assessed using Kaplan-Meier life-table and compared using a logrank test ..."

This simply means they will assess differential in the two curves through logrank test. The logrank test will say whether the whole curve (not just at a given time) are significantly different under the protocol hypothesis (Effect sized derived from their 10% superiority assumption). The "time frame" is just an estimated time frame. Log-rank test is meant for comparing overall curves.

From this document

"Comparing two Survival Curves: the Log-rank test

There are many circumstances when it is required to ascertain whether or not there are differences in the survival experiences of two groups. A naive approach would be to choose a time, e.g. 3 years, work out the proportions surviving beyond this time in the two groups and compare them.
This has two disadvantages:
- First it is difficult to define the proportions in a way that is both efficient and unbiased, largely due to the presence of censoring
- Second the choice of a time (such as 3 years) is arbitrary and it is potentially misleading to compare two survival curves at a single time

The log-rank test overcomes the problems of censoring and overcomes the problem of comparison at a single point by comparing at many time points simultaneously."

Now, how will they assess the 10% "superiority" mentioned many times by Geert but not at all in this CT site ? There is no way to calculate an "average survival superiority", only at one point in time (such as median survival). So it is not really a survival superiority that is measured, it is rather the "less death ratio" the hazard ratio with gives the effect size of the test group and is supposed to be constant at each point in time under the KM maths hypothesis.
- By sizing the sample (298 events) using the required effect size, if p<0.05 this will mean that logrank test has detected at least such overall effect size.
- They will calculate the hazard ratio which is "the commonest estimate of the difference between two survival curves", and provide a confidence interval : the more sample size is high, the more the measure of the harzard ratio will be narrowed to its true value and the effect size convincing for the FDA to approve