I have a 3 arm clinical study with two drugs and placebo on the three arms respectively. I intend to perform a Log-rank test on it.
I would like to know about the best software which is globally recognized, that can help me calculate the sample size.
Also, would anyone explain how do I get the effect size ?
Thank you.
Best Answer
There is a lot of reliable software for this task that is well established and widely recognized (e.g. NQuery, EAST, some R packages, the sample size procedure in SAS and so on). The effect size used in sample size calculations (usually a hazard ratio for the log-rank-test) can be based on a number of considerations. Options include the smallest effect size that would still be clinically relevant (e.g. for a cardiovascular endpoint such as major adverse cardiovascular events that might be a hazard ratio in the 0.8 to 0.9 region, depending on whom you ask, in other disease areas the answer might be different), the smallest effect size that would be commercially viable (often based on what competitor products have demonstrated), the smallest effect size that would change clinical practice, or perhaps a discounted version of what you would realistically expect from the drug in question (e.g. based on some initial data, which might of course be an overestimate - if it's based on a small study - or simply come with a certain amount of uncertainty, so some suitable discounting may be needed). Usually, the non-statistical team members of a team planning a study will have an idea on which of these options might be the most relevant.