Folks:
I have a short term project in clinical trials, and since I am a bit uneducated in this stuff, and since I know that this list is full of top level bio-statisticians, I dare to ask for some
help. Any ideas are greatly appreciated.
I have eight medical centers (hospitals). Let's call them sites S1,..., S8.
In each of this sites, a number of cancer-related measurement are performed (such as PSA, delays between diagnostics and treatment, mortality, etc. In each of sites, there are two groups: treatment and control. Treatment means a sort of medical support, controls do not have such a support. The groups in each site are small, so site-specific comparisons have very low statistical power. The site-specific numbers of groups, as well as the very nature of the measurements they perform are generally different. The goal of study is decide whether or not the intervention is efficient. The null hypothesis to reject is that there is no difference between treatment and control. However, unlike usual t-test we need to somehow compare the controls in ALL the sites with treatments in ALL the sites and to infer an overall summary measure of significance.
Are there standard methodologies to conduct such a test? Any ideas/references are greatly appreciated.
Thanks,
Kamil Toth
|