Destined to fail, page-28

Just my thoughts

But here's my thinking - if you give a treatment to a large number of people and it doesn't help some, but does help others (even to the extent of curing them) I'm not sure you can claim that the treatment was the cause of any improvement to any of them.

In my mind this is more like if you have 100 people with headaches and they all take panadol, and 90 of them are cured from the headache then this would be statistically significant although it would have to be based on your definition of what statistically significant (p value) for your hypothesis was.

And this is why in these medical trials they attempt to hold all variable constant except for the one being tested, although this is near on impossible for a medical test like this but much simpler for the one above, but that is how you determine if the results from the treatment group are caused by the treatment itself, obviously this is where your statistical significance threshold comes in. IE there is a 1 in 1000 chance that the results of this trial were caused by random chance rather than the sir spheres.

Also my thinking is along these lines where patients are treated with only chemo have a mean survival of 3 months, patients treated with both sir spheres and chemo have a mean survival of 6 months. Both sets of data would likely result in a bell curve distribution, and for arguments sake they have the same standard deviation. If you looked at these two data sets categorised as outlined you would say that there is a clear relationship between sir spheres and survival time.

Also i wanted to ask a question around statistical significance, say as per my example above except patients with sir spheres on average lived one week extra than patients without sir spheres. This would still be statistically significant, but it probably wouldnt be clinically significant?

so you could say they displayed a statistical significant increase in pfs unless of course PFS is defined as a tumor not progressing for example 6 months?

Appreciate your thoughts on this. Trying to get my head around the P value stuff been a while since i did this in uni.