100 people is a large sample? You're joking right? 100 people is nowhere near enough for an accurate result.
As I stated above, 100 samples is very large for narrow distributions. You have suggested (and I agree) that the cancer distribution is probably very broad and requires a larger sample.

No point doing this kind of study? Definitely not.
Imagine the situation. A doctor informs his cancer patient that they are currently undertaking a study to see if prayer is effective in cancer treatment. To gain an unbiased result has to be taken off ALL scientific remedies (chemotherapy, drugs, surgery etc) and rely only on prayer. How many do you think would take up that study? None, because they know that modern science is the best chance they have of survival.
OK, let's not choose cancer but another disease for which there is no treatment. Also don't inform the patients of the study.