The less variables factored in the less Faith one should have in the informations validity. In the case of treatment A or B in the example, if this is all the information you have to go on I'd probably tossing a coin.

That data is complicated or difficult to evaluate is not a good reason to ignore it.