"Two folds actually," L became more calm and patient than ever, "The simple one is a microscopic personal individual consideration. Let's take heart attack as the example. By the time the patient goes home, he should be taking aspirin, a beta blocker, and an ACE inhibitor. Right?"
"Yes, they have their benefits proved beyond doubt."
"Now, if we have a mega-trial and prove that, say, adding a second anti-platelet has some additional benefit, are you going to give it to your heart attack patient?"
"Of course."
"How about another trial on, say, a new aldosterone antagonist?" L smiled.
"Em ..." I began to hesitate.
"And yet another trial on- what about a direct thrombin inhibitor ?"
"In that case ..." I began to see what he was getting at.
"In that case your patient would be taking an increasing number of medicine, which means a higher risk of unexpected drug interaction, or, simply, a higher risk of non-adherence to the treatment ! Although your new gadget may achieve some marginal benefit under experimental setting, in real life it would deter the patient from taking the more important medicine - for example, aspirin in our above discussion."
"Or, to put it simply, there is a limit of how complicated one could put up a practical treatment regimen," I was forced to finish for him.