There are drug adverse events, and then there are the Voldemorts of drug adverse effects, whose which-must-not-be-named. Bleeding in your stomach lining is one. Basically, you can lose half of your blood volume before the blood appears at the “exit point” alerting you and others to the problem (other than paleness, dizziness, fainting and shock that is). If you happen to be more than a few yards from a hospital at that point, well, good luck.
More insidious is an adverse event not uncommon in the history of small molecule drugs called "agranulocytosis." In this case, your white blood cell count plummets to zero and you don’t know it at all. You don’t know that your body is now open for business for every virus, bacterium, fungus and parasite in the vicinity to which you are normally immune. By the time you find out what that pesky fever is from, it's too late. You basically suddenly got AIDS or chemotherapy and didn’t know it.
That’s what happened to Acorda's tozadenant. Five people died suddenly of infections in their Phase III trial of this Parkinson’s drug and, lo and behold, they had no white blood cell counts. Tozadenant is an adenosine A2a antagonist and its not clear why agranulocytosis would result from its use. Initially, the company considered playing with the dose to mitigate the effect, but failures in Phase III are past the point of no return, especially for unpredictable effects like agranulocytosis.
It would have been a different story if this terrible side effect were KNOWN prior to the start of clinical studies. Indeed, the widely used schizophrenia drug clozapine is burdened by agranulocytosis, but its health value overcomes the risk in most cases because it is the devil you know. If Acorda had a technology that could have alerted them to problems in hematopoietic tissue early in drug development, they could have either saved some money or advanced the trials with a heavy watchful eye for this side effect, an approach with which the FDA may have been lenient. Now if only they had access to a technology like that . . .