I agree with Brian Oblivion on the importance of this distinction. However the issue is not entirely medical. It really depends on what you are after:
1) Suppose I am a psychiatrist trying to legalize MDMA for therapeutic use. I want to know what the chances of a fatality are when I give the medicine to the patient during a session. Obviously here the statistics of the number of deaths caused by heat stroke is not going to be a factor, as one can expect to be able to control and prevent this in a clinical environment.
2) Suppose I am an average Joe and I have a teenager kid named Bobby. Bobby has just made some dodgy looking friends and I am concerned after overhearing their plan to drop some E and dance the weekend away. To me as a parent, the most relevant statistic is the one that includes both MDMA neurotoxicity deaths and MDMA related deaths such as hyponatremia, heat stroke, and crashing your car into a brick wall due to impaired judgment.
3) Suppose I am a mature responsible adult user of E. I carefully watch the dosage, use pure E crystals only, and always have a reliable sitter for the group, who can watch the temperature, make sure everybody drinks water and sports-drinks, nobody gets into a car while still under influence, etc. Then I would estimate that the risk is somewhere in between (1) and (2).
Would be quite interesting to see accurate estimate for (1) and for (2), especially with links showing where the original numbers come from.
TH