A 2008 report from the United Kingdom's Home Office Advisory Council on the Misuse of Drugs concluded that ecstasy (at least, MDMA) is not nearly as dangerous as was previously thought, either in deadliness or in long-term health consequences. The Council even recommended changing the classification of MDMA from its present status as Class A (heroin, crack, and amphetamines prepared for injection are Class A) to the less-dangerous Class B (which includes marijuana and Ritalin). (The recommendation was, of course, rejected.)
A February 2009 editorial in the New Scientist took the logic a step further:
Imagine you are seated at a table with two bowls in front of you. One contains peanuts, the other tablets of the illegal recreational drug MDMA (ecstasy). A stranger joins you, and you have to decide whether to give them a peanut or a pill. Which is safest?
You should give them ecstasy, of course. A much larger percentage of people suffer a fatal acute reaction to peanuts than to MDMA.
The implication is that, when acting upon a stranger, we should minimize his risk of death.
The lovely and talented Caledonian has a slightly different take: we should focus on the relative likelihood of harm, he says, rather than the relative likelihood of death.
Both of these goals - acting to minimize the risk of death to a stranger, and acting to minimize his risk of harm - are laudable and widely shared. But there's a glaring aspect of the utilitarian calculus that almost no one seriously considers in making the decision to administer a peanut or some ecstasy. This is the differential positive utility to be gained by the stranger in each case. A peanut is marginally sustaining, but unless it's been boiled with star anise and Sichuan peppercorns, it's not particularly enjoyable. Ecstasy, on the other hand, is fucking awesome. Why doesn't anybody consider the relative benefit to the stranger along with the relative harm?
While many of us would certainly consider the pleasure of ecstasy in deciding whether to eat the pill or the peanut ourselves, it's proper and coherent not to consider the pleasurable effects of a potentially harmful action when it will be inflicted upon a non-consenting stranger whose values we do not know. This illustrates Seana Shiffrin's principal that, while it's morally acceptable to harm a stranger without his consent in order to prevent worse harm (e.g., to administer ecstasy in order to avoid administering a peanut or to break someone's arm in order to pull him from a burning car), it's not morally acceptable to harm a stranger without his consent in order to provide a pure benefit. But the ecstasy example supports a stronger inference: when evaluating actions that will harm a non-consenting stranger, his potential pleasure doesn't count. When we're acting toward someone whose values we do not know, we should not think in terms of maximizing his utility, but in terms of minimizing our harm to him.
The distinction between acting toward a non-consenting stranger whose values we do not know, and acting toward ourselves (or toward someone whose values we know), is one that is ignored by S. D. Baum in his article "Better to exist: a reply to Benatar" (J. Med. Ethics 2008;34;875-876). Baum's "reply" (to David Benatar's position that it is always better not to bring people into existence) is, in relevant part, as follows:
The benefits/harms asymmetry is commonly manifested (including in Benatar’s writing) in the claim that no amount of benefit, however large, can make up for any amount of harm, however small. This claim comes from an intuition that while we have a duty to reduce harm, we have no duty to increase benefit. The corresponding ethical framework is often called "negative utilitarianism". Negative utilitarianism resembles maximin in its resolute focus on the worst off—as long as some of those worst off are in a state of harm, instead of just in a state of low benefit. Like maximin, negative utilitarianism can recommend that no one be brought into existence—and that all existing people be euthanised. I find negative utilitarianism decidedly unreasonable: our willingness to accept some harm in order to enjoy the benefits of another day seems praiseworthy, not mistaken. I thus urge the rejection of this manifestation of the benefits/harms asymmetry. [Emphasis mine; citations omitted.]
Our own willingness to accept suffering in the interest of pleasure (or any other value) is no reason to think that it is right to inflict that same suffering on a non-consenting stranger. Negative utilitarianism may not be the proper course to take in our own lives, but thought experiments like mine suggest that negative utilitarianism is the proper course to take toward the lives of others who do not consent to our interference. 
Many people think it's morally acceptable to have babies, despite the fact that the babies will certainly suffer a great deal during their lifetimes and may suffer an exceptional amount (that is, bringing someone into existence does him some harm). Pronatalists generally want to point out the good things in life - the pleasant effects of puppies and sunsets - and to balance them against life's harms. But bringing a child into the world necessarily entails harming a stranger (for one doesn't know the values of one's child prior to procreation). It is no different from dosing a stranger with ecstasy for no reason, except that the harms of life massively exceed the harms of ecstasy, and the pleasure of life, for many, is much less. Considering the non-consenting stranger's pleasure in the ecstasy/peanut case is unthinkable; procreation advocates need to explain why considering his pleasure in coming into existence is just fine.
The peanut/ecstasy example functions as a thought experiment that may be closer to real life than Shiffrin's ingenious example in which a wealthy person drops gold bars from an airplane, thereby benefiting some of the people below but also occasionally breaking their arms.
The only case in which it is widely accepted to inflict unconsented harm in order to provide a pure benefit is when acting toward one's children. This is an aspect of viewing one's children as property rather than persons. (Proprietariness is also the best explanation for why parents sometimes kill their natural children - and why men sometimes kill their wives or wife-equivalents - when they decide to commit suicide.)
1. Actually, the New Scientist is oversimplifying; there are two risks of death in each case. The first kind of risk is the risk that the stranger S has particular characteristics which will make any peanut, or any MDMA, lethal for him. The second kind of risk is that a particular ecstasy tablet or peanut will be lethal for any given stranger (e.g., the tablet purporting to be E is really, say, buprenophine, or the peanut is somehow infected with lethal levels of salmonella). The latter type of risk probably isn't that significant, though. UK studies don't seem to be finding lethal chemicals in street ecstasy. In Australia, the most common "fake ecstasy" is methamphetamine, which is not particularly lethal. As for peanuts, the CDC reports that the death rate from nontyphoidal Salmonella like the S. typhimurium that recently caused peanut recalls is about 00.78%.
2. I have to point out that the Mounties claim that "peanut" is a street name for ecstasy. I've never heard this in my life, but I don't go clubbing in Canada much.
3. We might also consider our own willingness to endure, on the one hand, a stranger's slight peanut breath, and on the other, a stranger clinging to our leg like a baby macaque for three hours, but that is a separate calculus.
4. Baum also assumes, contrary to Benatar's express position, that death is not a harm to already-existing people. In fact, Benatar's claims do not rest on any simplistic pleasure/pain conception of value; Benatar argues that death is a harm, even a painless death. It is, in fact, one of the great harms of life - every born person will suffer the harm of death.