Weight of the Spillover

The Scientist Who Broke the Script

Mara: David Nutt is partway through a BBC interview with one of the home affairs correspondents when he hears it. He's a psychiatrist and a neuroscientist, the professor who runs Imperial College London's unit on how drugs act in the brain. And he's talking like the government officials. He's told this story himself: he suddenly thought that this was not him saying these things. So he stops the interview. No, he says, they can't go on.

Eli: The listener's question is a plain one: is alcohol really more harmful than cigarettes or weed?

Mara: Nutt had spent years on exactly that kind of question. In his words, 'I spent nine years chairing a committee that did the most systematic analysis of drug harms that has ever been done.' What came out of that work was a famous ranking. The story is what that ranking counted, what it couldn't, and whether a single score can carry the answer at all. Today's angle is the paradox of bystander damage overturning individual substance mortality rankings.

Eli: Nine years of evidence, and he caught himself reciting someone else's script.

Mara: So he changed course. He'd been made chair of the UK's Advisory Council on the Misuse of Drugs in 2008. Britain sorts drugs into legal classes, and the class sets the penalty. Nutt had started ranking drugs to test that. 'I wanted to know whether that system was evidence-based.' He began saying in public that it wasn't, and within six months, in 2009, he was out. His words: 'I was sacked for stating publicly that drug laws were not evidence-based.'

Eli: The tool he and his colleagues used next is called multicriteria decision analysis—a scorecard combining each way a drug does damage into sixteen boxes, running from death and disease to healthcare costs and violence, wider than an earlier 2007 UK expert assessment on a four-point scale, zero to three.

Mara: The paper, 'Drug Harms in the UK: A Multicriteria Decision Analysis,' ran in The Lancet in November 2010. Alcohol came out as the most harmful drug overall. Then other panels ran their own versions. European and international panels put alcohol first. A New Zealand study in 2023 ranked it most harmful for the whole population and for young people. A Canadian study in 2026 put it on top too. Nutt's summary: 'All show that alcohol is the most harmful drug.'

Eli: A legal drink, top of a list built to test illegal ones.

Mara: And topping a sixteen-box scorecard still doesn't say which of those boxes put it there.

The Bystander's Burden

Eli: That Canadian study shows where alcohol's points come from. A panel of experts scored a long list of drugs on the same sixteen kinds of harm. Alcohol got 79. Tobacco, 45. Cannabis, 15. The researchers credited alcohol's score to how widely it's used and how hard it hits society. Tobacco's came from severe long-term health effects and high addiction. Those are two different kinds of harm. Harm to users is what a drug does to whoever takes it. Harm to others is everything that spills onto the people around them.

Mara: So tobacco's points come mostly from the smoker's own body. Alcohol's spill outward.

Eli: An American panel ran the same method on a longer list. Fentanyl came out worst. Alcohol scored 73, just behind heroin, and ahead of cocaine and prescription opioids.

Mara: That could say more about how common drinking is than about the chemistry.

Eli: The Canadian team did say widespread use is part of it. Washington State makes a useful check, because cannabis was legal there too. In 2016, adults were asked about harm from someone else's drinking, or someone else's marijuana, over the past year. About 35 percent reported at least one harm from another person's drinking. For marijuana, it was under 14 percent. And the gap is widest where it's worst. Physical harm from someone's drinking was about five times as common as from someone's marijuana. Every one of those people was hurt by somebody else's use.

Mara: It reaches into homes too. Family or marriage trouble from someone's drinking came up about three times as often as from someone's marijuana.

Eli: Family and social adversity is part of what alcohol's 79 is made of.

Mara: That still leaves a fork, Eli. Alcohol's top score could come from what it does inside the drinker, or from what lands on everyone near the drinker, and one number on its own doesn't show which.

The Value of a Score

Eli: The answer sits in how Nutt's 2010 score was built. Sixteen boxes don't become one number by themselves. They're weighted and added up, a weighted composite score, like a final grade where the exam counts more than the homework. In Nutt's version, harm to users carried 46 percent of the weight. Harm to others carried 54. And as Alan Parrott and the Advisory Council on the Misuse of Drugs describe it, the two were counted differently. Harm to users was measured per person. Harm to others mostly in aggregate.

Mara: So the more people drink, the bigger that second number grows.

Eli: For David Nutt, that's the point. He and other researchers who use the method hold that a combined score captures harm to individuals and to society together.

Mara: That's the other reading, Eli, and part of it comes from Nutt's old council. Parrott and the council see the weighting as a value judgment, not a measurement. The council's line: 'harm expresses a level of damage. Value, on the other hand, indicates how much that level of damage matters.' And the aggregate isn't small. A 2020 survey of American adults found about 34 percent had been harmed at some point by someone else's drinking. That's around 113 million people. For cannabis, it was under 6 percent.

Eli: So a single score quietly settles something no survey can measure: how much a broken family weighs against a death.

Mara: The ranking Nutt built carries a weight somebody had to choose. What isn't known yet is how the order looks with the weights taken off.

Eli: Count only who dies and who gets sick, and the ranking faces a different test entirely.

The Body Count

Mara: That test exists. It's the Global Burden of Disease, a worldwide tally of how many people actually die, and how many healthy years are lost, from each cause. It doesn't ask a panel how much something matters. It counts. For 2021, it put tobacco's toll at about seven and a quarter million deaths in that one year. It's an estimate, with a wide margin either way, but every version of it lands in the millions.

Eli: The WHO says over 8 million people worldwide each year.

Mara: So tobacco has its bystanders too. Now alcohol, same tally, same year: about 1.8 million deaths from high alcohol use. Set against tobacco, that's roughly one death for every four.

Eli: Drug use overall comes to under half a million, about a quarter of alcohol's.

Mara: Then look at one country. In the United States, cigarettes are behind the vast majority of tobacco disease and death. Smoking alone causes nearly a third of the country's cancer deaths.

Eli: Alcohol's share of those cancer deaths is about 6 percent.

Mara: So this is what the top spot was made of. The expert models put alcohol first because of violence, family damage and costs that land on other people, counted in aggregate and weighted at more than half. The death counts point the other way: tobacco is far more lethal to the person actually using it. Does the answer depend on how harm is measured? It does.

Eli: Canada scored tobacco 45, alcohol 79. The body count runs the opposite way.

Mara: And cannabis has barely shown up in any of these death counts, Eli, which is a different kind of finding altogether.

The Toxicity Threshold

Eli: Toxicologists place it with something called the margin of exposure: how far a typical dose sits from the dose where real damage starts. A wide margin means room to spare. A narrow one means an ordinary amount is close to a dangerous one. A 2015 assessment put alcohol, nicotine, cocaine and heroin in the high-risk band. THC, the active ingredient that gets you high, from cannabis, was the one substance outside the risk categories altogether.

Mara: Alcohol in the same band as heroin. Cannabis in none of them.

Eli: A New Zealand analysis puts it as safety ratios, where a bigger number means more room to spare. THC's is over a thousand. Alcohol's is ten, down near heroin. The authors call death from a cannabis overdose almost impossible, and alcohol poisoning deaths not infrequent there. The CDC is more measured: 'While a fatal overdose caused solely by cannabis is unlikely, cannabis is not harmless.'

Mara: Behind alcohol on bystander harm in Washington, far behind on overdose here.

Eli: That last clause covers everything a safety ratio doesn't measure.

Mara: One physician's account shows what that can look like. His name isn't given. He writes: 'During my residency and fellowship, marijuana became an increasingly frequent means of escape from the pressures of daily life. I began to use it nightly as a means of rewarding myself for success or appeasing my failures of the day.'

Eli: Nightly, good days and bad. A safety ratio over 1,000 has nothing to say about that.

The Three-Way Matrix

Mara: For months, he planned every morning to quit and found a reason every night not to. Friends fell away. He eventually left his wife and young children, in part so he could use the way he wanted. Then one summer night in 2005, he was on the highway in light rain, smoking a joint, when the car ahead swerved into a barrier and spun around. The two cars stopped a few feet apart, both drivers shaken, neither hurt.

Eli: A joint at the wheel, and a few feet to spare.

Mara: After that, he resolved to quit. That pattern, the planning and the failing, has a clinical name: cannabis use disorder. The CDC says research suggests 3 in 10 people who use cannabis may have some form of it.

Eli: So the danger isn't a dose that kills. It's a habit that won't let go.

Mara: And for some, it reaches the mind. Daily or regular use has been linked to about twice the risk of schizophrenia or psychosis. Scientists don't yet know whether cannabis causes that, though it may make symptoms more severe.

Eli: That's where Parrott and the council's case ends up: drug harm, they argue, isn't naturally one thing. Their recommendation is a harm matrix instead of a single rating: a table with its own column for each kind of harm, filled in and never added up. The question Nutt put to Britain's drug classes ends up with columns, not a rank.

Mara: Read the three drugs down those columns and the listener's question has its answer. Tobacco kills far more of its own users than the other two. Cannabis is the least likely to kill anyone directly, and it carries real risks of dependence and mental illness. Alcohol is the more harmful one in the column about everyone else, not in the column about who dies. Nutt once stopped an interview because he realized he was speaking like government officials. The evidence he went after doesn't fit in one line.

Eli: Three things could still move those columns. One is long-running psychiatric studies on whether cannabis causes psychosis or sets off a vulnerability already there.

Mara: Another is long-term health data on edibles and vapes, set against smoked cannabis.

Eli: And future Global Burden of Disease rounds that try to count the deaths cannabis use disorder causes indirectly.

Mara: Look up Multicriteria decision analysis.

Eli: And keep following health with us.

Mara: And that's the story. I'm Mara.

Eli: And I'm Eli. Thanks for listening, and until next time.