A systematic comparison of 20 drugs across 16 harm criteria found alcohol the most harmful overall (72/100), with cannabis scoring 20/100 — roughly one-quarter of alcohol's harm — demonstrating that drug classification correlates poorly with scientific evidence of harm.
Anyone interested in drug policy, cannabis scheduling debates, or understanding how cannabis's actual risks compare to other substances.
Alcohol: 72. Cannabis: 20. The gap between drug classification and scientific evidence of harm.
The Backstory
On October 30, 2009, David Nutt was fired.
As chairman of the UK government's Advisory Council on the Misuse of Drugs — the body responsible for providing scientific guidance on drug classification — Nutt had committed an unforgivable act. He had publicly stated, with supporting data, that ecstasy was statistically less dangerous than horse riding. One serious adverse event per 10,000 exposures for ecstasy versus one per 350 for equestrian sports. The comparison was deliberately provocative, but the numbers were real.
Home Secretary Alan Johnson dismissed him, explaining that Nutt "cannot be both a government adviser and a campaigner against government policy." The message was clear: the government's drug classifications were political decisions, and scientists who pointed out the gap between policy and evidence were not welcome.
One year later, Nutt published his answer. Not a letter of protest. Not an op-ed. A paper in The Lancet — one of the most prestigious medical journals in the world — that systematically ranked 20 drugs by their actual harms. The result was a chart that made drug scheduling look like what it was: a system built on politics, not pharmacology.
Alcohol was the most harmful drug. Cannabis ranked 8th. The paper has been cited thousands of times and remains the most influential drug harm assessment ever published.
The Scientist Who Wouldn't Stay Quiet
David John Nutt is a neuropsychopharmacologist based at Imperial College London. Before his firing, he had served on the ACMD's Technical Committee for seven years and was appointed chairman in January 2008. His career spans decades of research on the pharmacology of anxiety, depression, and addiction — work that positioned him to understand exactly how different drugs affect the brain and body.
After being sacked, Nutt did two things. He established the Independent Scientific Committee on Drugs (later rebranded as Drug Science), a charity dedicated to providing evidence-based drug information without political interference. And he assembled the committee to produce the study that his government employer wouldn't let him publish from the inside.
The co-authors were Leslie King, a former forensic scientist at the Home Office, and Lawrence Phillips, a decision analysis expert at the London School of Economics. The combination — a pharmacologist, a forensic chemist, and a decision scientist — was deliberate. This wouldn't be a subjective ranking. It would be a rigorous application of multicriteria decision analysis, a methodology used in everything from military strategy to corporate risk assessment.
The Method: How Do You Measure Drug Harm?
The fundamental problem with drug classification is that "harmful" isn't one thing. Heroin kills individuals efficiently but creates relatively contained social damage. Alcohol is less lethal per use but devastates families, generates violence, and costs health systems billions. Cannabis rarely kills anyone directly but may contribute to mental health problems. How do you compare these fundamentally different types of harm?
How They Did It
Multicriteria Decision Analysis Applied to Drug Harms
Define the criteria
Sixteen harm criteria, divided into two groups: nine measuring harms to the user (mortality, damage, dependence, mental impairment, loss of tangibles, loss of relationships) and seven measuring harms to others (injury, crime, environmental damage, family adversity, international damage, economic cost, community impact).
The split matters: harms to others carried slightly more total weight (54%) than harms to users (46%), reflecting the view that a drug's impact on society is as important as its impact on the individual user
Assemble the panel
Members of the Independent Scientific Committee on Drugs plus two invited specialists — experts in pharmacology, addiction medicine, forensic science, epidemiology, and decision analysis.
The panel included researchers with expertise across multiple drug classes, not just cannabis or alcohol specialists
Score each drug
In a one-day interactive workshop, the panel scored 20 drugs on all 16 criteria using a 0-100 scale, where 100 was assigned to the most harmful drug for each criterion and 0 indicated no harm.
The interactive format allowed panelists to challenge each other's scores and reach consensus, reducing individual bias
Weight the criteria
Panelists assigned relative importance weights to each criterion, recognizing that (for example) drug-specific mortality matters more than community reputation damage.
This is the key advantage of MCDA over simple ranking — it allows transparent value judgments while keeping the science separate from the weighting
Calculate overall scores
Weighted scores were combined to produce a single overall harm score for each drug, with the maximum possible being 100.
The final ranking reflects both the scientific evidence (how harmful is this drug on each criterion?) and explicit value judgments (how much do we care about each type of harm?)
Nutt, King & Phillips (2010), Lancet 376(9752):1558-1565
The Chart That Changed Everything
72 vs. 20
Alcohol's overall harm score versus cannabis. Alcohol scored 3.6 times more harmful than cannabis. Both substances are widely used. One is legal, celebrated, and available in every grocery store. The other remains a Schedule I controlled substance in many jurisdictions — classified alongside heroin as having 'no accepted medical use and a high potential for abuse.'
The disconnect between harm assessment and legal classification is the central finding of the paper. Tobacco (score: 26) also outscored cannabis while remaining legal. Mushrooms (score: 5) and LSD (score: 7) — often classified in the most restrictive drug categories — scored among the least harmful substances assessed.
Nutt, King & Phillips (2010)
What Cannabis's Score Means
Cannabis's harm score of 20 out of 100 breaks down differently depending on whether you look at harm to users or harm to others.
- Drug-specific mortality low
This does not mean cannabis is risk-free — it means the specific risk of acute fatal overdose is essentially zero
Cannabis has no established lethal dose in humans. Zero confirmed deaths from THC toxicity alone. This is where cannabis most dramatically separates from alcohol, opioids, and even tobacco.
- Dependence moderate
Cannabis dependence is a legitimate clinical concern, but its severity and treatment difficulty are substantially lower than alcohol or tobacco dependence
Approximately 9% of cannabis users develop cannabis use disorder. The withdrawal syndrome is real but rarely medically dangerous. Compare: alcohol dependence rate ~15%, tobacco ~32%, heroin ~23%.
- Drug-related mental functioning moderate
The mental health dimension is the primary driver of cannabis's harm score, particularly the dose-dependent psychosis risk with high-potency products
The cannabis-psychosis association, cognitive effects of chronic use, and adolescent vulnerability all contribute to this moderate score. This is where cannabis scores highest among harm-to-user criteria.
- Crime low
Most 'cannabis crime' would disappear under legalization — as Canada's experience has demonstrated
Cannabis-related crime is overwhelmingly a consequence of prohibition rather than pharmacological effects. Unlike alcohol, cannabis does not typically trigger violence or aggression.
- Family adversities low
This is one of the largest gaps between cannabis and alcohol in the harm assessment
While problematic cannabis use can strain relationships, the family damage associated with cannabis is a fraction of what alcohol produces (domestic violence, neglect, financial ruin).
- Economic cost low
Under legalization, the economic profile shifts from enforcement cost to tax revenue
Healthcare and criminal justice costs associated with cannabis are modest compared to alcohol and tobacco. The economic burden is heavily driven by enforcement costs rather than health costs.
Nutt, King & Phillips (2010)
The Aftermath: What the Paper Changed
Research Timeline
From Firing to Framework: The Impact of Nutt's Work
Nutt fired from ACMD
Lancet paper published
Drug Science founded
European replication
Canada cites evidence in legalization debate
US rescheduling momentum
DEA proposes rescheduling to Schedule III
Multiple sources; timeline compiled from news reports and policy documents
What People Get Wrong
Myth vs. Reality
The Evidence
Evidence-based analysis
The Deeper Question
Nutt's paper isn't really about drug rankings. It's about a question that every democracy should find uncomfortable: what happens when scientific evidence contradicts political consensus?
The UK government's answer was clear — fire the scientist. But the data didn't go away. And every year since 2010, the gap between cannabis's legal classification and its actual harm profile has become harder to defend. The paper doesn't tell governments what to do. It tells them what they can no longer pretend not to know.
For cannabis users, the practical takeaway is straightforward: cannabis carries real risks, particularly for adolescents, daily users, and those with mental health vulnerabilities. But those risks exist within a comparative framework where the legal substances most people consume without a second thought — alcohol and tobacco — score substantially higher on every systematic harm assessment ever conducted.
Key Takeaways
Drug harms in the UK: a multicriteria decision analysis
Nutt DJ, King LA, Phillips LD (2010) · The Lancet
Related Research
Key studies in this area
Cannabis prohibition origins: Harry Anslinger
Musto (1999)
The historical roots of cannabis scheduling — how racial politics, not pharmacology, drove the original classification decisions that Nutt's work contradicts
Racial disparities in marijuana arrests
Edwards, Bunting & Garcia (2020)
The human cost of classification based on politics rather than evidence — Black Americans arrested at 3.73x the rate of white Americans for a drug that scores 8th out of 20 in harm
Canadian legalization: five-year outcomes
Hall et al. (2023)
What happens when a country aligns its policy with the evidence Nutt's paper presents — the real-world results of treating cannabis proportionally to its actual harm
Lower-risk cannabis use guidelines
Fischer et al. (2017)
The harm reduction approach that follows logically from Nutt's framework — if cannabis isn't the most harmful drug, manage its risks rather than criminalizing its users
NASEM health effects of cannabis
National Academies of Sciences (2017)
The most comprehensive US evidence review, which — like Nutt's work — found cannabis's harm profile far below its Schedule I classification
Cannabis use disorder epidemiology
Hasin et al. (2015)
The dependence dimension of cannabis harm — roughly 9% of users develop CUD, the criterion where cannabis scores highest in Nutt's framework
Is alcohol really more harmful than heroin?
In Nutt's analysis, yes — when both personal and societal harms are counted. Heroin scores higher on harms to individual users (particularly mortality and dependence), but alcohol's massive impact on others — violence, family disruption, drunk driving, economic costs — pushes it to the top of the overall ranking. This doesn't mean alcohol is more dangerous per dose than heroin. It means that at the population level, accounting for how widely it's used and the breadth of damage it creates, alcohol causes more total harm.
Isn't it unfair to compare legal and illegal drugs?
This is the most common criticism, and it cuts both ways. Legal drugs benefit from wider availability, which inflates some harm dimensions (particularly harms to others). But legality also means regulated production, quality control, and age restrictions — factors that reduce some harms. The authors acknowledged this limitation. What they argued is that the current classification system isn't based on harm evidence at all, so the comparison — while imperfect — is still more evidence-based than the status quo.
Has this paper actually changed any drug laws?
Not directly in the UK, where cannabis remains Class B (one class below the most restricted). But it has been cited in legalization and rescheduling arguments worldwide. Canada's legalization framework was informed by comparative harm data. The 2024 US DEA proposal to move cannabis from Schedule I to Schedule III reflects growing acknowledgment of the gap between classification and evidence. Multiple countries have used Nutt's framework as a reference point in parliamentary debates.
Was Nutt biased because he was fired?
Nutt was fired in 2009 for stating his scientific views. The paper was published in 2010 in The Lancet — a journal with rigorous peer review. The methodology (MCDA) is transparent and reproducible. A 2015 European replication produced consistent results. The question of bias is worth asking, but the methodology is the answer: anyone can see the criteria, the weightings, and the scores. If the analysis were biased, it would be easy to demonstrate.
Does this mean we should legalize all drugs?
The paper doesn't make that argument. What it argues is that drug classification should be proportional to actual harm. Currently, some of the most harmful substances (alcohol, tobacco) are legal, while some of the least harmful (mushrooms, LSD) are classified in the most restrictive categories alongside heroin. Whether to legalize, decriminalize, or maintain prohibition for each substance is a separate policy question — but it should be informed by evidence, not inherited politics.
What the researchers found
Alcohol was the most harmful drug overall (score: 72/100). Cannabis ranked 8th (score: 20/100), roughly one-quarter of alcohol's harm. Heroin (55) and crack cocaine (54) ranked 2nd and 3rd. Mushrooms (5) and LSD (7) were least harmful. Drug classifications correlated poorly with actual assessed harm.
Why it matters
Published in The Lancet by the scientist fired from the UK government's drug advisory council, this paper is the most influential drug harm assessment ever published. It systematically demonstrated that drug scheduling is based on politics rather than pharmacology, and has been cited in rescheduling arguments worldwide.
The numbers in context
20 drugs scored across 16 criteria. Overall harm scores: Alcohol 72, Heroin 55, Crack 54, Methamphetamine 33, Cocaine 27, Tobacco 26, Amphetamine 23, Cannabis 20, GHB 18, Benzodiazepines 15, Ketamine 15, Methadone 14, Mephedrone 13, Butane 10, Khat 9, MDMA 9, Steroids 9, LSD 7, Buprenorphine 6, Mushrooms 5.
How the study worked
Multicriteria decision analysis (MCDA). Expert panel scored 20 drugs on 16 harm criteria (9 user harms, 7 societal harms) on a 0-100 scale in a one-day interactive workshop. Criteria were weighted for relative importance. Harms to others weighted 54%, harms to users 46%.
Who was studied
Expert panel assessment of UK drug harms. Not a population study — an expert-derived comparative harm framework applied to 20 substances.
What this study cannot tell us
Expert panel scoring involves subjective judgment. Alcohol's high availability may inflate some harm dimensions. The methodology conflates per-user harm with population-level harm. Replicated in 2015 European analysis with consistent results, partially addressing reproducibility concerns.
How to read the evidence
Strong: published in The Lancet using a formal decision analysis methodology (MCDA) with transparent criteria and weighting. Replicated in 2015 European analysis with consistent results. The methodology has known limitations (expert judgment, prevalence effects) but remains the most rigorous comparative drug harm framework available.
When this study was published
Published in 2010 in The Lancet. The harm rankings have been replicated in European analysis (2015) and remain widely cited. Cannabis potency has changed substantially since 2010, which could alter its harm score.
The bigger picture
This paper provided the most systematic evidence base for the argument that drug scheduling is political, not scientific. Published one year after the lead author was fired from the UK government's drug advisory council, it has been cited in cannabis rescheduling arguments worldwide and informed legalization debates in Canada, the US, and multiple other countries.
Questions still open
- Should drug scheduling be overhauled to reflect evidence-based harm rankings? How would harm scores change if cannabis potency increases continue? Should harm-to-others criteria be weighted differently from harm-to-users?
Common questions
Is alcohol really more harmful than heroin?
Does this mean cannabis is safe?
Has this paper changed any drug laws?
Read the original research
Drug harms in the UK: a multicriteria decision analysis.
Lancet, 376(9752), 1558-1565
One of the most prestigious medical journals in the world, published continuously since 1823.
Citation
Nutt, David J; King, Leslie A; Phillips, Lawrence D. (2010). Drug harms in the UK: a multicriteria decision analysis.. Lancet, 376(9752), 1558-1565. https://doi.org/10.1016/S0140-6736(10)61462-6
Explore the wider topic
- Cannabis Legalization and Your Brain: A Decade of Legal Weed
- 10 Lower-Risk Cannabis Use Guidelines (Based on Published Research)
- Cannabis and the Developing Brain: What Every Teenager (and Parent) Should Know
- Is Weed Addictive? What the Research Actually Shows
- Cannabis-Induced Psychosis: What It Is, Who's at Risk, and What to Do