The interactive layer did not finish loading. The article remains readable below; rankings and controls may be unavailable.
Read first Every drug on this page can seriously hurt you. The ranking says which is worse — never that anything is safe. If you or someone close to you is using: carry naloxone, test the supply, don't use alone. Full safety information →

The Shape of Harm

Thirteen drugs. Five kinds of harm. You decide what counts.

Some drugs do more damage than others — but "damage" isn't one thing. A drug can wreck your health slowly, kill you in a single night, or mostly hurt the people around you. Rank them one way and alcohol comes first. Rank them another way and it doesn't.

This page scores thirteen drugs on five kinds of harm, using published research. Then it hands you the sliders. There is no neutral setting. Deciding how much each kind of harm counts is a judgment about values, not a fact you can look up — so you make it.

Nothing here is safeThe ranking compares drugs to each other. Last place is still a drug that has hurt people. What that means →
The numbers are roughEstimates with wide margins. About half the orderings here cannot be told apart from noise — which is why the result is grouped into broad bands rather than a list from one to thirteen.
Everything shows its workingTap any drug for its full profile — where each number came from, and how good that evidence is.
Before you look at the ranking

Build the ranking yourself

Move a slider to say how much that kind of harm matters to you, and the table regroups as you go. There is no correct setting — that is the point.

The result is grouped into bands, not ranked one to thirteen. A band break falls where one drug outranks another with probability 90% or higher under the model's normal approximation, and a band is held together only where every pair inside it stays below that. The bands are a reading aid, not a statistical stratum: the grouping is one of several the model permits, and it moves when you move the sliders. For any specific pair, use Compare two substances below the chart rather than reading the horizontal rules. That probability is a closed-form approximation rather than a count of simulation runs, and it does not apply the simulation's clamping at 0 and 100.

Status after the evidence audit: legacy exploratory model The controls below still run model v5 so its assumptions remain inspectable. A July 2026 evidence audit found that several cells use evidence answering a different question from the column — withdrawal prevalence as medical danger, an acute cocaine heart-event study as chronic damage, and a selected “worst trip” survey as an episode rate. The numbers are retained for transparency, not promoted as calibrated measurements. The proposed v6 design separates outcomes from dependence, withdrawal, route, co-use and supply modifiers. Read the audit and redesign →
How much does each kind of harm matter to you?

All five columns are in play, so this is the combined model — four columns about the person using and one about everyone else, added together. It is not a per-person risk figure.

1–3 barely counts 4–6 counts some 7–8 counts a lot 9–10 matters most

Everything matters the same. Simple, but it's a real stance — it says a drug's damage to society counts exactly as much as its damage to your liver.

Compare two substances
vs

This is the model's own uncertainty under a normal approximation — not a significance test, and not a count of simulation runs. The interval widths it propagates are assigned by evidence grade rather than estimated from data, so the percentage describes this model's confidence in itself, not the strength of the underlying evidence.

Try to break it — move the sliders and see what you can force
The crossover
How to read the numbers

Open any drug and every cell carries an evidence grade. A 55 backed by a cohort study and a 55 backed by nothing look identical on a bar chart, so each is now marked Measured Indirect or Judgment. Of the 65 cells in this table, 21 rest on a drug-specific measured figure, 19 on a class proxy or extrapolation, 24 on judgment alone, and one has been suspended. That ratio is the single most useful thing on this page for deciding how much weight to give any particular number.

Every bar now carries its interval, and the numbers are ranges rather than points. Each cell's width comes from its evidence grade — ±8 where a drug-specific measurement exists, ±15 for a proxy, ±20 for judgment — propagated to the composite as though errors within a drug move together, which is the conservative direction. The result is uncomfortable and correct: of the 78 pairwise comparisons between these thirteen drugs, 41 have overlapping intervals. Only about half the ordering on this page is distinguishable from noise. The bands hold; the neighbours do not.

One drug is drawn as a hatched band rather than a bar. Methamphetamine's damage to others cell has been suspended — the only value available for it came from a formula that its neighbours in the same column are not held to, in a country where the drug was rare (section 07). Rather than print a number nobody should trust, the band shows every score meth could take, from that cell being 0 to it being 100. It spans four rank positions. That is the honest width of what is not known, and no other cell in this table has been examined closely enough to be sure it deserves a point rather than a band.

Every drug on this list is dangerous; the ranking is relative, not a verdict on safety. Grouped into bands under your current weights, worst first. Ordering within a band is not supported by the evidence and is not claimed. Tap any drug for its full profile — all five scores, how many Americans use it,39 its federal status, deaths per year, dependence risk, what is least certain about it, and every source behind it.

Go deeper

Everything above is the short version. Each of these is a full section.

Safety information

The part of this page that matters if the subject is more than academic.

If this is more than academic Read this part first

The numbers on this page describe populations, not you. Individual risk swings hugely with dose, frequency, mixing, mental-health history, and setting.

Interactions are where people actually get hurt. Depressant stacking — alcohol plus benzodiazepines plus opioids — multiplies overdose risk. For serotonergic drugs the risks are not equivalent: interaction risk differs substantially by drug, dose and mechanism. MAOIs and lithium warrant particular caution. SSRIs and SNRIs may alter or blunt effects rather than amplify them, but the evidence is incomplete and does not support treating every antidepressant as equivalent. If blunting leads someone to redose, that is a plausible hazard rather than a demonstrated one. Check a combination chart before mixing anything, and don't treat "antidepressant" as one category.

A personal or family history of psychosis or bipolar disorder meaningfully raises the acute psychiatric risk of psychedelics and stimulants. Research protocols exclude people with that history outright, and treat preparation, a safe setting and the presence of a sober companion as the real safeguards rather than optional extras.44

For dependence or withdrawal — especially alcohol or benzodiazepines, where quitting unsupervised can be fatal — talk to a clinician. Medically supervised detox exists for exactly this reason.

Carry naloxoneIt reverses an opioid overdose — and only that — and is sold over the counter in the US. It does nothing if you aren't there to use it — so the person nearby matters more than the person using.
Test the supplyIllicit fentanyl can appear unexpectedly across the drug supply, including in counterfeit pills and in products sold as stimulants. Test strips are cheap. Much of section 04's death curve runs through this.
Don't use aloneUsing alone sharply reduces the chance that anyone can recognise an emergency, give naloxone if it is an opioid, or call for help at all. Bystanders are often present and still don't act — knowing what an overdose looks like matters as much as being there. If alone is the only option, a hotline that stays on the call with you is better than nothing.
SAMHSA National Helpline — 1-800-662-4357. Free, confidential, 24 hours a day, every day, in English and Spanish. It handles treatment referral and information for substance use and mental health, for the person using and for their family. Treatment search at findtreatment.gov. For an immediate mental-health crisis, call or text 988.
A low rank is not a safety rating

Every substance on this page can seriously harm you. The ranking says which is worse. It does not say that anything here is safe, and rank 12 of 13 is still a psychoactive drug with a documented way of killing or damaging people.

This is a comparative scale, and comparative scales are easy to misread. Alcohol scoring 81 and psilocybin scoring 9 does not mean psilocybin is harmless — it means the two do different amounts of a thing that both do. Not one of these thirteen has a score of zero, and not one of them is without a documented serious harm.

And the supply has largely collapsed the low end of this scale. In the current US market, illicitly obtained pills and powders routinely contain something other than what they are sold as. Counterfeit benzodiazepines and prescription opioids frequently contain illicit fentanyl; substances sold as LSD have turned out to be NBOMe compounds, which — unlike LSD — have killed people at ordinary doses; material sold as MDMA is often adulterated. A drug's position in this table describes the molecule, not the thing in your hand. Anything not from a pharmacy should be treated as an unknown, which is why testing and naloxone appear in the notice above.

The floor is not zeroPsilocybin scores lowest of the illegal drugs. In Griffiths' survey of difficult experiences, 11% put themselves or others at risk of physical harm and three cases were linked to persisting psychotic symptoms.43 Wild mushrooms are also misidentified, and some look-alikes are lethal.
Legal is not safeThe two highest-mortality substances here are both legal. Alcohol kills roughly 178,000 Americans a year and smoking about 480,000 — together more than every overdose death combined, several times over.
Mild is not harmlessCaffeine, the calibration floor, has killed people through concentrated powder and pills, and produces a withdrawal syndrome that impairs 13% of those who stop.40
Rare is not neverKetamine's bladder damage can be permanent. Cannabis has no overdose deaths but a real dependence rate and a psychosis association. Low frequency is not the same as low stakes for the person it happens to.

If this is more than academic

The short form of what is above, worth repeating:

These numbers describe populations, not you. Interactions are where people actually get hurt — depressants stack, and MAOIs are the antidepressant class that kills. A personal or family history of psychosis or bipolar disorder raises the psychiatric risk of psychedelics and stimulants. Alcohol and benzodiazepine withdrawal can be fatal unsupervised; supervised detox exists for that reason.

Carry naloxone, test the supply, don't use alone. Those three do more to change outcomes than anything else here for opioids and for the counterfeit-pill supply, which is where most overdose deaths now are. They are not universal: naloxone reverses opioid overdose and nothing else — not alcohol poisoning, not a stimulant cardiac event, not severe withdrawal, not a psychiatric emergency. For those, the answer is emergency care.

Reliable sources

PsychonautWiki — per-drug pharmacology, dosing, interactions.

TripSit combination chart — what's dangerous to mix.

DrugBank / Drugs.com — prescription interactions.

US support: SAMHSA National Helpline, 1-800-662-4357 — free, confidential, 24/7, English and Spanish, for families as well as users. Treatment search at findtreatment.gov. Mental-health crisis: call or text 988.

← Back to the ranking

What the five harms mean

What each column measures, what each of the thirteen rows actually covers, and how well sourced every drug is.

05

What the five kinds of harm mean

Plain-language names for the five, with what each one actually covers. Later sections use the formal terms from the research — "harm to others," "acute crisis" — for the same things.

The whole point of keeping them separate is that a single score hides the shape. Alcohol and benzodiazepines are both depressants whose withdrawal can kill — but alcohol is bad at everything, while benzodiazepines are mild on most axes and terrifying on one.

And what the thirteen rows mean

Where the categories are too broad

A score has to attach to a defined substance, route and pattern — not to a drug name. Several rows here are categories doing the work of several distinct things, and in at least one case the underlying sources disagree about which thing they meant. Stating the intended referent is the minimum fix; splitting the rows would be the real one.

Alcoholwell definedEthanol, oral. A standard drink is 14 g in the lethal-dose table (US convention) and 10 g in the GBD thresholds (section 03). The two are not interchangeable and both appear on this page.
Opioidstoo broadScored as illicit heroin and fentanyl-adulterated supply. Nutt's referent is heroin; Gable's ratio of 6 is intravenous heroin; the death counts are dominated by illicit fentanyl. Prescribed pharmaceutical opioids are a materially different risk and are not what this row scores.
CocaineconflatedIntended as powder cocaine, intranasal — that is Gable's ratio of 15 and Nutt's overall 27. But the harm-to-others part-score borrowed in section 07 is Nutt's crack figure of 17, and crack scored 54 overall against powder's 27. These are two drugs and this row mixes them.
Nicotine (no smoke)addedPouches, vapes and nicotine replacement — the same addictive molecule as the row below, without combustion. Added because the table previously admitted that non-combustible nicotine "is not represented anywhere", which left the single clearest illustration of route-of-administration missing. It is a compound category and the weakest-sourced row here; the organ score leans on Swedish snus cohorts, and snus is fermented tobacco rather than pure nicotine.
Tobacco (smoked)renamedThis row was labelled Nicotine until an audit pointed out that its organ and mortality scores are smoking figures — tar and combustion, not the molecule. It has been renamed to match what it actually scores. Nicotine gum, pouches and vapes have materially different profiles and are not represented anywhere in this table.
Benzodiazepinestoo broadA class of many drugs with different half-lives and potencies, scored as one. The safety ratio shown is flunitrazepam specifically. Prescribed therapeutic use and high-dose recreational use are not distinguished, though their risk profiles diverge sharply.
Methamphetaminecontext-boundSmoked or oral. Its damage-to-others score is derived from a UK 2010 panel where meth was rare, and carries that context (section 07).
Cannabispotency-sensitiveSmoked herbal cannabis. Potency is the dominant variable and has risen sharply — daily high-potency use carries roughly 4.8× the psychosis odds of non-use. A single row cannot hold that range.
MDMA · Ketamine · LSD · Psilocybin · Caffeinereasonably specificSingle compounds at typical recreational oral doses (ketamine also insufflated). These rows are the least ambiguous, largely because there is less evidence to disagree about.
And the rows are not independent of each other A second structural problem, harder to fix than the naming. This model scores thirteen drugs one at a time, as though each were used in isolation. They are not, and the effect is measurable. Jones and Griffiths ran chronic oral caffeine against placebo in a double-blind design, then gave intravenous nicotine: caffeine maintenance significantly raised the subjective drug effect, the stimulation, and the participants' willingness to pay for nicotine.42

That is one row of this table changing the reinforcing value of another row, between two of the most ordinary substances here — coffee and cigarettes. And it runs in both directions: in a pilot from the same laboratory, psilocybin inside a structured cessation programme left 12 of 15 smokers abstinent at six months.45 Small and uncontrolled, so it settles nothing clinically — but a table where psilocybin sits at rank 11 and nicotine at rank 5 cannot express the sentence "one of these is being tested as a treatment for the other" at all.

The same logic runs through the rest: depressants stack, stimulants and opioids are now routinely co-used, and most cocaine and methamphetamine deaths in section 04 also involve an opioid. A thirteen-row table has no place to put any of it. Real use is combinations, and this instrument can only score elements — one at a time, each assumed to be acting alone.
One pair in this table isolates a single variable, and the result is unusually clean Nicotine (no smoke) and Tobacco (smoked) are the same drug. The addictive agent is identical; the delivery differs. Under equal weighting they score 31 and 49, landing one band apart — and the separation is settled at 99%, making it one of the most confident comparisons anywhere on this page, more confident than alcohol against methamphetamine.

That 18-point gap is combustion. Not nicotine, which both rows share, and which is why both score near the top on getting hooked. Smokeless nicotine still carries roughly 28% higher all-cause mortality than using no tobacco at all,48 so this is emphatically not a story about a safe product. It is a story about which part of the cigarette does the killing — and it is the only place in this table where a single variable has been isolated cleanly enough to answer that.

Three of thirteen rows are categories, not substances, and two of those — cocaine and nicotine — silently mix sources that meant different things. That is a design defect, not a data defect: no amount of extra citation fixes it. It is fixed by splitting the rows, which would mean rescoring.

06

The evidence, column by column

Where the numbers come from, what they actually say, and where the published sources disagree with each other. Every figure below is traceable to a citation at the foot of the page.

Addiction

Strongest evidence

The hardest data in the table, and now covering eight of the thirteen drugs rather than four. NESARC followed a nationally representative US sample across two waves — 15,918 nicotine users, 28,907 alcohol, 7,389 cannabis, 2,259 cocaine — and estimated lifetime probability of moving from first use to dependence.2

67.5%Tobaccomedian 27 yrs
22.7%Alcoholmedian 13 yrs
20.9%Cocainemedian 4 yrs
8.9%Cannabismedian 5 yrs

Lifetime transition probability, and the point at which half of all dependence cases had appeared.2

Correction to the original table The earlier draft said nicotine and cocaine "hook fastest." Two different things were being conflated. Nicotine has by far the highest lifetime risk, but it is the slowest to get there — half of nicotine dependence cases took about 27 years to appear, versus roughly 4 years for cocaine and 5 for cannabis.2 Highest risk and fastest onset are separate axes, and nicotine sits at opposite ends of them.
  • At the ten-year mark the ordering compresses sharply: 15.6% nicotine, 14.8% cocaine, 11.0% alcohol, 5.9% cannabis — closely matching earlier National Comorbidity Survey estimates.23
  • Speed of progression to a use disorder is fastest for heroin (median ~0 months), then cocaine (0–4 yrs), cannabis (1–6 yrs), tobacco (1–27 yrs), and alcohol (3–15 yrs) — a synthesis across several cohorts, which is why it's given as ranges rather than points.3
  • Psychedelics sit at the lower end for mechanistic reasons, not just statistical ones — lower relative risk, which is not the same as low risk. Psilocybin has limited reinforcing effects and only marginal, transient self-administration in animals — the standard abuse-liability model.4 It can produce tolerance, but there's no evidence of a withdrawal syndrome after chronic administration.5 NIDA does not consider it addictive, as it doesn't drive uncontrollable drug-seeking.5
  • A telling detail: on the ARCI euphoria scale used to predict abuse liability, psilocybin doesn't reliably score — it instead raises a dysphoria measure historically used to predict the absence of abuse potential.5

All thirteen drugs, and what is actually measured

Two survey families

The four NESARC figures cover only a third of the table. Adding the National Comorbidity Survey family — Anthony, Warner & Kessler's comparative epidemiology, summarised across drug classes — extends coverage to eight. Four drugs have no comparable figure at all, and saying so is more useful than inventing one.

DrugNESARC lifetimeNCS / AnthonyStatus
Tobacco (smoked)67.5%32%Measured in both
Alcohol22.7%15%Measured in both
Cocaine20.9%17%Measured in both
Opioids (heroin)23%Measured — NCS only
Cannabis8.9%9%Measured in both
Methamphetamine11%Class proxy — "stimulants"
LSD5%Class proxy — "psychedelics"
Psilocybin5%Class proxy — "psychedelics"
BenzodiazepinesClinical evidence instead
CaffeineSeparate literature
MDMANo comparable figure
KetamineNo comparable figure
5 measured — a real use-to-dependence probability for that specific drug 3 class proxy — the survey reports a drug class, not the drug 4 no figure — evidence exists, but not as a comparable percentage

Why the two columns disagree. NESARC and NCS were run two decades apart on different diagnostic criteria, and they measure subtly different things — NESARC estimates lifetime cumulative probability of transition using survival analysis, while the NCS figures are lifetime dependence histories among users.232 For nicotine the gap is large (67.5% vs 32%); for cannabis the two nearly agree (8.9% vs 9%). The ordering is stable across both, which is why the column's ranking is trustworthy while its exact digits are not.

One finding sits awkwardly across this whole column. Nicotine has the highest population dependence rate ever measured — and in the laboratory it is a weak and inconsistent reinforcer. Griffiths gave oral nicotine against placebo to eighteen people who had never smoked; some chose it reliably, others avoided it, and the drug is unreliable as a reinforcer in non-humans altogether.41 Population capture rates and individual reinforcement are measuring different things, and only one of them is what "getting hooked" sounds like it means. This column reports the first.

What the class proxies actually hide Three rows use a drug-class figure because no drug-specific one exists. "Stimulants 11%" pools methamphetamine with prescription amphetamines — meth is generally considered at or above that figure, so 11% is likely a floor, not an estimate. "Psychedelics 5%" pools LSD, psilocybin, mescaline and others; it is the only number available and it happens to align with the mechanistic evidence that these drugs are not reinforcing.45 Using a class figure for a single drug is a real approximation, and the table now says so rather than presenting it as measurement.

The four with no percentage still have evidence — just not in this form. Benzodiazepines: roughly four million daily US users, many meeting dependence criteria, with withdrawal possible after about a month of daily use.28 Caffeine: use disorder estimated at 8–20% of consumers, with a DSM-5-recognised withdrawal syndrome.2223 MDMA and ketamine: NESARC-III assesses "club drug" and hallucinogen use disorders, but publishes no use-to-dependence transition probability comparable to the figures above.33

Withdrawal danger

Sources disagree — range shown

The distinction that matters is unpleasant versus lethal. Only the GABA depressants can kill by cessation.

Alcohol. Severe withdrawal can progress to delirium tremens, a medical emergency. Reported frequency is roughly 3–5% of hospitalized withdrawal patients.6 Mortality figures vary meaningfully by source and era, and it would be false precision to pick one:

  • StatPearls: historically as high as 20%; now around 1% with advances in critical care and prompt treatment.6
  • Medscape: 5–15% even with appropriate treatment, closer to 5% under modern ICU management, versus as high as 35% before intensive care existed.7
  • Clinical reviews commonly cite untreated mortality in the 15–37% band.78

The honest reading: treated DTs is now a low-single-digit-mortality event; untreated it is a substantial fraction. That gap — not any single percentage — is the argument for supervised detox, and it's why alcohol scores 95 on this column.

Benzodiazepines. Withdrawal parallels alcohol's through the shared GABA-A mechanism, spanning mild anxiety to life-threatening delirium or seizures, with documented cases progressing to convulsive seizures and nonconvulsive status epilepticus.9 Hence the 90 — the one extreme column in an otherwise moderate profile.

Opioids and stimulants produce severe withdrawal — and for stimulants a psychologically dangerous crash — but not the classically lethal physical syndrome of GABA-depressant cessation. Scored high, below 95.

Organ and body harm

Mixed: strong + estimated
  • Tobacco's high organ score is really combustion and tar rather than the molecule — which is why its acute column is near zero while its chronic column is severe.
  • Opioids score lower here (~55) than their composite implies: the danger concentrates in addiction and overdose rather than in progressive tissue damage. That is a comparison rather than an absence — chronic use still carries endocrine effects, severe constipation, hypoxic injury from repeated overdose, and infection risk where injected. A genuinely different harm shape from alcohol.
  • Ketamine's 40 is the most under-appreciated number in the table. Chronic heavy use is strongly linked to ketamine-induced uropathy — ulcerative cystitis, urothelial ulceration, reduced bladder capacity and fibrosis — first described as a clinical entity in 2007.1011 It can progress to hydronephrosis and renal failure, and case reports also link chronic use to cholangiopathy affecting the biliary tract.1112 In the UK, ketamine misuse and related uropathy among 16–24 year-olds doubled between 2010 and 2020.12
  • Psychedelics have low physiological toxicity and no established fatal overdose at recreational doses; their real risk is acute and psychological, not bodily.4

Harm to others

Published figures

Alcohol is the clear outlier, and this single column is what drives its overall #1 ranking. In the Lancet MCDA, alcohol's harm-to-others part-score was 46 — more than double the next drug, with heroin at 21 and crack cocaine at 17.1 Meanwhile the most harmful drugs to individuals were crack (37), heroin (34), and methamphetamine (32).1 That split is the entire per-user-versus-societal crossover, in the source's own numbers.

Acute crisis

Mixed: strong + estimated
  • Opioids (95): highest single-episode death risk — respiratory depression, compounded by a fentanyl-adulterated illicit supply.
  • Cocaine (75) and meth (70): acute cardiovascular catastrophe — MI, stroke, hyperthermia — possible even in young or first-time users; cocaine raised myocardial-infarction risk 23.7-fold in the first hour in the original case-crossover study — on a very wide interval (95% CI 8.5–66.3) resting on nine exposed cases.31
  • Psychedelics (35–45): the one column where they leave the floor. Not toxicity — the acute psychological event, plus dangerous behavior in unprepared or unsupervised users and exacerbation of illness in those with or predisposed to psychotic disorders.4 LSD scores above psilocybin largely because the experience runs roughly twice as long, widening the window.
  • Tobacco (2): essentially no acute-crisis risk. Its harm is entirely chronic — the mirror image of the psychedelic profile.
08

What backs each drug

Every drug in the table now rests on at least three independent sources, and most on four or more. This is the audit: what each one is actually built from, and where the evidence is thinnest.

Where coverage is weakest Psilocybin rests on four sources rather than five or six. Caffeine was the thinnest row here until Griffiths' review of 57 withdrawal studies was added,40 which is a reminder that "thinly sourced" sometimes means the literature is thin and sometimes only means it had not been looked for. For psilocybin this is partly unavoidable — there is little mortality data because there are very few deaths to study, and Gable had to extrapolate its lethal dose from animals. Methamphetamine previously sat here as a conflict case, because Nutt's panel scored its societal harm far below this table's figure. That has now been corrected — its damage-to-others score was derived from Nutt's part-scores rather than judged, and has since been suspended entirely — the only cell in the table with no value at all (section 07).
← Back to the ranking

Dose and lethality

How much is a normal dose, how much would kill you, and why that ratio is less clean than it looks.

03

The missing dimension: how much

Everything above scores drugs as categories — as if "cocaine" were one fixed thing. It isn't. The same substance is a different risk at a different dose, and the gap between a normal dose and a fatal one varies more than a hundredfold across this table — from about six doses at one end to over a thousand at the other. That gap has a name.

Safety ratio — lethal dose ÷ typical recreational dose Gable 2004 · log scale

Robert Gable estimated, for 20 common substances, how many typical recreational doses it would take to kill a healthy 70 kg adult with no tolerance.13 A ratio of 10 means ten doses could be fatal. A ratio of 1000 means the lethal dose is effectively unreachable by accident.

Note the scale is logarithmic — each gridline is a tenfold jump. Gable is emphatic that these are ordinal, not arithmetic: you can say nitrous oxide is safer than GHB, but not that it is "20 times" safer.13 Two of the main thirteen are absent: Gable's benzodiazepine figure is for flunitrazepam specifically rather than the class, and nicotine's acute lethal dose is disputed — neither belongs on a chart of comparable ratios. Two substances shown here — GHB and nitrous oxide — sit outside the thirteen, included only to place the others on Gable's full scale.

What this reorders By this measure alone, alcohol sits in the bottom band of the scale, alongside heroin, GHB and methamphetamine. Sellman puts its ratio of about 10 nearer heroin (6) and GHB (8) than MDMA (16).14 That particular comparison is arithmetic, and on the logarithmic scale Gable insists on it is weaker than it sounds — alcohol is roughly equidistant from heroin and MDMA in log space. The substantive point survives the correction: alcohol shares the narrow-margin band with heroin, GHB and meth (also 10), and is orders of magnitude away from LSD or cannabis, both above 1000. The drug with the widest legal availability has one of the narrowest margins for error.

Why a small ratio is so dangerous in practice

The caveat Gable states in his own abstract, which this page had omitted The majority of published acute-toxicity deaths involve a co-intoxicant, usually alcohol — and Gable excludes those cases from the calculation. Both halves matter. It means these ratios describe single-substance lethality in isolation, which is the cleaner question to ask; but it also means each ratio rests on whatever minority of fatalities happened to be uncomplicated, and that remainder is small and probably unrepresentative. For six substances he had fewer than three human fatalities to work from at all, and extrapolated from animals.

So the most common way people actually die on this list — combining drugs — is precisely what these numbers exclude by design. Gable also notes the ratios ignore tolerance, and that accidents, aggression and addiction were deliberately left out.

A ratio of 6 doesn't mean you're safe until dose six. It means the distance between the effect you want and the effect that kills you is small enough that ordinary variation can close it. Four things close it faster than people expect:

  • Purity is unknown in an illicit supply. The ratio assumes you know what you took. If a bag is twice as strong as the last one, your "one dose" was two — and against a ratio of 6, you just spent a third of your margin without deciding to.
  • Mixing collapses the margin. Safety ratios are single-substance figures. Depressants stack: alcohol with benzodiazepines with opioids share a respiratory-depression mechanism, so the combined margin is far narrower than any individual number implies.
  • Tolerance moves the wanted dose, not always the lethal one. Gable notes explicitly that his ratios don't reflect tolerance, and that where tolerance to the desired effect grows faster than tolerance to the toxic effect, the ratio narrows.13 This is why a long-time user chasing the old feeling is in more danger than a beginner, not less — and why returning to a previous dose after a break is a classic fatal error.
  • Route changes everything. Gable found intravenous heroin carried the greatest combined risk of dependence and acute lethality of the 20 substances he assessed; oral psilocybin the least.15 Same molecule, different route, different risk.

Dose-response: the curve inside a single drug

Alcohol as the worked example

Alcohol is the best-studied case of "the amount is the whole story," and the honest answer has moved. A pooled analysis of 599,912 current drinkers across 83 prospective studies found all-cause mortality rose with consumption in a curved relationship, with minimum risk at or below roughly 100 g of alcohol per week — about five to six UK glasses or pints as the paper puts it, or roughly seven US standard drinks at 14 g each.16

The older claim that light drinking is protective — the J-shaped curve — is now contested rather than settled. A meta-analysis of 87 studies reproduced the classic J-shape without adjustment, showing reduced mortality among low-volume drinkers, but the shape substantially reflects study-design bias, including comparing drinkers against ex-drinkers who quit because they were already unwell.17 The GBD 2016 analysis concluded that the consumption level minimising an individual's risk is zero.18

Where the field genuinely disagrees A later GBD analysis using 2020 data reached a more nuanced position — that risk thresholds vary substantially by age, sex and region, with some older adults potentially seeing benefit from small amounts, while risks for young people are significant at any level.19 This is an active disagreement between competent groups, not a settled fact. What all of them agree on: risk climbs steeply and monotonically at higher intakes, and the dispute is confined to the very bottom of the curve.

What the 2020 analysis actually varies by. It estimated two thresholds for every region, five-year age group, sex and year: the TMREL, the intake that minimises health loss for a population, and the non-drinker equivalence, the intake at which a drinker's risk matches a non-drinker's.19 Age moved them a lot. Region moved them some. Sex did not move them at all — and that null result is one of the paper's headline conclusions, not a footnote.

GBD 2020 · global thresholds by age · standard drinks per day
Minimum-risk level (TMREL)
02 drinks/day
Non-drinker equivalence (NDE)
07 drinks/day

One standard drink here is 10 g of ethanol, the GBD convention — not the 14 g US standard drink used in the lethal-dose table further down. The two are not interchangeable.19

Read the numbers above as a property of a population, not a permission A TMREL is the intake that minimises total health loss across a population, given that population's background rates of disease. It is not a personal safe limit, and it is not advice about you. The reason it rises with age is not that older bodies handle alcohol better — it is that the mix of things alcohol causes changes. In the young the burden is overwhelmingly injury, which rises from the first drink. Later, cardiovascular causes make up more of the burden, and those are the causes where small amounts show an apparent benefit — the same apparent benefit that [17] argues is substantially study-design artefact. The disagreement between [18] and [19] lives entirely inside that shift, which is why the age axis exists and the sex axis does not.

Deaths per user, not deaths per drug

Cohort data

Your other question — deaths per person — has real data behind it. A Danish register study followed 20,581 people in treatment for substance use disorders, recording 1,441 deaths across 111,445 person-years, and computed standardised mortality ratios: how many times more likely each group was to die than the general population.20

9.1×Heroin
7.7×Other opioids
6.4×Cocaine
6.0×Amphetamine
4.9×Cannabis

Standardised mortality ratios versus the general population. MDMA's crude mortality was 1.7 per 1,000 person-years and its SMR was not significantly elevated.20

Read this one carefully — it is the most misinterpretable number on the page. These are people in treatment, which selects for the most severe cases, and the SMR captures all-cause death, not just the drug. That cannabis appears at 4.9× does not mean cannabis kills at half the rate of heroin; it largely reflects who ends up in treatment and what else is going on in their lives. The ordering is informative. The absolute values do not transfer to a casual user.

For scale at population level: US overdose deaths peaked at 107,941 in 2022, fell to 105,007 in 2023, and then dropped to 79,384 in 2024 — an age-adjusted rate of 23.1 per 100,000, down 26.2% in a single year and the largest fall in the decade.37 The counts in the table below are 2024 finals for that reason; earlier drafts of this page used 2023, which now understates how fast this is moving.21

All thirteen: dose, lethal dose, and deaths

Audited cell by cell

First, what these three things actually mean, because they are easy to confuse and they answer different questions:

  • Typical dose — the amount someone actually takes to get the effect they want. Varies with tolerance, purity, and route.
  • Lethal dose — roughly how much would kill a healthy adult with no tolerance. For several drugs this has never been established in humans, because there aren't enough deaths to study.
  • Safety ratio — lethal dose divided by typical dose. It describes how much room a substance leaves for error, not how many doses any particular person could survive. A ratio of 6 means the fatal amount sits close enough to the ordinary amount that normal variation can close the gap; a ratio of 1000 means it is effectively unreachable by accident. These are ordinal population estimates, not a personal count. And they describe a pure substance in controlled conditions: under illicit-market conditions there is no stable lethal dose, because purity, adulteration, route, tolerance, individual sensitivity, co-use and how fast help arrives all move it. The ratio is a property of the molecule, not of the situation anyone is actually in.
  • Deaths per year — how many people actually die. This is a completely different question from lethal dose, and the two often disagree, which is the most important thing on this page.
DrugTypical doseLethal doseAcute safety marginUS deaths / year
Opioids (illicit) heroin 5–10 mg IV ~50 mg IV, no tolerance ~6 47,735 synthetic opioids (mostly fentanyl); 2,743 heroin (2024; was 72,776 and 3,984 in 2023)37
Alcohol 1 drink = 14 g ethanol ~10 drinks fast, BAC ≈0.40% ~10 ~178,000 all alcohol-attributable causes35
Methamphetamine 20–40 mg oral ~10× a typical dose ~10 28,722 psychostimulants, mostly meth (2024; was 34,855)37
Cocaine 20–30 mg per line ~15× a typical dose ~15 21,945 (2024; was 29,449)37
MDMA 75–125 mg oral ~16× a typical dose ~16 Rare. No separate CDC category — deaths appear under other headings27
Benzodiazepines Varies by drug Very high alone; far lower with opioids 20–80? Thousands, but almost always alongside opioids — rarely alone28
Ketamine 50–100 mg Not established in humans ~25? Rare. Chronic harm is bladder and kidney damage, not overdose11
Caffeine ~100 mg = 1 coffee ~10 g ≈ 100 cups ~100 Very rare; case reports, usually concentrated powder or pills23
Tobacco (smoked) 1–2 mg per cigarette Acutely high; disputed ~100 ~480,000 — but from smoking over decades, not overdose2930
Cannabis Varies widely None established 1000+? No confirmed overdose deaths. Harms are psychiatric and chronic2425
LSD ~100 µg None established 1000+? Essentially none from toxicity. Deaths are accidents while impaired13
Psilocybin 10–25 mg ~1000× — not reachable via mushrooms 1000+? One documented overdose death — a heart-transplant patient4
Measured — human fatality data exists; both dose and death figures are real Partial — dose figure solid, death count not separately tracked Weak — lethal dose extrapolated from animals because human deaths are too rare to study Split — the death figure and the lethal dose measure completely different things Combined — dangerous mainly in combination, so a single-drug number misleads
The single most important thing on this page Lethal dose and death count do not line up, and the mismatch runs both ways.

Nicotine has a wide safety ratio — you cannot realistically smoke yourself to death tonight — yet it kills roughly 480,000 Americans a year. Those deaths come from decades of smoking, not from overdose. Its acute score is near zero and its chronic score is near maximum, and both are correct.

Psilocybin is the mirror image: essentially no deaths ever recorded, one documented case in a patient with a transplanted heart. Cannabis has no confirmed overdose deaths at all.

Alcohol sits in the worst possible position — a narrow safety ratio of about 10 and a death toll of roughly 178,000 a year, more than double the entire US overdose count. (Smoked tobacco kills more, at about 480,000, but it has no comparable acute danger: nobody dies of an overnight cigarette overdose. Alcohol is the only substance here that is near the top on both counts.) It is dangerous both ways at once, which no other drug in this table manages.
Four things that make these death numbers slippery 1. Deaths are counted more than once. CDC states plainly that a death involving both fentanyl and cocaine is counted in both categories, so these figures cannot be added together.34 Stimulants were involved in about 57% of fentanyl overdoses by 2023.36

2. Deaths per year is not risk per user. A drug used by fifty million people will produce more deaths than one used by fifty thousand, even if it is far safer per person. Alcohol's enormous total partly reflects how many people drink.

3. "No recorded deaths" is not "safe." It means low acute toxicity. Psychedelics still carry psychiatric risk, and accidents while impaired are real. Cannabis has no overdose deaths but a measurable psychosis association.24

4. The lethal dose assumes conditions almost nobody meets — no tolerance, no other substances, known purity, healthy adult.13 Real overdoses usually involve a mixture, an unknown dose, or a lost tolerance after a break.

The audit, stated plainly. Of the thirteen rows above: four rest on human toxicity evidence plus category-level mortality surveillance (opioids, alcohol, meth, cocaine) — and the qualifier carries weight: the CDC figures are for headings such as psychostimulants, mostly meth and synthetic opioids other than methadone, most of those deaths involve more than one substance, and a death is counted under every drug it involved. Two have a solid dose figure but no separate death tracking (MDMA, caffeine). Four have lethal doses extrapolated from animal studies because human deaths are too rare to study — Gable marks these with a question mark, and so does this table (ketamine, cannabis, LSD, psilocybin). One is split between two different measures entirely (nicotine). One is dangerous mainly in combination, so its solo number understates it badly (benzodiazepines). Only a third of this table rests on direct human measurement, and not one of the death columns is a clean per-substance count. The honest presentation is to say which third, and to say what the counts actually count.

Why there is no dose calculator here It would be easy to turn these ratios into a tool that takes your weight and tells you a number. That tool would be actively dangerous, and Gable's own caveats explain why: the ratios assume a healthy 70 kg adult, no tolerance, no residue from previous use, a single substance, and known purity — and for several substances (four of them in this table) the lethal dose is extrapolated from animal studies rather than human data.13 Almost no real situation satisfies those conditions. These figures are for understanding why some substances leave room for error and others don't. They are not a personal safety margin, and treating them as one inverts their meaning.
← Back to the ranking

Who gets harmed

Overdose deaths by age, sex and race — and the eleven drugs with no equivalent data.

04

The other missing dimension: who

The alcohol model above varies by age and region because someone built one. No equivalent exists for the other eleven drugs — there is no published minimum-risk intake for cannabis or cocaine by age and region, and inventing one would be the exact failure this page keeps correcting. But that does not mean nothing is known. It means the evidence takes a different shape for each drug, and the honest move is to say which shape.

What is resolved by demographics, and resolved well, is who dies of an overdose. US vital statistics break that down by age, sex, and race and ethnicity — and the 2024 figures record the largest single-year fall in a decade.

2024 2023

What the collapse does and doesn't mean Overdose deaths fell 26.2% between 2023 and 2024, from 105,007 to 79,384 — the steepest drop in the ten-year series, driven mostly by fentanyl, whose rate fell 35.6%.37 That is real and large. It is also one year, it does not touch alcohol or tobacco, which kill far more people and are not counted here at all, and the rate is still roughly 1.6 times what it was in 2014.

All thirteen: what is actually known about who

Four tiers

Applying the same audit to the demographic question that section 07 applies to the harm columns. Only one drug has the full treatment, and five have essentially nothing.

Tier 1 — modelled by age, sex and region1 of 13
Alcohol Thresholds estimated for 21 regions × 14 age groups × sex × year, 1990–2020.19 This is the only drug in the table with a purpose-built model of how the right answer changes with who you are — and, as section 03 shows, its own authors found the sex axis empty.
Tier 2 — mortality resolved by age, sex and race3 of 13
Opioids · Cocaine · Methamphetamine National vital statistics give overdose death rates by age band, by sex, and by race and ethnicity, plus per-drug totals.37 But the breakdowns are not crossed. The demographic tables cover all overdose deaths together; the per-drug tables give national totals only. So "the rate for men aged 35–44" is knowable and "the cocaine rate for men aged 35–44" is not published here. Roughly three-quarters of overdose deaths involve an opioid, so the demographic shape is heavily an opioid shape.
Tier 3 — age-resolved, but measuring something else3 of 13
Tobacco (smoked) · Cannabis · Benzodiazepines Real age gradients, none of them a threshold. Nicotine: the damage is reversible on a steep age curve — stopping at 30, 40, 50 or 60 regains roughly 10, 9, 6 and 3 years of life.30 Cannabis: the psychosis association concentrates in early-onset and heavy use, with daily high-potency use at 4.8 times the odds of non-use.2524 Benzodiazepines: harm rises sharply with age — hip-fracture risk in older people raised by at least 50%.28 Three different quantities, none comparable to each other or to a TMREL.
Tier 4 — nothing comparable published5 of 13
Caffeine · MDMA · Ketamine · LSD · Psilocybin No age-, sex- or region-resolved comparative harm estimate exists for any of these. For several the reason is the same one that thins their rows elsewhere on this page: too few deaths and too little longitudinal use data to stratify. You cannot break a demographic curve out of a denominator this small. Ketamine is the one likely to move first — UK misuse and related uropathy among 16–24 year-olds doubled between 2010 and 2020, which is an age signal, but it is prevalence of a harm rather than a rate that can be set against the others.12

The audit, stated plainly. One drug of thirteen has a demographic model. Three have demographic mortality data that cannot be crossed with the drug itself. Three have an age gradient measuring a different quantity. Five have nothing. An age-and-region control for all thirteen would be about ninety per cent fabrication, which is why this section is a map of the gap rather than a twelfth copy of the alcohol widget.

← Back to the ranking

How reliable is this?

Agreement with two expert panels, an audit of every column, and what happens when the uncertainty is propagated.

02

How closely does this agree with Nutt 2010?

Nutt, King & Phillips (2010, The Lancet) had an expert panel score 20 drugs on 16 weighted criteria. Plotting their published overall-harm score against this table's composite at equal weighting, the two track closely: Spearman's ρ = 0.93 across the eleven overlapping drugs, and tightest at the extremes. (This was 0.94 before methamphetamine's damage-to-others cell was suspended — suspending a cell moves the composite, and every figure downstream of it.)

This is a calibration check, not an independent validation An earlier version of this section called Nutt "the strongest validation" and described it as an outside check. That claim was circular and has been withdrawn. Nutt's results did not sit outside this table — they helped build it. Methamphetamine's entire harm-to-others score is derived arithmetically from Nutt's part-scores (section 07). Alcohol's harm-to-others is anchored to Nutt's 46. Several per-drug notes cite Nutt's figures as their justification. Correlating the finished table against the source that shaped it measures internal consistency, not external agreement.

ρ = 0.94 is therefore an upper bound, inflated by an unknown amount. Some of it is real — most columns were not built from Nutt — but no honest reading treats this as validation. The fix is not to delete the plot. It is to stop calling it something it isn't, and to go find a panel that had nothing to do with this table.

Each point is one drug. Both axes run 0–100, but the two instruments were built differently, so what matters is that points rise together, not that they sit on the dashed line of exact agreement. Psilocybin and LSD anchor the floor in both; alcohol tops both. This plot is pinned to equal weighting — moving the sliders above does not move it, so it stays a fixed calibration reference rather than something the weights can tune. The two informative disagreements — meth and benzodiazepines — are explained below.

Why meth ranks lower for Nutt Nutt's aggregate lifts alcohol far above meth on harm-to-others weighting. On harm-to-user alone he scored meth 32 — essentially tied with heroin (34) and crack (37) at the very top, matching its high placement here.
Why benzos look mild there Benzodiazepine harm is lopsided: low on most axes, extreme on withdrawal lethality. A single composite dilutes that spike. This table gives withdrawal its own column precisely so the spike stays visible.

A fresh panel not used to construct this table

Canada 2026 · agreement drops

The proper test is a panel that never informed these scores. In January 2026 a Canadian MCDA did exactly that: 20 experts from six provinces scored 16 drugs on 16 harm dimensions at a two-day decision conference.38 Alcohol came first at 79, then tobacco 45, nonprescription opioids 33, cocaine 19, methamphetamine 19, cannabis 15.

DrugThis tableCanada 2026Rank shift
Alcohol81791st in both
Tobacco (smoked)53455th here · 2nd there
Opioids75332nd here · 3rd there
Methamphetamine68193rd here · 4th there
Cocaine67194th here · 5th there
Cannabis2515Last in both
Agreement falls from 0.93 to 0.55 — but that gap does not measure the leakage Across the six overlapping drugs, Spearman's ρ = 0.55, and at n=6 that is not statistically significant. It is tempting to read the drop from 0.94 as the size of the contamination. It isn't, and an earlier version of this page said so wrongly. The two comparisons differ in sample size (11 drugs versus 6), in which drugs overlap, in country, in era, in criteria, in the presence of tied ranks, and — most importantly — in construct, since Canada scored population-level harm and this table scores harm per user. The drop is consistent with the concern that the Nutt figure is inflated. It cannot quantify it.

One component can be isolated, and it cuts the other way. Restricting the Nutt comparison to the same six drugs Canada scored gives ρ = 0.94, not a lower figure — so the smaller drug set does not explain the gap, and controlling for it makes the contrast sharper rather than softer. Everything else remains confounded, and no arithmetic here separates leakage from the genuine methodological differences below.

The disagreement is not random either. It is almost entirely tobacco — 2nd in Canada, 5th here. That is a real methodological difference rather than an error on either side: the Canadian panel scored population-level harm, explicitly weighting how many people use each drug, so a legal product used by millions rises. This table scores harm per user and keeps prevalence out. Both are defensible; they answer different questions, and the gap between them is the same per-user-versus-societal crossover the model at the top of this page is built around.

One caveat on independence, which cuts against this page's own framing. David Nutt and Lawrence Phillips — two of the three authors of the 2010 paper — are co-authors on the Canadian study, which also uses the same MCDA method and decision-conference format. So this is a fresh panel, a different country, a different decade and different data, but it is not a methodologically independent tradition. A genuinely independent check would have to come from outside the MCDA family altogether. That check does not currently exist, and this page should not pretend otherwise.
07

Could these numbers come from a formula?

A fair challenge to everything above: if the digits are judgment, why not replace the judgment with arithmetic? This section tries. The answer splits cleanly by column, and the exercise surfaced one real error in the table.

Derivable — applied to one cell

Harm to others

This column needs the least judgment of any. Nutt's panel already published weighted harm-to-others part-scores; rescaling them so alcohol anchors at 100 yields a value directly, grounded in an expert panel rather than in a single author's intuition.

score = 100 × (nutt_others ÷ 46)

Nutt published part-scores for six of the thirteen drugs here.1 Running the formula on all six is the test the earlier version of this section skipped — it agrees closely on four and diverges sharply on two:

DrugFormulaTableStatus
Alcohol10095Agrees
Cannabis17.420Agrees
LSD2.23Agrees
Methamphetamine2.22Formula applied
Opioids (Nutt: heroin)45.765Table 19 higher
Cocaine (Nutt: crack)37.065Table 28 higher
This exposed a real error — now corrected The table originally scored methamphetamine 70 for damage to others. Derived from Nutt, it should be about 2 — a roughly seventy-point overstatement, by far the largest discrepancy in the document. The original score reasoned that meth "obviously" harms families and communities badly. But Nutt's panel scored meth's overall harm at 33 while scoring its harm to the individual user at 32 — which leaves a harm-to-others part-score of only about 1, near the floor, because in the UK context of the study meth's societal footprint was small. Rescaled against alcohol's 46, that lands near 2, essentially tied with LSD. The formula was right and the original judgment was wrong.

The chart used 2 for a time. It no longer uses any value. Correcting 70 to 2 moved methamphetamine from 2nd to 3rd under equal weighting and from 2nd to 4th under the societal preset — but a value obtained by a method the rest of the column is not held to, in a country where the drug was then rare, is not a value this table can defend. The cell is now suspended, and methamphetamine is drawn as a band spanning every score it could take. It still ranks at or near the top for harm to the person using, which the evidence does strongly support.
And the correction was applied to only one row — stated plainly Meth is the sole cell that was actually overwritten with the formula's answer. Opioids and cocaine sit 19 and 28 points above what the same formula gives them, and were left alone. As a rule that is indefensible: it holds one drug to a standard the others escape, and it happens to be the drug whose score the correction lowered.

There is a reason not to simply rewrite them, but it undercuts the column rather than rescuing it. Nutt's figures are UK-2010 and drug-specific in ways that don't transfer: his 17 is crack, not powder cocaine, and his heroin 21 is not the same object as this table's "opioids" in a fentanyl era. Rewriting those two cells would import that mismatch rather than resolve it — but the identical objection applies to meth, whose near-zero score reflects how little methamphetamine circulated in Britain in 2010, not what it does to families in the US now.

So the honest label for this column is not "derivable" but "checked against a formula, and corrected in one place." For the record, applying it to all six would drop cocaine from 65 to 37 and opioids from 65 to 46, which under the everyone else preset narrows cocaine's lead over meth from 12 points to 3. The ordering survives. The margins do not.
Half derivable

Acute crisis

Gable's safety ratios give a principled toxic component — take a log, because the ratios span three orders of magnitude:

toxic = 1 − log(safety_ratio) ÷ log(1000)

That works well for the poisons: alcohol lands within 7 points of the table's value, meth within 3. Then it collapses. It scores LSD's acute risk at zero, because LSD has no reachable lethal dose — while the table says 45.

Both are correct about different things. Gable measures lethality; the acute column also contains psychological crisis, which has no lethal-dose analogue and no published scale. Adding a second component and taking whichever dominates cut the error from 23.5 to 14.3 — a genuine structural finding: acute risk is two independent things wearing one label. But the psychological values are this page's own, so that half remains judgment, and the improvement is partly circular.

Ordinal only

Withdrawal danger

Derivable as tiers, not digits. The published evidence supports a four-step ladder — lethal (alcohol, benzodiazepines), severe but not classically lethal (opioids, stimulants), mild, none — because real mortality data exists only for the top tier.67 The gap between 90 and 95 for benzos versus alcohol is not measuring anything; the gap between tier 1 and tier 2 is.

Not derivable

Addiction

The surprise. NESARC gives hard lifetime-dependence figures for only four drugs; adding the NCS/Anthony family reaches eight, but four still have no comparable percentage at all, and three of the eight rely on a drug-class proxy.2 Worse, no scaling of that data reproduces the table. Rescaling linearly puts cocaine at 31; the table says 70.

Adding a speed-of-onset term fixes cocaine and breaks nicotine, because nicotine has the highest lifetime risk but the slowest capture — a median of 27 years. The best fit available still missed by about 22 points on average, well outside the ±10 the document claims.

The honest conclusion: the addiction column is not a formula in disguise. It is judgment that blends lifetime probability, onset speed, and compulsion intensity in a ratio never made explicit — and only five of its thirteen values rest on a drug-specific measured percentage.

Not derivable

Organ harm · Harm to self

Organ harm would need disability-adjusted life-years per user-year, standardised across substances. That data doesn't exist in comparable form — alcohol's liver and cancer burden is heavily studied, ketamine's bladder toxicity is known mostly from case series.11

Harm to self has a worse problem: it isn't independent. It overlaps addiction, organ harm and acute crisis, so a composite that includes all four double-counts. Nutt avoided this by defining sixteen criteria — nine harms to the user, seven to others — designed to minimise overlap, and weighting them explicitly. This table did not, which is a structural flaw in its column design rather than a bad number within it. It was six columns when that criticism was first written; it is five now, because one of them was deleted for exactly this reason.

Illustrative sensitivity analysis

If the digits are shaky, how much does the ranking move?

A stated error bar is worth nothing until you push it through the model. So: take each cell's interval — ±8 where a drug-specific measurement exists, ±15 for a proxy or extrapolation, ±20 for judgment, as marked on every cell in the drug panels — treat those as 95% intervals, resample all 65 cells 20,000 times, and re-rank each draw. An earlier version used one flat width per column, which assumed a measured cell and a guessed cell were equally trustworthy. They are not, and the grades now say so.

Call this what it is This is a sensitivity analysis, not an uncertainty quantification. It answers one question only: if the assumed error widths, the normal distribution, and the assumed size of the shared bias are about right, how much does the ranking move? It does not establish that any of those assumptions are right. The ±10 and ±20 bands are author estimates rather than measured standard errors; nearly every drug gets the same width within a column, when the evidence grades above show they plainly deserve different ones; clamping at 0 and 100 distorts the drugs sitting near the ends; the correlation between columns is assumed rather than estimated; and the biggest uncertainty of all — whether five columns is the right structure at all — is not simulated, because a model cannot resample its own shape.

Run one, independent errors. Each cell wobbles on its own. The ranking barely moves: alcohol takes first place in 97% of draws under equal weights and 100% under the societal preset. But this is the flattering assumption, and it flatters for a structural reason — five independent errors partly cancel when you average them, so the composite comes out steadier than its inputs.

Run two, correlated errors. Real mistakes are not independent. If this table has misjudged methamphetamine, it has probably misjudged it across several columns at once. Adding a shared per-drug bias — one author, one blind spot — gives the honest picture:

Where the answer holds, it holds firmly. Where it doesn't, it doesn't at all: under the per-user view the top spot is a genuine three-way contest between methamphetamine, alcohol and opioids, and this page's habit of naming a single winner there is not supported by its own error bars.

Rank intervals, equal weighting — the bar is the 90% interval, the mark is the median:

rank 1 — most harmfulrank 13
Simulation method — parameters, and how to reproduce it

These figures are computed in your browser when this section loads, not looked up from a stored table. Press re-run to resample; the numbers should move by a point or two and no more.

Distribution
Gaussian, via Marsaglia polar transform
Cell standard deviation
Derived from each cell's evidence grade, not assumed flat: measured 4, indirect 7.5, judgment 10 — i.e. the ±8 / ±15 / ±20 intervals shown in the drug panels, read as 95% intervals. A suspended cell is excluded from the draw and handled as a range instead.
Shared per-drug bias
sd 5, drawn once per drug per iteration and added to all five of its cells (correlated mode only)
Implied column correlation
Not a single number — it depends on the two cells' grades, because the shared bias is fixed at sd 5 while cell noise varies. Between two measured cells it is 0.61; two indirect, 0.31; two judgment, 0.20; measured against judgment, 0.35. Range 0.20 to 0.61. An earlier version of this panel quoted 0.50 and 0.32, which were the values from the previous flat-width model and were left behind when the grades took over.
Truncation
Clamped to [0, 100] after perturbation
Iterations
20,000 per run
Seed
7 on first load (mulberry32); the re-run button uses a fresh random seed
Weight vectors
equal 5/5/5/5/5 · per-user 8/6/8/1/7 · societal 4/4/4/10/5 · one-bad-night 2/4/1/1/10, in column order addiction, withdrawal, organ, others, acute
Weights perturbed?
No. Weights are the reader's values, not an uncertain quantity — only the evidence cells are resampled
Reference implementation
Published alongside this page as stability-simulation.js. The full analysis is also reproducible from data files — see below.

The correlated model is itself an assumption, and a consequential one — it is what moves alcohol from 97% to 79%. A shared bias of sd 5 says roughly that a systematic misjudgment about one drug is about as large as the noise on any single cell. That is a plausible guess about how a single author goes wrong, not a measured quantity. Both modes are exposed above so you can see how much rests on it.

The broad bands survive: nothing at the top ever falls to the bottom, and no psychedelic ever climbs into the top half. But adjacent ranks are not distinguishable. Opioids spans ranks 1–4. Cocaine and methamphetamine overlap almost completely. Ketamine, MDMA and cannabis are one undifferentiated block, and so are LSD, psilocybin and caffeine. Reporting these as an ordered list of thirteen implies a precision the model does not have — five or six broad bands, depending on the weighting, is what the evidence supports.

A column was removed. This section used to defend it. Earlier versions of this page had a sixth column, damage to your life, and admitted in this very section that it overlapped addiction, organ harm and acute crisis — meaning a reader who raised all four was counting the same outcomes repeatedly. The defence offered was that the distortion was small. Measuring it: the column correlated with addiction at r = 0.93 and with organ harm at r = 0.92, and deleting it moved two of thirteen drugs under equal weights and none at all under the societal preset.

That was presented as reassurance. It is the opposite. A column that changes almost nothing when removed is a column carrying almost no information — while still letting the same harm be counted twice by anyone who moved two sliders. It has been deleted, and every score, preset and simulation on this page now runs on five columns. The reader-facing consequence is small; the honest description of the instrument is meaningfully different.
What this means for the table One column could be largely replaced by arithmetic tomorrow — doing so already corrected a nearly seventy-point error in one cell, and exposed two further cells the correction was never applied to. One more is half derivable. Two are ordinal at best. Two are judgment that no available data can currently replace. And propagating the stated uncertainty shows the ordering is robust in bands but not in ranks. The right conclusion isn't that the table is worthless; it's that ±10 was too confident, thirteen ordered positions is too many, and one author is not enough.
← Back to the ranking
10

Known weaknesses

Everything wrong with this page that its author is aware of, collected in one place so nobody has to reconstruct it from ten sections. The first group describes what is missing from the evidence itself; the three after it describe what is wrong with the instrument built on top of that evidence, ordered roughly by how much damage each does to the conclusions. None of these are hidden elsewhere on the page — this is a summary, not a confession.

Evidence-audit update — the next step is redesign before rescoring A cell-by-cell review found that several citations are real and correctly quoted but placed under the wrong construct. Cannabis and caffeine studies establish that withdrawal occurs; they do not measure whether stopping is medically dangerous. Cocaine's 23.7-fold first-hour myocardial-infarction trigger is acute evidence, not chronic organ damage. The psilocybin “11% risked physical harm” figure comes from people selected for their most difficult experience, not from all users or episodes. The benzodiazepine evidence separates use disorder from physical dependence, while this model combines them.

That means the model cannot become valid by replacing a few scores. Outcome rates must be separated from causal and contextual modifiers first. The full review is published as INSTRUMENT_AUDIT_2026-07-23.md, every cell is classified in evidence-audit.csv, and the unscored v6 specification is in instrument-v6.json.

The evidence base — what has never been measured

Not faults in this instrument but absences in the literature it is built from. Most of the structural problems in the next group are downstream of these, and no amount of care in scoring would remove them.

No unified dataset exists to score againstNever measured
The dataset this table would need does not exist for any substance, let alone for all thirteen collected the same way. It would follow large groups of people over years and record the exact substance and its purity, dose, route and frequency, combinations with other drugs, age, physical health and psychiatric history, setting and access to medical care, and then addiction, injury, disease and effects on other people as they accumulate. What exists instead is a patchwork in which each study measures whichever population it could reach — lifetime users, past-year users, people in treatment, hospital admissions, or the dead — and those denominators are not interchangeable. Nothing on this page can be better than that patchwork allows, and several weaknesses below are this one fact reappearing in a particular column.The undefined denominator, the collapsed dose–response curve, the missing co-use and adulteration axes and the mixed damage-to-others column are all this absence seen from different sides
Addiction probabilities are not comparable across the row setSerious
Getting hooked is the best-evidenced column in the table, and even there only four of thirteen cells rest on a drug-specific transition probability — alcohol, tobacco, cocaine and cannabis, all from the same NESARC analysis.2 Seven are class-level proxies: psilocybin, LSD and MDMA all inherit the National Comorbidity Survey's 5% psychedelic-class figure, methamphetamine an 11% stimulant-class figure, illicit opioids a 1994 heroin figure from before fentanyl,32 smokeless nicotine an inference from tobacco, and caffeine a use-disorder prevalence range rather than a transition rate at all. Ketamine and benzodiazepines have no comparable percentage. Evidence about those two exists; it answers a different statistical question, and folding it into a column that answers this one would be worse than leaving the cell as judgment.Graded per cell in section 06
Rare severe outcomes cannot be distinguished from unrecorded onesSerious
Detecting an event that happens perhaps once in tens or hundreds of thousands of uses takes enormous samples and reliable reporting, and neither exists for most of this table. Where the record shows few cases, that is equally consistent with genuinely low risk, with underreporting, with low total exposure, or with any mixture of the three, and the available data cannot separate them. This bites hardest at the bottom of the ranking: psilocybin and LSD score low partly because little has been recorded against them, and "little recorded" is not the same claim as "little happens." A low score in a thinly studied row should be read as a wide interval rather than a small number — which is what the intervals are there to say.Visible as interval width on the ranking, not as a separate figure

Structural — the design of the instrument

Problems with how the thing is built, which no amount of extra evidence would fix.

Single-author scoringFundamental
Every cell was scored by one person reading published evidence. There was no panel, no independent scorers, no blinding, and no reconciliation between disagreeing raters. Published drug-harm analyses use twenty or more experts scoring independently and then reconciling in a structured decision conference. This is the single biggest gap between this page and a validated instrument, and it is not one that more careful writing can close.Discussed in section 08
Undefined exposure denominatorFundamental
"Per user" does not distinguish someone who drinks twice a year from someone who drinks daily, and those two cannot meaningfully share a harm score. A rigorous version would state a target — expected harm per 1,000 user-years at a specified pattern of use — and score against it. Every number here should be read as an implicit average over a mixed and unstated population.Stated in the intended-use box
Mixed substance categoriesFundamental
Three of the thirteen rows are categories rather than substances. Opioids (illicit) spans heroin, diverted pharmaceuticals and illicit fentanyl, whose risks differ enormously. Benzodiazepines covers a class with different half-lives and potencies, and does not separate prescribed from recreational use. Cocaine mixes powder and crack — which Nutt scored at 27 and 54 respectively, and one of this table's own cells borrows the crack figure. Renaming two rows to Tobacco (smoked) and Opioids (illicit) made the labels honest; it did not make the categories homogeneous.Section 05 lists the definition for each row
Rows are scored as though used in isolationSerious
There is no co-use axis. Real harm overwhelmingly involves combinations, and the effect is measurable in both directions: chronic caffeine significantly raises willingness to pay for nicotine, while psilocybin is under study as a treatment for tobacco addiction. A thirteen-row table has nowhere to put either result, and most cocaine and methamphetamine deaths in section 04 also involve an opioid.Section 05, with sources [42] and [45]
No route-of-administration axisSerious
Smoking, injecting, insufflating, swallowing and transdermal use produce materially different risks from the same molecule. The scores collapse all routes into one number, with an unstated assumption about which route is typical.
Dose–response collapsed to a pointSerious
Each cell flattens a nonlinear curve into a single value, so a substance that is modest at one pattern of use and severe at another receives one score. Section 03 discusses dose at length and the main model still cannot represent it.
No supply or adulteration axisSerious
For opioids and counterfeit pills, much of the present danger comes from unknown potency and contamination rather than from the pharmacology. That is a property of a market at a moment in time, not of a molecule, and a properly built instrument would separate intrinsic harm, route harm and current-market supply harm into different columns.
No benefit axisBy design
This scores harm only. The omission is deliberate — a harm ranking is coherent and a harm-minus-benefit ranking is not, since the two are not measured in comparable units — but it leaves the table silent on a substantial clinical literature. Psilocybin sits near the floor while under study for depression and end-of-life anxiety; ketamine is a licensed anaesthetic and approved antidepressant; benzodiazepines and opioids are prescribed daily for good reason. A low score is not an endorsement and a high score is not a verdict of uselessness.
The five columns may double-count one causal chainFundamental
The columns are added as though they were parallel outcomes, but at least three of them sit on a causal path: getting hooked raises exposure, which raises both acute and chronic harm, and danger in quitting is conditional on having become dependent in the first place. Summing all four can therefore count different stages of one process more than once, and it inflates precisely those drugs that score high on dependence. The instrument has to pick a type and commit: a consequence inventory, where each score is realised burden and overlaps must be explicitly excluded, or a causal risk model, where dependence modifies expected exposure downstream instead of entering as a fifth peer score. This page is currently neither, and the composite is the place that shows.Related to the acknowledged overlap in section 06, but larger — that concerns one column, this concerns the addition itself
Prevalence excludedBy design
Scores are per person using, not multiplied by how many people use. This is why smoked tobacco sits mid-table here and second in the Canadian population-level analysis. Defensible, and stated throughout — but it means the ranking should not be read directly as a policy priority list, because policy cares how many people are affected.

Evidential — what is actually behind the numbers

Problems with the inputs rather than the architecture.

The composite mixes two denominators, and the odd one decides the winnerFundamental
Four columns are individual-level: what a drug does to the person taking it. Damage to others is not — it descends from expert-panel scores covering crime, public cost, family harm and community effect, all of which depend on how widely and under what conditions a drug is used. The page states that this column alone carries societal footprint, but stating it does not remove it from the arithmetic.

And it is load-bearing. Alcohol leads this table only while that column is weighted at 3 or above. Below that, methamphetamine takes the top. Remove the column entirely and score on the four genuinely individual-level criteria and the order is methamphetamine 78, alcohol 75, opioids 74 — alcohol is no longer first. So the headline result of this page depends on the one column whose denominator differs from the other four.

That is not a reason to discard the finding: alcohol's societal harm is real, well evidenced, and the reason most serious analyses put it near the top. It is a reason to stop describing the composite as harm per user without qualification. The honest description is four per-user columns plus one societal column, combined by a weighting the reader chooses. Try it: set damage to others to 1 and watch the top of the table change
A third of cells rest on a measurementFundamental
Of the 65 cells, 21 are backed by a drug-specific measured figure, 19 by a class proxy or extrapolation, and 24 by judgment alone. One is suspended. Every cell now carries its grade in the drug panels, and the intervals are derived from those grades, but grading a weak cell does not strengthen it.Visible on every cell; tallied in section 01
Half the ordering is not distinguishableFundamental
Propagating each cell's interval to the composite gives overlapping ranges for 41 of the 78 pairwise comparisons. Only about half the orderings this page displays can be separated from noise at all. The broad bands are robust; adjacent positions are not, and should not be quoted as though they were.Section 01 and the cold-open questions
One cell has no defensible valueSerious
Methamphetamine's damage to others is suspended. The only available figure came from rescaling a UK 2010 expert part-score — a method that the same column's other cells are not held to, produced in a country where the drug was then rare. It is drawn as a band spanning four rank positions rather than printed as a number.Section 07 sets out the full correction history
The damage-to-others column is methodologically mixedSerious
It contains one arithmetically derived cell, one suspended cell, and nine judgments. Applying the derivation consistently would drop cocaine from 65 to 37 and opioids from 65 to 46 — which is not done, because those figures carry a UK-2010, crack-not-powder, pre-fentanyl frame that does not transfer. Both choices are defensible; holding one drug to a standard the others escape is not.Section 07
The evidence grades are themselves author-assignedSerious
Nobody audited the audit. The M / P / J assignment on each of the 65 cells reflects the same single judgment as the scores, and a second reader would grade some cells differently. Since the uncertainty intervals are now derived from those grades, an error in grading propagates directly into the stated confidence.
The 0–100 scale has no external anchorKnown
There is no fact in the world that corresponds to a 70 on organ damage. The scale is anchored only internally — alcohol's damage-to-others is set at 95 because Nutt's panel put it highest — which makes the numbers comparative within the table and meaningless outside it.
The choice of thirteen substances is unexaminedKnown
Why these thirteen? Fentanyl has no row of its own despite dominating the mortality data; GHB, inhalants, kratom, nitrous oxide and novel synthetic cathinones are absent. The selection reflects what is commonly discussed rather than any principled inclusion rule, and a different thirteen would produce a different-looking scale.

Validation — what has not been checked

Where agreement has been tested, and why the tests are weaker than they look.

No independent validation existsFundamental
Agreement with Nutt 2010 (ρ = 0.93) is not validation, because Nutt's results helped build this table — one cell is derived from them outright and several notes cite them as justification. It is a calibration check. The Canadian MCDA 2026 is a fresher panel and agreement drops to ρ = 0.55, but Nutt and Phillips co-author that paper and it uses the same method, so it is not methodologically independent either. A genuinely external check would need scorers blinded to these numbers, a separately assembled evidence base, and predefined scoring rules. That check does not exist.Section 02, including the withdrawn version of this claim
The sensitivity analysis is illustrative, not quantificationSerious
It answers one question: if the interval widths, the normal distribution and the assumed size of the shared bias are about right, how much does the ranking move? It does not establish that any of those hold. Clamping at 0 and 100 distorts drugs near the ends, correlations between columns are assumed rather than estimated, and the largest uncertainty — whether five columns is the right structure — cannot be simulated, because a model cannot resample its own shape.Section 07, with the parameters disclosed
Not transportable across place or timeSerious
Anchored to US conditions in 2026, in a fentanyl-adulterated supply. Several columns would move substantially elsewhere or in another decade — the mortality data alone fell 26.2% in a single year between 2023 and 2024. Nothing here has been tested against another region, era or policy environment.
Reproducible is not validKnown
The published data files let another person regenerate every figure from the same inputs. They say nothing about whether the inputs are right. Twenty-four of sixty-five cells are judgment, and no amount of reproducible arithmetic converts a judgment into a measurement.
Context is assumed and unstatedKnown
Supervised settings, sterile equipment, a known dose, a companion present and access to treatment all substantially change outcomes, and none are inputs. Under screening, preparation and monitoring, persisting adverse reactions to psychedelics are rare — which means a large share of what the acute column measures is circumstance rather than chemistry.Source [44]

What a stronger source can — and cannot — fix

The audit changed the earlier claim that every weak cell only needed a better literature search. Some do. Others are attached to the wrong question, so stronger evidence would only make the mismatch more confidently wrong.

Two apparent evidence promotions were construct mismatchesCorrected
Caffeine and cannabis withdrawal are real and well documented. The caffeine review reports headache and functional impairment; the cannabis meta-analysis reports withdrawal prevalence. Those sources strengthen a variable called withdrawal symptom burden. They do not strengthen a column defined as “whether stopping can physically kill you.” The prior version counted them as measured support for that column. That was an evidence-classification error. Keep the sources; move the construct
Benzodiazepines expose the dependence-versus-addiction conflationRedesign required
The 2025 joint clinical guideline states that physical dependence is expected with ongoing regular benzodiazepine use and is distinct from a substance use disorder. It cites an estimate that 1.5% of people who use benzodiazepines met criteria for benzodiazepine use disorder, while warning that rapid cessation in a dependent person can cause seizures or delirium. One “getting hooked” score cannot represent low use-disorder prevalence, common physical dependence under regular exposure, and severe conditional withdrawal at once. Split use disorder, physical dependence and withdrawal hazard
Cocaine's strongest chronic-harm citation belongs in the acute columnMove evidence
Mittleman's 23.7-fold estimate concerns myocardial-infarction onset during the first hour after cocaine use. It strongly supports acute cardiovascular risk. It cannot calibrate years of organ damage. The chronic column needs longitudinal morbidity, mortality or health-loss evidence under a specified use pattern.
The psilocybin denominator is conditional, not generalRelabel
Carbonaro's survey asked respondents about their most difficult psilocybin experience. The 11% who put someone at physical risk are 11% of that selected difficult-experience sample, not 11% of all psilocybin episodes or users. It is useful evidence about what can happen during a severe challenging experience, but it cannot directly set an episode-level acute-risk score.
Tobacco's external-harm count is real but uses a different denominatorDo not compare directly
Annual US deaths attributed to secondhand smoke are population-burden data. Most neighbouring cells are per-user judgments or UK panel scores. A measured number is not automatically comparable merely because it is measured. The column must choose external harm per user-year or total harm in a region-year and use that denominator for every row.
What can still be improved by searching Several cells genuinely need better studies: chronic opioid outcomes, route-specific nicotine harm, ketamine use-disorder incidence, benzodiazepine impaired-driving externality, and stimulant withdrawal outcomes. But searching comes after the construct, population, exposure window and denominator are fixed. Otherwise the project will continue finding good evidence for incompatible questions.

Future work

In rough order of how much each would improve the result per unit of effort. The first three are the ones that would change what this is, rather than how good it is.

Define the three estimandsSeparate one episode, one year of regular use, and total population burden. Name the denominator, geography, year, dose, route, frequency, supply and target population for each output.Do this first
Separate outcomes from modifiersAcute and chronic outcomes belong in the score. Use-disorder probability, physical dependence, withdrawal hazard, route, co-use, setting and supply uncertainty change exposure or risk and should not automatically add peer points.Changes what this is
Split acute harm in twoMedical toxicity is not the same construct as panic, psychosis, impaired judgment or accident risk. Report acute medical and acute behavioral/psychiatric harm separately per episode.Changes what this is
Split the compound rowsIllicit fentanyl, heroin and counterfeit pills; powder and crack cocaine; prescribed and nonmedical benzodiazepines; and route-specific nicotine products should be distinct scenarios.Changes what this is
Preregister the protocolFreeze intended use, target population, criteria, denominators, evidence rules, missing-data rules and value functions before any v6 scoring begins.High value, low cost
Build systematic evidence dossiersUse predefined searches, eligibility rules and risk-of-bias assessment for each scenario-by-domain estimate. Record natural units before translating anything to a common scale.High value, high cost
Convene an independent panelHave addiction medicine, toxicology, epidemiology, psychiatry, health economics, criminology and lived-experience experts score identical dossiers independently before discussion. Preserve and report disagreement.Required for an instrument
Anchor any 0–100 transformationDefine what 0 and 100 mean in natural units and use explicit value functions. Where evidence supports only ordering, publish ordinal bands instead of invented distance.Required for interval scores
Replace grade-width uncertaintyUse study confidence intervals, model uncertainty, scenario uncertainty and between-rater disagreement. M/P/J labels are useful provenance, but fixed ±8/±15/±20 widths are not empirical error estimates.Required for uncertainty claims
Validate outside the construction setTest predefined hypotheses against outcomes and panels not used to build the model, replicate in another region, and freeze a version for prospective prediction.Required for validity
Audit all 24 judgment cellsThe attached cell audit identifies which need a literature search, which need a new denominator, and which should remain missing. Searching is useful after the redesign, not instead of it.Cheap after protocol
SubtractOnce v6 has separate views, the public page should show the question and result first, then reveal technical detail. Preserve the audit trail without making every reader traverse it.Overdue
The short version One author, no panel, an undefined denominator, heterogeneous rows, outcome and causal variables added together, acute medical and behavioral risk collapsed into one column, several citations attached to the wrong construct, 24 of 65 cells resting on judgment, and no independent validation. The v5 chart remains useful as an interactive map of assumptions, but its 0–100 values are not calibrated measurements. The defensible next product is not a cleaner universal ranking. It is three separately defined views — one episode, regular use, and population burden — with missing values left missing and modifiers kept out of the outcome sum.
← Back to the ranking
11

References

Forty-nine sources. Tags mark what kind of evidence each one is — primary for original peer-reviewed studies, clinical for practice references, review for syntheses of a literature.

[1]
primaryNutt DJ, King LA, Phillips LD. Drug harms in the UK: a multicriteria decision analysis. The Lancet 2010;376(9752):1558–65. The load-bearing calibration reference — not an external check, see section 02. Expert panel scored 20 drugs on 16 weighted criteria. Overall harm: alcohol 72, heroin 55, crack 54, meth 33, cocaine 27, tobacco 26, cannabis 20, benzodiazepines 15, ketamine 15, MDMA 9, LSD 7, mushrooms 5. Harm-to-others part-scores: alcohol 46, heroin 21, crack 17. Harm-to-individual: crack 37, heroin 34, meth 32. Meth's harm-to-others is not published separately but follows from the additive model — overall 33 minus harm-to-user 32 leaves roughly 1, which rescales against alcohol's 46 to about 2. pubmed.ncbi.nlm.nih.gov/21036393
[2]
primaryLopez-Quintero C, Pérez de los Cobos J, Hasin DS, Okuda M, Wang S, Grant BF, Blanco C. Probability and predictors of transition from first use to dependence on nicotine, alcohol, cannabis, and cocaine: results of the NESARC. Drug and Alcohol Dependence 2011;115(1–2):120–30. Re-verified against the published record, July 2026 — every figure exact. Drug and Alcohol Dependence 2011;115(1–2):120–30. Cumulative probability of transition from first use to dependence: nicotine 67.5%, alcohol 22.7%, cocaine 20.9%, cannabis 8.9%. Subsamples of lifetime users: nicotine n=15,918, alcohol n=28,907, cannabis n=7,389, cocaine n=2,259. Half of dependence cases appeared roughly 27, 13, 5 and 4 years after first use for nicotine, alcohol, cannabis and cocaine respectively — the source of the onset-speed figures quoted on this page. Population: US civilian non-institutionalised adults 18+, which excludes incarcerated and unsheltered people and so undercounts the heaviest use. Source of every dependence percentage in this table. Lifetime transition: nicotine 67.5%, alcohol 22.7%, cocaine 20.9%, cannabis 8.9%. Ten-year: nicotine 15.6%, cocaine 14.8%, alcohol 11.0%, cannabis 5.9%. Median time to half of all cases: ~27, 13, 5 and 4 years for nicotine, alcohol, cannabis and cocaine respectively. pubmed.ncbi.nlm.nih.gov/21145178
[3]
primaryWagner FA, Anthony JC. From first drug use to drug dependence: developmental periods of risk for dependence upon marijuana, cocaine, and alcohol. Neuropsychopharmacology 2002;26(4):479–88. National Comorbidity Survey, n=8,098, ages 15–54, survival analysis. The independent replication that NESARC's ten-year figures agree with. Speed-of-progression ranges (heroin median ~0 months; cocaine 0–4 yrs; cannabis 1–6; tobacco 1–27; alcohol 3–15) are a synthesis across this and later cohorts, not a single study's output. pubmed.ncbi.nlm.nih.gov/11927172
[4]
reviewJohnson MW, Griffiths RR, Hendricks PS, Henningfield JE. The abuse potential of medical psilocybin according to the 8 factors of the Controlled Substances Act. Neuropharmacology 2018;142:143–66. Anchors the psychedelic floor. Limited reinforcing effects; only marginal, transient non-human self-administration. Names the real harms too: dangerous behavior in unprepared or unsupervised users, and exacerbation of illness in those predisposed to psychotic disorders. pubmed.ncbi.nlm.nih.gov/29753748
[5]
primaryCarbonaro TM, Johnson MW, Griffiths RR, et al. Subjective features of the psilocybin experience that may account for its self-administration by humans. Psychopharmacology 2020;237(8) (PMC10013695). Supports the "not addictive" claim in detail: NIDA does not consider psilocybin addictive as it doesn't produce uncontrollable drug-seeking; not reliably self-administered by monkeys; tolerance occurs but no evidence of a withdrawal syndrome after chronic administration. Also the ARCI euphoria/dysphoria scale point. pmc.ncbi.nlm.nih.gov/PMC10013695
[6]
clinicalAlcohol Withdrawal Syndrome. StatPearls, NCBI Bookshelf (NBK441882). Withdrawal delirium historically carried mortality as high as 20%, now around 1% with critical care and prompt treatment. Also documents withdrawal seizures at 8–48 hours and warns that patients with prior complicated withdrawal should not reduce intake without clinical consultation. ncbi.nlm.nih.gov/books/NBK441882
[7]
clinicalDelirium Tremens: Practice Essentials, Pathophysiology, Etiology. Medscape / eMedicine. The higher end of the mortality range: 5–15% despite appropriate treatment, closer to 5% with modern ICU management, versus as high as 35% before intensive care. Most common causes of death are respiratory failure and cardiac arrhythmias. Included precisely because it disagrees with [6]. emedicine.medscape.com/article/166032
[8]
reviewDelirium tremens — overview and cohort data. ScienceDirect Topics; incl. Ferguson et al. 1996. Cohort detail behind the range: in 200 consecutive alcohol-dependent admissions at an inner-city US hospital, 24% developed DTs and 8% of those died of complications. Risk concentrated among the unemployed, homeless, and those with more severe end-organ disease — a reminder these averages hide steep social gradients. sciencedirect.com/topics/delirium-tremens
[9]
clinicalMader EC Jr, Rathore SH, England JD, Branch LA, Copeland BJ. Benzodiazepine withdrawal catatonia, delirium, and seizures in a patient with schizoaffective disorder. J Investig Med High Impact Case Rep 2020;8:2324709620969498. Documents the full severity span — withdrawal ranging from mild anxiety to life-threatening delirium or seizures — in a case where lorazepam cessation triggered convulsive seizures and nonconvulsive status epilepticus. Concrete grounding for the benzodiazepine withdrawal score of 90. pubmed.ncbi.nlm.nih.gov/33138643
[10]
primaryWinstock AR, Mitcheson L, Gillatt DA, et al. The prevalence and natural history of urinary symptoms among recreational ketamine users. BJU International 2012;110(11):1762–6. The prevalence study behind ketamine's elevated organ score — urinary symptoms among recreational users, rather than inference from clinical case series alone. pubmed.ncbi.nlm.nih.gov/22416998
[11]
reviewJhang JF, Birder LA, Kuo HC. Pathophysiology, clinical presentation, and management of ketamine-induced cystitis. Tzu Chi Medical Journal 2023;35(3) (PMC10399845); with Shahani R, et al., Urology 2007;69(5):810–12, first describing ketamine-associated ulcerative cystitis. Mechanism and progression: irritable bladder symptoms advancing to painful ulcerated bladder, reduced capacity, detrusor overactivity and contracture; metabolites drive urothelial inflammation, barrier deficits, oxidative stress and wall fibrosis. pmc.ncbi.nlm.nih.gov/PMC10399845
[12]
reviewKetamine-induced uropathy: clinical recognition, epidemiology, and multisystem effects. Cureus case literature 2025 (PMC12276041, Shkoukani et al.; PMC12828242, Nazir et al.). Progression to hydronephrosis and renal impairment; ketamine-induced cholangiopathy with concomitant hemorrhagic cystitis as an emerging cause of cholestasis; UK misuse and related uropathy among 16–24 year-olds doubled 2010–2020. ncbi.nlm.nih.gov/pmc/PMC12276041
[13]
primaryGable RS. Comparison of acute lethal toxicity of commonly abused psychoactive substances. Addiction 2004;99(6):686–96. Re-verified against the published abstract, July 2026. Bands as published: below 10 — GHB oral, heroin intravenous, isobutyl nitrite. 10–20 — alcohol oral, cocaine intranasal, codeine, dextromethorphan, MDMA oral, methadone, methamphetamine oral. Above 20–80 — DMT, flunitrazepam oral, ketamine inhaled, mescaline, phenobarbital. 100 or above — fluoxetine, LSD oral, marijuana oral, nitrous oxide inhaled, psilocybin oral. Every figure used here falls inside its published band, and Gable reports ratios varying by more than a factor of 100, which is the source of the hundredfold claim in section 03. Two figures could not be confirmed from the 2004 paper: caffeine at 100 and nitrous oxide at 150 are not among the substances its abstract names and appear to come from Gable's 2006 book chapter; both are now marked estimated on the chart. Method: LD50 estimated from postmortem blood concentrations, reported overdose doses and pharmacokinetic modelling; fatalities involving co-intoxicants or trauma excluded, though Gable notes those are the majority of published cases. Values assume a healthy 70 kg adult with no tolerance and no residue from prior use. For six substances (DMT, ketamine, LSD, marijuana, mescaline, psilocybin) fewer than three human fatality reports existed, so the lethal dose is extrapolated from animal studies — marked "?" in the chart. Gable warns the ratios are ordinal and must not be arithmetically manipulated, and that tolerance narrows them. pubmed.ncbi.nlm.nih.gov/15139867
[14]
reviewSellman D. If alcohol was a new drug. New Zealand Medical Journal 2009;122(1303):6–8 (editorial). States the comparison directly: alcohol's safety ratio of 10 places it nearer heroin (6) and GHB (8) for overdose danger than MDMA (16), and far above LSD (1000) or cannabis (>1000). Draws its ratios from Gable (Addiction 2004). nzmj.org.nz — vol 122 no 1303
[15]
primaryGable RS. Toward a comparative overview of dependence potential and acute toxicity of psychoactive substances used nonmedically. American Journal of Drug and Alcohol Abuse 1993;19(3):263–81. The earlier study combining dependence potential with acute lethality across 20 substances. Intravenous heroin carried the greatest combined risk; oral psilocybin the least. Explicitly excludes behavioural deficits, perceptual distortion and chronic illness — the route-of-administration point. pubmed.ncbi.nlm.nih.gov/8213692
[16]
primaryWood AM, Kaptoge S, Butterworth AS, et al. Risk thresholds for alcohol consumption: combined analysis of individual-participant data for 599,912 current drinkers in 83 prospective studies. The Lancet 2018;391:1513–23. 40,310 deaths over 5.4 million person-years. All-cause mortality showed a positive curvilinear association with consumption; minimum mortality risk around or below 100 g of alcohol per week. HRs corrected for long-term variability using 152,640 serial assessments. thelancet.com/S0140-6736(18)30134-X
[17]
reviewStockwell T, Zhao J, Panwar S, et al. Do "moderate" drinkers have reduced mortality risk? A systematic review and meta-analysis. J Stud Alcohol Drugs 2016;77:185–98. Meta-analysis of 87 studies. Unadjusted, it reproduces the classic J-shaped curve — low-volume drinkers at 1.3–24.9 g/day showed reduced mortality (RR 0.86). The paper's contribution is showing how much of that apparent benefit is study-design artefact, notably sick-quitter bias. pubmed.ncbi.nlm.nih.gov/26997174
[18]
primaryGBD 2016 Alcohol Collaborators. Alcohol use and burden for 195 countries and territories, 1990–2016. The Lancet 2018;392:1015–35; with linked Comment "No level of alcohol consumption improves health." Concluded the level minimising individual risk is zero, as estimated protective effects for ischaemic heart disease and diabetes in women are offset by monotonic associations elsewhere. Alcohol was the seventh leading risk factor for deaths and DALYs in 2016 — 2.2% of female and 6.8% of male deaths. thelancet.com/S0140-6736(18)31571-X
[19]
primaryGBD 2020 Alcohol Collaborators. Population-level risks of alcohol consumption by amount, geography, age, sex, and year. The Lancet 2022. Included because it disagrees with [18]. Bryazka D, Reitsma MB, Griswold MG, et al; Lancet 2022;400:185–235. Burden-weighted dose–response curves across 22 outcomes, for 21 regions and 204 countries, by 5-year age group, sex and year, ages 15–95+, 1990–2020. Source of every figure in the age explorer above; one standard drink = 10 g ethanol. Global TMREL: 0.136 (males) and 0.273 (females) at 15–39, 0.527 and 0.562 at 40–64, 0.636 and 0.656 at 65+. NDE: 0.249/0.546, 1.69/1.82, 3.19/3.51. Across all regions the TMREL spans 0–0.603 at 15–39 and 0.114–1.87 at 40+. Differences between males and females were not statistically significant in any age band, and the authors explicitly recommend guidelines differentiate by age rather than by sex. Of those consuming harmful amounts in 2020, 59.1% were aged 15–39 and 76.9% were male. thelancet.com/S0140-6736(22)00847-9
[20]
primaryArendt M, Munk-Jørgensen P, Sher L, Jensen SO. Mortality among individuals with cannabis, cocaine, amphetamine, MDMA, and opioid use disorders: a nationwide follow-up study of Danish substance users in treatment. Drug and Alcohol Dependence 2011. 20,581 individuals in treatment 1996–2006; 1,441 deaths over 111,445 person-years. SMRs: heroin 9.1 (CI 8.5–9.8), other opioids 7.7 (6.6–8.9), cocaine 6.4 (3.9–10.0), amphetamine 6.0 (4.2–8.3), cannabis 4.9 (4.2–5.8). MDMA crude rate 1.7/1000 person-years, SMR not significantly elevated. Injection use raised hazard ratios significantly. Treatment-population sample — does not generalise to casual users. pubmed.ncbi.nlm.nih.gov/20971585
[21]
clinicalGarnett MF, Miniño AM. Drug Overdose Deaths in the United States, 2003–2023. NCHS Data Brief no. 522, December 2024. 105,007 overdose deaths in 2023; age-adjusted rate 31.3 per 100,000, down from 32.6 in 2022 after nearly quadrupling from 8.9 in 2003. Note deaths involving multiple drugs are counted in each category, so drug-specific rates are not mutually exclusive. cdc.gov/nchs/products/databriefs/db522
[22]
primaryÁgoston C, Urbán R, Richman MJ, Demetrovics Z. Caffeine use disorder: an item-response theory analysis of proposed DSM-5 criteria. Addictive Behaviors 2018;81:109–16; with Abdoli F, et al. Scientific Reports 2024;14:7644. Caffeine dependence is not negligible. Studies across countries put caffeine use disorder at roughly 8–20% of consumers; the Iranian cross-sectional study (n=1,228) estimated CUD at 19.5% and caffeine withdrawal at 46.6%. Caffeine withdrawal is a formal DSM-5 diagnosis; caffeine use disorder remains in Section 3, "conditions for further study." pubmed.ncbi.nlm.nih.gov/29454178
[23]
clinicalCaffeine Withdrawal. StatPearls (NBK430790); with PsychDB clinical summary. Headache is reported in up to 50% of people who stop, and roughly 13% experience clinically significant distress or functional impairment — the same figures as [40]. Onset is 12–24 hours, peak 20–51 hours, resolution 2–9 days. Real syndrome, no mortality — which is exactly why caffeine scores low on withdrawal danger despite scoring non-zero on withdrawal frequency. ncbi.nlm.nih.gov/books/NBK430790
[24]
primaryMarconi A, Di Forti M, Lewis CM, Murray RM, Vassos E. Meta-analysis of the association between the level of cannabis use and risk of psychosis. Schizophrenia Bulletin 2016;42(5):1262–9. Dose-response, pooled across 10 studies: odds of psychosis 1.97 (CI 1.68–2.31) at median use, rising to 3.90 (CI 2.84–5.34) among the heaviest users. Independently corroborated by Moore et al. 2007 (AOR 2.09) and Kiburi et al. 2021 in adolescents (OR 2.7). pubmed.ncbi.nlm.nih.gov/26884547
[25]
primaryDi Forti M, Quattrone D, Freeman TP, et al. The contribution of cannabis use to variation in the incidence of psychotic disorder across Europe (EU-GEI). Lancet Psychiatry 2019;6(5):427–36. The potency/dose finding, and a direct illustration of the "how much" point: irregular low-potency use roughly doubles psychosis odds, daily use raises it to 3.2 (CI 2.2–4.1), and daily high-potency use to 4.8 (CI 2.5–6.3). The 2015 south London study found daily high-potency users about five times more likely than non-users to have a psychotic disorder. thelancet.com/S2215-0366(19)30048-3
[26]
primaryHenry JA, Jeffreys KJ, Dawling S. Toxicity and deaths from 3,4-methylenedioxymethamphetamine ("ecstasy"). The Lancet 1992;340:384–7. The foundational MDMA toxicity series. Among 7 fatalities the pattern was fulminant hyperthermia, convulsions, disseminated intravascular coagulation, rhabdomyolysis and acute renal failure — following recreational misuse of small amounts, which is the key point for the acute column. pubmed.ncbi.nlm.nih.gov/1353554
[27]
reviewHall AP, Henry JA. Acute toxic effects of ecstasy (MDMA) and related compounds. British Journal of Anaesthesia 2006;96(6):678–85; with Parrott AC, Pharmacol Biochem Behav 2002. Mechanism and environment-dependence: severe serotonin syndrome has historically been quoted at 10–15% mortality — a claim about the syndrome at its most severe, not the fatality risk of MDMA use, and reported MDMA-associated cases usually involve another serotonergic drug; MDMA hyperthermia is amplified by high ambient temperature and crowding ("aggregation toxicity"). Also documents dose-nonlinearity — a small dose increase can produce a large plasma-level jump. Directly supports treating MDMA's risk as dose- and setting-dependent rather than fixed. bjanaesthesia.org/S0007-0912(17)35084-5
[28]
clinicalJohnson B, Streltzer J. Risks associated with long-term benzodiazepine use. American Family Physician 2013;88(4):224–6; with Brett J, Murnion B, Aust Prescr 2015. Beyond withdrawal: many of the ~4 million US daily benzodiazepine users meet DSM-IV dependence criteria. Driving risk comparable to a blood alcohol level of 0.05–0.079%; hip-fracture risk in older people raised by at least 50%; cognitive decline that did not resolve three months after discontinuation. Withdrawal symptoms are possible after only a month of daily use. aafp.org/afp/2013/0815/p224
[29]
clinicalTobacco-Related Mortality; Diseases and Death. CDC; and The Health Consequences of Smoking — 50 Years of Progress (Surgeon General, 2014). More than 480,000 US deaths a year including ~41,000 from secondhand smoke — about one in five deaths. Overall mortality among smokers is roughly three times that of never-smokers. Smokers die about 10 years earlier on average. ncbi.nlm.nih.gov/books/NBK294316
[30]
primaryDoll R, Peto R, Boreham J, Sutherland I. Mortality in relation to smoking: 50 years' observations on male British doctors. BMJ 2004;328:1519. The independent long-run cohort behind the CDC figures. About half of persistent cigarette smokers are eventually killed by smoking; the higher "two-thirds" figure comes from later cohorts (Pirie 2013, Banks 2015) and is not this paper’s; continuous smokers born 1900–1930 died about 10 years younger than lifelong non-smokers. Stopping at 60, 50, 40 or 30 regained about 3, 6, 9 and 10 years respectively — the strongest dose/duration-response evidence in the whole table. pubmed.ncbi.nlm.nih.gov/15213107
[31]
reviewSchwartz BG, Rezkalla S, Kloner RA. Cardiovascular effects of cocaine. Circulation 2010;122:2558–69. Anchors cocaine's acute score. Reviews Mittleman 1999 (Circulation 99:2737–41), whose case-crossover design put the risk 23.7-fold above baseline in the first hour, 95% CI 8.5–66.3 from just nine exposed cases; risk falls falling to about 4-fold in the second and third hours. Cocaine users have 3.8–6.9 times the overall MI incidence of non-users, and cocaine contributes to roughly one in four MIs in people aged 18–45. ahajournals.org/circulationaha.110.940569
[32]
primaryAnthony JC, Warner LA, Kessler RC. Comparative epidemiology of dependence on tobacco, alcohol, controlled substances, and inhalants: basic findings from the National Comorbidity Survey. Experimental and Clinical Psychopharmacology 1994;2(3):244–68. The source that extends dependence coverage beyond NESARC's four drugs. Americans aged 15–54: about 24% had a history of tobacco dependence, 14% alcohol, 7.5% dependence on an inhalant or controlled drug. Among users, roughly a third of tobacco smokers became dependent and about 15% of drinkers. Anthony's later per-class summary gives heroin 23%, cocaine 17%, alcohol 15%, stimulants 11%, cannabis 9%, psychedelics 5%, tobacco 32%. pure.johnshopkins.edu — Anthony 1994
[33]
primaryGrant BF, Saha TD, Ruan WJ, Goldstein RB, Chou SP, Jung J, et al. Epidemiology of DSM-5 drug use disorder: results from NESARC-III. JAMA Psychiatry 2016;73(1):39–47. Establishes what the modern survey does and does not measure. 36,309 adults; 3.9% had a 12-month drug use disorder and 9.9% a lifetime diagnosis. Use disorders are assessed for amphetamine, cannabis, club drugs, cocaine, hallucinogens, heroin, non-heroin opioids, sedatives/tranquilizers and inhalants — but as prevalence of disorder, not as a use-to-dependence transition probability, which is why MDMA and ketamine have no comparable figure in the table above. pubmed.ncbi.nlm.nih.gov/26580136
[34]
clinicalDrug overdose deaths by substance, 2022–2023. CDC / NCHS provisional and final counts; About Overdose Prevention. Per-drug US death counts: synthetic opioids (mostly fentanyl) 74,702; psychostimulants including methamphetamine 36,251; cocaine 29,918; natural/semisynthetic opioids 10,171; heroin about 4,000. Roughly 76% of 2023 overdose deaths involved an opioid. CDC states explicitly that deaths involving multiple drugs are counted in every applicable category, so these figures must not be summed. cdc.gov/overdose-prevention
[35]
clinicalEsser MB, et al. Deaths from excessive alcohol use — United States, 2016–2021. MMWR 2024; with CDC Alcohol-Related Disease Impact (ARDI) and NIAAA statistics. Average annual deaths from excessive alcohol use rose 29.3%, from 137,927 in 2016–2017 to 178,307 in 2020–2021 — about 5% of all US deaths. The earlier 2011–2015 ARDI estimate was 95,158 per year with 29 years of life lost per death. Leading causes: alcohol-associated liver disease, heart disease and stroke, poisonings, accidents, and alcohol-related cancers. cdc.gov/alcohol/facts-stats
[36]
primaryZhu DT, Bajaj SS, Sen A. Methamphetamine and cocaine overdose deaths in the United States, 1999–2023. Substance Use & Misuse 2025;60(12):1946–9. Why single-drug death counts mislead. Methamphetamine-involved deaths rose from 547 in 1999 to 34,855 in 2023; cocaine-involved from 3,822 to 29,449. Critically, stimulants were involved in about 8% of fentanyl overdoses in 1999 but 56.67% by 2023 — the "fourth wave" of the overdose crisis is largely a polysubstance phenomenon, so attributing a death to one drug is increasingly artificial. pubmed.ncbi.nlm.nih.gov/40509756
[43]
primaryCarbonaro TM, Bradstreet MP, Barrett FS, MacLean KA, Jesse R, Johnson MW, Griffiths RR. Survey study of challenging experiences after ingesting psilocybin mushrooms: acute and enduring positive and negative consequences. Journal of Psychopharmacology 2016;30(12):1268–78. The evidence behind psilocybin's acute-crisis score, which was previously the thinnest-supported non-zero cell on the page. 1,993 respondents (mean age 30, 78% male) described their single worst psychologically difficult experience. 39% rated it among the five most challenging experiences of their life; 11% put themselves or others at risk of physical harm; 2.6% behaved aggressively or violently; 2.7% sought medical help. Among those whose experience was over a year earlier, 7.6% had sought treatment for enduring psychological symptoms, with three cases linked to onset of persisting psychotic symptoms and three to attempted suicide. Read the denominator carefully: this is a self-selected online sample recalling their worst experience, so these are proportions of people who had a difficult trip and chose to report it — not rates among psilocybin users. The risk factors identified are the useful part: dose, duration, difficulty, and absence of physical comfort and social support. doi.org/10.1177/0269881116662634
[44]
reviewJohnson MW, Richards WA, Griffiths RR. Human hallucinogen research: guidelines for safety. Journal of Psychopharmacology 2008;22(6):603–20. Source for the screening advice in the safety notice at the top of this page. Classical hallucinogens are described as relatively safe physiologically and not drugs of dependence, with the risk being specifically psychological: the most likely adverse event is overwhelming distress during drug action, which can lead to dangerous behaviour such as leaving the session; prolonged psychoses are less common. The safeguards are exclusion of anyone with a personal or family history of psychotic or other severe psychiatric disorder, rapport with monitors established beforehand, careful preparation, a safe physical setting, and interpersonal support from at least two monitors throughout, with follow-up probing for hallucinogen persisting perception disorder. Persisting adverse reactions are rare when these conditions hold — which is the point: the risk is largely a property of the circumstances, not the molecule. doi.org/10.1177/0269881108093587
[48]
pooled cohortByhamre ML, Araghi M, Alfredsson L, et al. Swedish snus use is associated with mortality: a pooled analysis of eight prospective studies. International Journal of Epidemiology 2020;49(6):2041–50. The organ-harm anchor for the nicotine row, and the best long-run evidence on nicotine without combustion anywhere. Eight prospective Swedish cohorts. Exclusive current snus users, against never-users of any tobacco: all-cause mortality aHR 1.28 (95% CI 1.20–1.35), cardiovascular mortality 1.27 (1.15–1.41), other-cause mortality 1.37 (1.24–1.52), cancer mortality 1.12 (1.00–1.26). Risk rose with duration of use but not with weekly amount. Read it carefully in both directions: smokeless nicotine is clearly not harmless, and it is just as clearly nowhere near cigarettes, which carry two to three times all-cause mortality. A separate cohort of 41,162 adults found no association with major heart disease after adjustment, but did find raised stroke risk in never-smokers (HR 1.52, 1.01–2.30). doi.org/10.1093/ije/dyaa197
[49]
reviewClarke E, Thompson K, Weaver S, Thompson J, O'Connell G. Snus: a compelling harm reduction alternative to cigarettes. Harm Reduction Journal 2019;16:62. Read alongside Valen H, et al., International Journal of Cancer 2023, and the 2025 umbrella review of comparative disease risks. The population-level counterpart to [48]. Sweden reports the lowest daily cigarette prevalence in the EU at around 5%, alongside roughly 20% daily oral tobacco use, and the lowest tobacco-related mortality and male lung cancer incidence in Europe. Comparative reviews find lower risks of lung cancer, COPD, mouth cancer and heart disease for smokeless products than for cigarettes. Two cautions. The Valen systematic review reports raised risks of oesophageal, pancreatic, stomach and rectal cancer among snus users, rating confidence in those estimates from moderate down to very low. And this literature attracts industry funding, so it should be read as a set of contested estimates rather than a settled conclusion \u2014 which is why the nicotine row's organ score leans on the independent cohort pooling in [48]. Harm Reduction Journal 2019;16:62
[46]
meta-analysisBahji A, Stephenson C, Tyo R, Hawken ER, Seitz DP. Prevalence of cannabis withdrawal symptoms among people with regular or dependent use of cannabinoids: a systematic review and meta-analysis. JAMA Network Open 2020;3(4):e202370. Upgrades cannabis withdrawal from a judgment cell to a measured one. Eight databases, 3,848 citations screened, 47 studies, 50 cohorts, 23,518 participants. Pooled prevalence of cannabis withdrawal syndrome 47% (95% CI 41–52) — but the stratification is what matters here: 17% (13–21) in population-based samples, against 54% in outpatient and 87% in inpatient groups. Since this page scores harm per person using rather than per person in treatment, 17% is the applicable figure and the widely quoted 47% is treatment-skewed. Heterogeneity was extreme (I² = 99.2%), and prevalence rose with daily use and with concurrent tobacco or other substance use. Note the limit: this establishes that the syndrome is real and common, not that it is dangerous — which is what the quitting column actually measures. doi.org/10.1001/jamanetworkopen.2020.2370
[47]
primaryUS Centers for Disease Control and Prevention; Office of the Surgeon General. Secondhand smoke: health problems, and tobacco-related mortality. CDC; The Health Consequences of Smoking — 50 Years of Progress. Upgrades tobacco's damage-to-others from judgment to a counted figure, which it should have been from the start. Secondhand smoke causes an estimated 41,000 deaths a year among US adults who do not smoke — about 33,951 from heart disease and 7,333 from lung cancer — plus roughly 400 infant deaths. Around 58 million non-smoking Americans are exposed, and living with a smoker raises a non-smoker's lung cancer risk by 20–30%. Since 1964 about 2.5 million non-smokers have died from secondhand exposure. One of very few harm-to-others values in this table that is directly counted rather than inferred from an expert part-score. cdc.gov — secondhand smoke
[45]
primaryJohnson MW, Garcia-Romeu A, Cosimano MP, Griffiths RR. Pilot study of the 5-HT2AR agonist psilocybin in the treatment of tobacco addiction. Journal of Psychopharmacology 2014;28(11):983–92. Open-label pilot, psilocybin within a 15-week structured smoking-cessation protocol: 12 of 15 participants were abstinent at six-month follow-up, a rate well above those usually reported for behavioural or pharmacological treatments. Small, uncontrolled and unblinded, so it proves little on its own. It is cited here for a structural reason rather than a clinical one — like [42] it is one row of this table acting on another row, but in the opposite direction. This model has no way to represent either. doi.org/10.1177/0269881114548296
[40]
reviewJuliano LM, Griffiths RR. A critical review of caffeine withdrawal: empirical validation of symptoms and signs, incidence, severity, and associated features. Psychopharmacology 2004;176(1):1–29. The reason caffeine is no longer the thinnest row here. Systematic review of 57 experimental and 9 survey studies. Of 49 candidate symptom categories, 10 met validity criteria — headache, fatigue, decreased energy, decreased alertness, drowsiness, decreased contentedness, depressed mood, difficulty concentrating, irritability, and feeling foggy. Headache incidence 50%; clinically significant distress or functional impairment 13%. Onset 12–24 hours after abstinence, peak at 20–51 hours, duration 2–9 days. Symptoms appeared at habitual doses as low as 100 mg/day — roughly one cup of coffee. Expectancy was found not to be a prime determinant, which rules out the obvious placebo explanation, and avoidance of withdrawal was central to habitual use. The review concluded the syndrome was well enough characterised to warrant inclusion in the DSM; caffeine withdrawal is now a DSM-5 diagnosis, with caffeine use disorder listed as a condition for further study. doi.org/10.1007/s00213-004-2000-x
[41]
primaryDuke AN, Johnson MW, Reissig CJ, Griffiths RR. Nicotine reinforcement in never-smokers. Psychopharmacology 2015;232(23):4243–52. Eighteen never-smokers (fewer than 50 lifetime nicotine exposures), double-blind oral nicotine against placebo. The first demonstration that nicotine functions as a reinforcer in people who have never smoked. The result that matters for this table is the variability: nicotine is a weak and inconsistent reinforcer in non-humans, and in never-smokers some individuals chose it reliably while others avoided it. Nicotine carries the highest population dependence rate of any drug measured [2] while being, in controlled conditions, an unreliable reinforcer — a combination Griffiths described as making it a very unusual drug from an addiction standpoint. doi.org/10.1007/s00213-015-4053-4
[42]
primaryJones HE, Griffiths RR. Oral caffeine maintenance potentiates the reinforcing and stimulant subjective effects of intravenous nicotine in cigarette smokers. Psychopharmacology 2003;165(3):280–90. A measured interaction between two rows of this table. Double-blind, within-subject: chronic oral caffeine (200 mg/70 kg three times daily for at least 12 days) against placebo, then intravenous nicotine. Caffeine maintenance significantly increased subjective drug effect, stimulation, and willingness to pay for nicotine. Cited here not for the caffeine row or the nicotine row but because it belongs to neither — this model scores thirteen drugs independently and has no structure that can hold a result like this. doi.org/10.1007/s00213-002-1262-4
[39]
primarySAMHSA, Center for Behavioral Health Statistics and Quality. Key Substance Use and Mental Health Indicators in the United States: Results from the 2024 National Survey on Drug Use and Health. HHS Publication PEP25-07-007, NSDUH Series H-60, 2025. Source of every prevalence figure in the per-drug panels. Past year, aged 12+, 2024: marijuana 64.2M, hallucinogens 10.4M, prescription opioid misuse 7.6M, tranquilliser/sedative misuse 4.6M, cocaine 4.3M, methamphetamine 2.4M (0.8%, flat since 2021), heroin ~556,000. Past month: alcohol 134.3M, tobacco products 48.0M, marijuana 44.3M, nicotine vaping 27.7M. 73.6M (25.5%) used any illicit drug; 48.4M had a substance use disorder; 10.2M received treatment. Three limitations matter for this page. Counts are not mutually exclusive — polysubstance use is counted in every applicable row. LSD, psilocybin, MDMA and ketamine are pooled into one "hallucinogens" figure and cannot be separated, so four of the thirteen rows here have no drug-specific prevalence. And NSDUH samples the civilian non-institutionalised population, excluding people who are incarcerated or unsheltered — which systematically undercounts the heaviest use. The 2024 overall response rate was 11.3%, with a weighted interview response rate of 45.3%. samhsa.gov/data — 2024 NSDUH
[38]
primaryCrépault JF, Russell C, Asbridge M, et al. Drug harms in Canada: a multi-criteria decision analysis. Journal of Psychopharmacology 2026;40(2):286–95. DOI 10.1177/02698811251409147. The nearest thing to a holdout this page has. Twenty experts from six provinces, 16 drugs, 16 harm dimensions — ten to the person using, six to others — scored 0–100 at a two-day decision conference, then swing-weighted. Alcohol 79, tobacco 45, nonprescription opioids 33, cocaine 19, methamphetamine 19, cannabis 15. The authors are explicit that these are population-level harms reflecting Canadian prevalence and policy context, not individual-level harmfulness — which is why tobacco ranks far higher there than here. Independence is partial: David Nutt and Lawrence Phillips co-author this paper as well as [1], and the method is the same MCDA tradition. journals.sagepub.com/10.1177/02698811251409147
[37]
clinicalGarnett MF, Miniño AM. Drug Overdose Deaths in the United States, 2023–2024. NCHS Data Brief no. 549, January 2026. Supersedes [21] and [34] for 2024. Final NVSS counts: 79,384 overdose deaths in 2024, age-adjusted rate 23.1 per 100,000, down 26.2% from 31.3 — the largest single-year fall in the 2014–2024 series. By sex: male 44.3→32.2, female 18.3→14.1. By age (2023→2024): 15–24 13.5→8.5, 25–34 45.6→30.4, 35–44 60.8→44.2 (highest), 45–54 53.3→41.0, 55–64 49.2→38.6, 65+ 14.7→13.4. By drug: synthetic opioids 22.2→14.3, psychostimulants 10.6→8.5, cocaine 8.6→6.3, heroin 1.2→0.8. Race rates are not adjusted for death-certificate misclassification, which understates American Indian and Alaska Native rates by about 34%. Deaths involving multiple drugs are counted in every applicable category. cdc.gov/nchs/products/databriefs/db549
Which version this is A reviewer pointed out that describing revisions is good and publishing several of them at once is not — parts of this page were still describing a six-column model with methamphetamine scored at 2, while the chart ran five columns with that cell suspended. That has been fixed, and this block exists so the current state is stated once, authoritatively, rather than inferred from whichever section you happen to read.

Legacy model v5 — 13 substances × 5 criteria = 65 cells. Criteria: getting hooked, danger in quitting, damage to your body, damage to others, one bad night. A sixth criterion, damage to your life, existed in earlier versions and was deleted for overlapping the other four. Exactly one cell is suspended — methamphetamine's damage to others — and that drug is drawn as a band rather than a bar in consequence. Cell intervals derive from evidence grades at ±8 measured, ±15 indirect, ±20 judgment. Band breaks are drawn at P ≥ 0.90 under complete linkage. Grade tally: 21 measured, 19 indirect, 24 judgment, 1 suspended.

Anything on this page that contradicts the paragraph above is a bug, and the data files are the tiebreak. The legacy score files are cross-checked against the page; the audit files are separate and deliberately do not overwrite v5.
Which sources have actually been re-checked Not all of them, and pretending otherwise would be the same failure this page keeps correcting. Sixteen of the 49 references have been opened and checked against the primary source during this project, and those carry a dated verification note in their entry. The rest were compiled from the literature without a second pass, and their figures should be treated as accurate-to-the-best-of-one-author's-reading rather than confirmed.

Re-checking found real errors. Two examples. Gable's own abstract states that most published acute-toxicity fatalities involve a co-intoxicant — a caveat this page had omitted entirely from a section built on his ratios, and which materially changes what a safety ratio means. And two of the values plotted from him, caffeine at 100 and nitrous oxide at 150, are not in the 2004 paper the page cites; they appear to come from a later book chapter, and are now marked estimated. Neither error was invented — both were inherited by not looking closely enough.

The checks that came back clean are worth as much. NESARC's four dependence probabilities and all four sample sizes are exact. Gable's every plotted value falls inside its published band. The remaining 33 references are the obvious next piece of work, and the honest position until then is that this page's numbers are as good as one careful reading, not as good as an audit.
Reproducing this analysis The legacy v5 figures can be regenerated from the original five files, and the audit adds three more artifacts: scores.csv — all 65 cells with their evidence grade, interval and the specific basis for each; references.json — the 49 sources structured, with the note explaining what each one supports; weighting.json — the criteria, the weight presets, and every simulation parameter; reproduce.py — standard library only, no dependencies; stability-simulation.js — the rank-stability analysis as a standalone script; evidence-audit.csv — all 65 cells classified by evidential and construct problem; instrument-v6.json — the unscored redesign specification; INSTRUMENT_AUDIT_2026-07-23.md — the reasoning and development protocol.

Running python3 reproduce.py rebuilds the ranking under each preset, the composite intervals, the pairwise-overlap count, the correlations with Nutt 2010 and Canada 2026, the rank-stability analysis, and a reference audit. Anything it prints that disagrees with this page means the page is wrong. That check has already earned its place: suspending methamphetamine's damage-to-others cell moved the Nutt correlation from 0.94 to 0.93 and the Canada correlation from 0.64 to 0.55, and the script caught that the page was still quoting the old figures.

Reproducibility is not validity. It means someone else can get these numbers from these inputs — not that the inputs are right. Twenty-four of sixty cells are author judgment, and no amount of reproducible arithmetic turns a judgment into a measurement. The script makes the analysis checkable, not correct, and it says so when you run it.
What this is, and what it would take to be something else Stated plainly, so nobody has to infer it: this is an evidence-informed exploratory multi-criteria decision analysis by a single author, and an interactive argument about why "the most harmful drug" is a question with more than one answer. It is not a validated instrument, and the gap between those two things is not a matter of adding references.

A validated version would need, at minimum: a preregistered protocol fixing the intended use, denominator and time horizon before any scoring; criteria redesigned to remove overlap, with an operational definition and a stated endpoint for each — "getting hooked" currently blends transition probability, onset speed, compulsion severity and relapse into one number; a systematic evidence review with predefined search strategies and risk-of-bias assessment for every drug-by-criterion cell, of which there would be well over a hundred; a multidisciplinary panel — addiction medicine, toxicology, epidemiology, psychiatry, health economics, criminology, lived experience — scoring independently from identical dossiers, then reconciling through structured elicitation; swing weighting for any default weights, since setting every slider to 5 asserts that a 0→100 move in acute mortality matters exactly as much as a 0→100 move in public cost, which is a claim and not a neutrality; distributions instead of digits, propagated through the ranking; reliability testing — inter-rater, test–retest, and replication by a second panel working from the same protocol; validation against sources deliberately held out from construction; transportability testing across regions, eras and supply conditions; and publication of the protocol, data and code for independent replication.

Section 07 does a small piece of this — it propagates the stated uncertainty and finds the ordering robust in bands but not in ranks. Section 02 does another small piece, badly at first and then honestly. Everything else on that list is undone, and most of it is a research programme rather than a revision. The sliders should stay exactly where they are, though: separating what the evidence says from how much each harm should count is the one structural thing this page already gets right.
What is still not sourced Being explicit about the boundary. The references support the orderings, the dependence percentages, the withdrawal-lethality ranges, the Nutt part-scores, the safety ratios, the mortality ratios, and the specific organ-harm mechanisms. They do not support the individual 0–100 digits in each cell.

Section 07 tests whether a formula could replace that judgment. Short version: harm-to-others can be derived wherever Nutt published a part-score — doing so revealed a nearly seventy-point error in methamphetamine and two further gaps, at opioids and cocaine, that remain uncorrected in the table; acute crisis is half derivable; withdrawal is ordinal only; and addiction, organ harm and harm-to-self cannot currently be derived at all. The stated ±10 accuracy is too confident — for the addiction column the real spread is closer to ±20, and harm-to-self double-counts other columns by construction.
← Back to the ranking

About this project

What this instrument claims to measure, what it does not, and how to check it yourself.

The research-development track is now separate

The chart on this page remains the transparent legacy model. The scientist-facing framework defines the future evidence process, and four exact pilot estimands are now provisionally frozen for feasibility and content review.

Open the Research Framework · Inspect the four defined pilot questions · See the first evidence pilot

What this instrument claims to measure — and what it does not

A structured, single-author comparison of the relative harm associated with recreational use of thirteen substance categories among US adults, as of 2026 — scored per person using, not per population.

  • The denominator is mostly harm per user — but not entirely. Prevalence is excluded from four of the five columns: a drug is not scored as worse simply because more people take it. Damage to others is the exception, since it descends from expert-panel judgments of societal footprint, and that column is what puts alcohol first. Weight it below 3 and methamphetamine leads instead. US prevalence is shown in each drug's profile39 as context, never as an input. This is why nicotine sits mid-table here and second in a population-level analysis (section 02). Only the damage to others column carries societal footprint, and only because its source did.
  • It is an exploratory MCDA, not a validated instrument. The scores come from one author reading published evidence, not from a panel, a systematic review, or a preregistered protocol. Section 07 sets out precisely what would have to change, and how much of it is missing.
  • It is not personalised risk. No number here applies to a person. Individual risk turns on dose, route, frequency, tolerance, supply purity, co-use, and health history — none of which are inputs to this model.
  • It is not stable across place or time. It is anchored to US 2026 conditions in a fentanyl-adulterated supply. Several columns would move substantially in another country or another decade.
  • The exposure denominator is undefined, and that is a real limitation rather than a quibble. "Per user" does not distinguish someone who drinks twice a year from someone who drinks daily, and those two cannot share a meaningful harm score. A rigorous version would state a target such as expected harm per 1,000 user-years at a specified pattern of use, then score against it. This page cannot yet, so every number should be read as an implicit average over a mixed and unstated population.
  • Four things that move real-world risk have no axis here at all. Route — smoking, injecting, insufflating and swallowing produce very different risks from the same molecule. Dose–response — each cell collapses a nonlinear curve into one point, so a drug that is modest at one pattern and severe at another gets a single number. Co-use — most real harm involves combinations, and the rows here are scored as though each were taken alone (section 05). Supply — for opioids and counterfeit pills, much of the present danger is uncertain potency and adulteration, which is a property of the market rather than the pharmacology, and a properly built instrument would separate intrinsic harm, route harm and current-market supply harm into different columns.
  • The scores assume an unstated context. Supervised settings, sterile equipment, a known dose, a companion present, and access to treatment all substantially change outcomes, and none of them are inputs. Griffiths' hallucinogen safety work is the clearest case: under screening, preparation and monitoring, persisting adverse reactions are rare,44 which means a large share of what the acute column measures is circumstance rather than chemistry.
  • It scores harm only, and harm is not the whole of a drug's effect. There is no benefit column, no therapeutic column, and no column for why anyone takes any of these in the first place. That omission is deliberate — a harm ranking is a coherent thing to build and a harm-minus-benefit ranking is not, since the two are not measured in comparable units — but it means the table is silent on a substantial clinical literature. Psilocybin sits near the floor here while being studied as a treatment for tobacco addiction,45 depression and end-of-life anxiety; ketamine is a licensed anaesthetic and an approved antidepressant; benzodiazepines and opioids are prescribed daily for good reasons. A drug scoring low here is not endorsed, and a drug scoring high is not without use.
09

The big-picture reads

Legal status is poorly aligned with this model's estimated harm

The two legal, culturally normalized drugs — alcohol and tobacco — do much of the real damage, while several substances people moralize about most sit at the floor. Nutt found the same: LSD, mushrooms and MDMA all ranked near the bottom despite Class A status. Note this is misalignment, not an inverse relationship — opioids, cocaine and meth are illegal and genuinely near the top. The claim is that legal status is a poor predictor of harm, not that it runs backwards.

"Most harmful" has two different answers

Per user, opioids and meth are arguably worst. For society, alcohol wins on its enormous harm-to-others. That's not a contradiction — it's the crossover. Toggle the presets above and you can watch the top of the list swap.

Dopamine explains a lot of the pattern — but not addiction

How directly a drug acts on the dopamine reward system tracks the addiction column reasonably well here: psychedelics (5-HT2A, little dopamine reward) sit at the floor, stimulants and opioids at the ceiling, the rest strung between. That is a useful intuition, not an explanation. Modern accounts of addiction also involve stress systems, learned cues and habit formation, relief of withdrawal, executive control, how fast a drug reaches the brain, and social context — which is why nicotine, a weak and inconsistent reinforcer in the laboratory,41 nonetheless produces the highest dependence rate ever measured. One axis orders this table; it does not account for it.

← Back to the ranking
Compiled July 2026 · 49 references · Exploratory synthesis Known weaknesses Legacy model v5 · v6 specified but not scored · Values are not calibrated