Addiction

TL;DR. Addiction is compulsive use of a substance or behaviour despite harm, and it happens because drugs of abuse hijack the brain circuit that learns what is worth pursuing. Repeated use makes that circuit hypersensitive to cues associated with the drug while the pleasure it delivers declines, so wanting grows as liking shrinks. That dissociation is the core of the disorder and the reason "just stop" fails as advice. Addiction is neither a moral failing nor a purely biological inevitability: genetics contributes about half the risk, adversity and availability supply much of the rest, and the most effective treatments (methadone and buprenorphine for opioids, varenicline for nicotine, contingency management for stimulants) are underused everywhere, largely because of beliefs about what addicted people deserve.

Key takeaways

• Tobacco kills over 7 million people a year and alcohol about 2.6 million, far more than illegal drugs, which is a permanent corrective to how the subject is usually framed.
• Opioid agonist therapy (methadone or buprenorphine) roughly halves mortality in opioid use disorder. It is the single most effective intervention in this chapter and is restricted or stigmatised in much of the world.
• Dependence is not addiction. Someone taking opioids correctly for cancer pain becomes physically dependent without being addicted. Conflating the two harms both pain patients and addicted patients.
• Contingency management, providing modest incentives for verified abstinence, has the best evidence for stimulant addiction, for which no effective medication exists.
• Naloxone reverses opioid overdose in minutes and can be used by anyone. Distributing it to people who use drugs and their families saves lives.
• Language matters measurably: clinicians shown a vignette describing a "substance abuser" recommended more punitive responses than those shown "person with a substance use disorder."

What it is

In short: Compulsive use despite harm, which is a different thing from tolerance and from physical dependence, and conflating them harms both pain patients and addicted ones.

Modern diagnostic systems avoid the word addiction and use substance use disorder, graded mild to severe by how many of eleven criteria are met over twelve months. The criteria cluster into four groups:

Notice that the last group is neither necessary nor sufficient. This is the formal recognition of a crucial distinction:

Don't be confused: tolerance, dependence, and addiction are three different things. Tolerance is reduced response to the same dose. Physical dependence is the body having adapted, so stopping abruptly causes withdrawal. Both occur with many medicines that nobody calls addictive, including beta blockers, antidepressants, and corticosteroids. Addiction is compulsive use despite harm, with loss of control. A patient on long-term morphine for cancer pain is dependent and usually not addicted. Failing to make this distinction has produced two opposite policy errors: prescribing opioids too freely on the grounds that pain patients do not become addicted, and then abruptly cutting off patients who were stable and not addicted at all.

Behavioural addictions: gambling disorder is formally recognised and shares the neurobiology and much of the treatment response. Gaming disorder is recognised in ICD-11 and remains contested. Whether "food addiction" is a valid construct is actively debated.

The history

In short: Framed successively as sin, crime, and disease, with a marketing campaign in 1996 setting off the overdose crisis that followed.

Societies have used psychoactive substances throughout recorded history, and problems with them have been framed successively as sin, then crime, then disease.

What actually goes wrong

In short: Wanting grows while liking shrinks, which is why people keep using a drug they no longer enjoy and relapse years after withdrawal has passed.

The circuit. The mesolimbic dopamine system runs from the ventral tegmental area to the nucleus accumbens and prefrontal cortex. Its normal job is not to produce pleasure but to signal reward prediction error: the difference between what you expected and what you got. That signal drives learning about what to pursue.

Every addictive drug increases dopamine in this pathway, by different routes:

Crucially, drug-induced dopamine release is larger, faster, and does not habituate the way natural rewards do. Food and sex produce dopamine signals that decline as you become sated; drugs do not.

Incentive sensitisation. With repetition, the circuit becomes hypersensitive to drug-associated cues, so the sight of a syringe, a pub, or a lighter triggers intense wanting. Meanwhile the liking produced by the drug declines with tolerance. The result, developed by Robinson and Berridge, explains the phenomenology that puzzles observers: people continue using a drug they no longer enjoy, and relapse years after withdrawal has passed, triggered by a place or a smell.

Allostasis and the dark side. Koob and Le Moal's model adds the other half. Repeated intoxication drives compensatory changes: reward systems are downregulated and stress systems (CRF, dynorphin) are upregulated. The set point shifts, so the person needs the drug to feel normal and feels dysphoric, anxious, and irritable without it. Use moves from seeking pleasure (positive reinforcement) to escaping distress (negative reinforcement), which is why addiction deepens even as it stops working.

Loss of top-down control. Prefrontal regions responsible for inhibiting impulses, weighing future consequences, and monitoring behaviour show reduced function in imaging studies. The capacity to override a strong urge is diminished at exactly the moment it is most needed. This is the finding behind the "brain disease" framing, and it is also where the framing draws legitimate criticism: it can imply an absence of agency that does not match the observation that many people recover, often without treatment, and that contingencies (employment, incentives, relationships) reliably change behaviour in ways a purely deterministic model would not predict.

The defensible position is that addiction involves real, measurable changes in a motivation-and-control system, and that those changes shift probabilities rather than abolishing choice, which is exactly why treatments that change contingencies and environments work.

Withdrawal differs by drug and is worth knowing precisely:

What it does to the body

Organ damage is substance-specific and often exceeds the harm from the addiction itself.

• Alcohol: liver disease (Chapter 32), pancreatitis, cardiomyopathy, hypertension, atrial fibrillation, at least seven cancers (mouth, throat, oesophagus, liver, colorectum, breast), Wernicke-Korsakoff syndrome from thiamine deficiency, peripheral neuropathy, fetal alcohol spectrum disorder, and injury.
• Tobacco: lung and a dozen other cancers, COPD, coronary disease, stroke, peripheral vascular disease. Roughly half of long-term smokers die of a smoking-related cause.
• Opioids: respiratory depression and overdose death, constipation, hormonal suppression, and, with injection, endocarditis, abscesses, HIV, and hepatitis C.
• Stimulants: heart attack, arrhythmia, stroke, aortic dissection, hyperthermia, psychosis, dental damage from methamphetamine.
• Cannabis: bronchitis when smoked, cannabinoid hyperemesis syndrome, cognitive effects with heavy adolescent use, and psychosis risk (Chapter 42).
• Injecting, specifically: bloodborne virus transmission, bacterial and fungal infections, vein damage, and wound complications from adulterants such as xylazine.

Is it deadly?

Yes, and the ranking is not the one media coverage suggests.

• Tobacco: over 7 million deaths a year from direct use, plus more than a million from secondhand smoke.
• Alcohol: approximately 2.6 million deaths a year, roughly 4.7 percent of all deaths, and a leading cause of death in young adults.
• Illicit drugs: several hundred thousand deaths a year globally, with opioids responsible for the majority. North America's overdose crisis has run at tens of thousands of deaths a year, peaking above 100,000 annually in the United States, driven overwhelmingly by fentanyl.
• Suicide risk is elevated across all substance use disorders.

Is it contagious?

Addiction is not an infectious disease. But the phrasing hides two real phenomena.

Social transmission of use. Initiation is strongly influenced by peers, family, and availability. Prescribing patterns are also a form of exposure: regions with higher opioid prescribing had higher subsequent addiction and overdose rates.

Infections that travel with injecting. HIV, hepatitis B and C, endocarditis, and skin and soft tissue infections spread through shared equipment. This is the entire rationale for needle and syringe programmes, which reduce transmission substantially without increasing drug use, a finding replicated often enough to be considered settled.

Who gets it

In short: Roughly half genetic, plus early use, adversity, mental illness, and, at population level, price and availability.

• Genetics: heritability is roughly 50 percent across substances, with some substance-specific components (the ALDH2 variant common in East Asian populations causes intense flushing and nausea with alcohol and substantially reduces alcohol use disorder risk, while increasing oesophageal cancer risk in those who drink anyway).
• Age of first use: earlier initiation strongly predicts later disorder, plausibly because the adolescent brain's reward system matures before its control system.
• Adverse childhood experiences: strong dose-response relationships with later substance use disorder.
• Mental illness: high co-occurrence in both directions. Self-medication is part of it, and shared vulnerability is part of it.
• Availability and price: among the strongest population-level determinants. Alcohol consumption and harm track price and availability closely, which is why minimum unit pricing and tax measures work.
• Poverty, unemployment, trauma, and social dislocation: consistently associated. The overdose crisis in North America has been analysed as part of a broader pattern of "deaths of despair" alongside suicide and alcoholic liver disease.

Treatment, and how it works

In short: Methadone and buprenorphine roughly halve mortality in opioid addiction, and the best-evidenced treatment for stimulants is paying people small incentives to stop.

Opioid use disorder

Opioid agonist therapy is the treatment with the strongest evidence in this chapter.

• Methadone: a long-acting full agonist. Taken once daily, it occupies opioid receptors steadily, eliminating withdrawal and craving without the peaks and troughs that drive compulsive use. Usually dispensed daily under supervision at first.
• Buprenorphine: a partial agonist with high receptor affinity. It relieves withdrawal and craving with a ceiling on respiratory depression, so it is much safer in overdose, and it blocks other opioids from acting. Available as sublingual tablets or monthly injections.
• Naltrexone: a full antagonist, blocking opioid effects entirely. Requires full detoxification first, and adherence is the limiting factor.

A cohort meta-analysis found all-cause mortality roughly halved during treatment with methadone or buprenorphine compared with time out of treatment. Retention is the mechanism, and retention is worse with shorter and more restrictive programmes. Despite this, treatment is unavailable, criminalised, or heavily restricted in much of the world, and abstinence-only programmes remain common despite substantially worse mortality outcomes.

Naloxone reverses overdose within minutes by displacing opioids from receptors. It is available as a nasal spray, works in the hands of untrained bystanders, and has no effect on someone who has not taken opioids, so it cannot be misused. Take-home naloxone programmes save lives and are among the cheapest interventions in this book.

Alcohol use disorder

• Withdrawal management with benzodiazepines, plus thiamine to prevent Wernicke's encephalopathy. This is a medical necessity, not an optional comfort: unmanaged alcohol withdrawal can kill.
• Naltrexone: blocks opioid receptors, reducing the reinforcement from drinking and reducing heavy drinking days. It works without requiring abstinence first.
• Acamprosate: modulates glutamate signalling, supporting maintenance of abstinence.
• Disulfiram: blocks aldehyde dehydrogenase, so drinking causes flushing, nausea, and palpitations. It works through deterrence and requires supervision to be effective.
• Psychosocial: CBT, motivational interviewing, and twelve-step facilitation, which a Cochrane review found produced abstinence rates at least as good as other approaches.

Nicotine

• Varenicline: a partial agonist at the nicotinic receptor, reducing both craving and the reward from smoking. The most effective single agent.
• Nicotine replacement: combining a patch with a fast-acting form (gum, lozenge, spray) outperforms either alone.
• Bupropion, and cytisine, a cheap plant-derived partial agonist used for decades in Eastern Europe and increasingly recognised elsewhere.
• E-cigarettes: randomised evidence, including a UK trial, found them more effective than nicotine replacement for quitting. Long-term health effects are less harmful than smoking and not harmless, and youth uptake is a legitimate separate concern. The policy debate is genuinely difficult and the smoking cessation evidence is reasonably clear.
• Behavioural support roughly doubles the success of any pharmacological approach.

Stimulants

No medication has established efficacy. Contingency management, which provides vouchers or prizes contingent on drug-negative urine tests, has the strongest and most consistent evidence of any intervention for stimulant use disorder. It is used sparingly, largely because paying people to stop using drugs is politically unpopular, which is a clear case of evidence losing to moral intuition.

Harm reduction

The principle is that reducing the damage from drug use is a legitimate goal even when use continues, because dead people do not recover. The measures with good evidence: needle and syringe programmes, take-home naloxone, supervised consumption facilities (where no fatal overdose has been recorded in decades of operation across many sites), drug checking services, and low-threshold treatment access.

Portugal decriminalised personal possession in 2001, treating it as an administrative matter while substantially expanding treatment and harm reduction. Drug-related deaths, HIV infections among people who inject drugs, and incarceration for drug offences fell substantially, without the large increase in drug use that critics predicted. The evidence is genuinely positive and is sometimes overstated, since expanded treatment funding accompanied the legal change and more recent years have seen some deterioration alongside funding cuts. The reasonable conclusion is that decriminalisation plus investment worked, and that neither element alone would have.

What treatment costs

• Methadone: overdose risk during induction, QT prolongation, constipation, sweating, and the practical burden of daily supervised dosing.
• Buprenorphine: precipitated withdrawal if started too soon after a full agonist, which is why timing matters.
• Naltrexone: liver enzyme elevation, and it blocks opioid analgesia, which matters if the person has an accident.
• Varenicline: nausea, vivid dreams. Earlier neuropsychiatric concerns were not confirmed by the large randomised EAGLES trial, and the warning was removed.
• Disulfiram: the reaction is unpleasant by design and can be dangerous with a large drinking episode.

What the person can do

In short: Ask for medication rather than counselling alone, carry naloxone, never use opioids alone, and know that tolerance falls fast after any break.

• Recognise that treatment exists and works. Substance use disorders have outcome rates comparable to other chronic relapsing conditions such as asthma and hypertension, which is a useful frame: relapse is a reason to adjust treatment, not evidence that treatment failed.
• Carry naloxone if you or someone close to you uses opioids, including prescribed ones at high dose.
• Never use alone if using opioids, and be aware that tolerance falls fast after any break, including release from prison or hospital, which are periods of drastically elevated overdose risk.
• Ask for medication, not just counselling, for opioid, alcohol, and nicotine use disorder. Effective drugs exist for all three and are frequently not offered.
• Get alcohol withdrawal managed medically if you drink heavily and want to stop. Do not do it alone.
• Deal with the co-occurring condition. Untreated depression, anxiety, PTSD, and ADHD all make relapse more likely.
• Change the environment, not just the intention. Removing cues, changing routes home, and building non-drug sources of reward and structure are not soft measures; they act directly on the mechanism described above.

Living with it

Stigma is the field's largest structural obstacle. It shapes what treatments are funded, whether clinicians offer them, whether people seek help, and whether relapse is met with support or punishment. The evidence on language is unusually concrete: in an experimental study, clinicians presented with an identical case described as a "substance abuser" rather than "a person with a substance use disorder" were more likely to endorse punitive measures. This is why terminology has shifted, and it is not merely politeness.

Recovery is common, and the picture presented by treatment settings is biased because it sees people during their worst periods. Population surveys find that a majority of people who ever meet criteria for a substance use disorder are in remission later in life, often without formal treatment. That is not an argument against treatment; it is an argument against fatalism.

What's next

• Fentanyl and xylazine: the immediate crisis. Higher-dose naloxone formulations, drug checking, and treatment models designed for a supply that is far more potent and less predictable than heroin ever was.
• GLP-1 receptor agonists: observational and early trial data suggest reduced alcohol and nicotine consumption in people taking semaglutide and related drugs, a genuinely intriguing signal that is not yet established.
• Anti-drug vaccines and monoclonal antibodies, which bind the drug in the bloodstream so it cannot reach the brain, in trials for fentanyl and methamphetamine.
• Psychedelic-assisted therapy for alcohol and tobacco addiction, in early trials with the same blinding caveats as elsewhere.
• Scaling what works: the largest available gains are not new technologies. They are wider access to buprenorphine and methadone, universal naloxone distribution, contingency management for stimulants, and alcohol pricing and availability policy, all of which are established and underused.

Sources and notes

Tobacco mortality (over 7 million from direct use, over a million from secondhand smoke) and alcohol mortality (approximately 2.6 million deaths, 4.7 percent of all deaths) are WHO figures. Opioid agonist therapy mortality effect: Santo et al., JAMA Psychiatry, 2021, systematic review and meta-analysis of mortality during and out of treatment. Incentive sensitisation: Robinson and Berridge, Brain Research Reviews, 1993. Allostasis model: Koob and Le Moal, Science, 1997. Rat Park: Alexander et al., 1978, with subsequent methodological critique. Contingency management evidence: multiple meta-analyses, including for stimulant use disorder. E-cigarettes for cessation: Hajek et al., NEJM, 2019, and Cochrane reviews. Varenicline safety: EAGLES trial, The Lancet, 2016. Language and clinician attitudes: Kelly and Westerhoff, International Journal of Drug Policy, 2010. Portugal: Hughes and Stevens, British Journal of Criminology, 2010, and subsequent evaluations. Remission rates: Lopez-Quintero et al., Addiction, 2011, and national survey analyses. Deaths of despair: Case and Deaton, PNAS, 2015.

Open questions. Whether GLP-1 drugs genuinely reduce addictive behaviour is not established. Effective pharmacotherapy for stimulant and cannabis use disorders does not exist. How much of the "brain disease" model is explanatory versus rhetorical remains a live and useful argument.

Next: the disease that turns breathing, the most automatic thing you do, into work. 👉