The treatments — what actually works, what doesn't, and why the loudest part of the industry has the weakest evidence

Canadian Managed Alcohol Programs — measured pours, in housing

The premise is operant. Chronic, street-entrenched alcoholics are not going to stop drinking. The choice is regulated wine at measured intervals inside housing, or hand sanitiser, mouthwash and Listerine in a doorway. The Canadian model picks the first.

The flagship research programme is the Canadian Managed Alcohol Program Study (CMAPS), run from the Canadian Institute for Substance Use Research at the University of Victoria by Tim Stockwell, Bernie Pauly and colleagues since 2013. The model: regulated, measured beverage-alcohol doses (typically wine, every 60–90 minutes) inside a housing-first or shelter setting, alongside meals, medical care, social work, and a route to treatment if the resident wants one. Sip, not abstain.

Outcomes from the 2018 multi-site comparison of 175 MAP residents at six Canadian sites versus 189 matched controls (Stockwell et al., Drug & Alcohol Review, 2018): MAP participants drank less overall, switched away from non-beverage alcohol (mouthwash, hand sanitiser), and had fewer alcohol-related harms. A linked paper found MAP residents coped better when alcohol became unaffordable, with fewer collapses into non-beverage substitutes. The 2022 scoping review in Harm Reduction Journal consolidated the picture.

Toronto's Annex at Seaton House is the original — Canada's first MAP, opened 1996. Operator data summarised at a CMAPS webinar in 2015 claims a 93 per cent reduction in emergency-department visits and roughly CAD $3,300 per day per client in avoided public costs across ER, police and ambulance. Read those as programme self-report, not RCT — but consistent with the wider CMAPS evidence base. BC's Centre on Substance Use has issued operational guidance so the model can be replicated without amateur improvisation.

The caveats Pauly and Stockwell put on the record themselves: selection bias favouring MAP enrollees cannot be ruled out; some participants still die — these are end-stage alcoholics; the point is not abstinence, it is fewer ambulance calls, fewer assaults, fewer rough-sleeping deaths, and a door back into the medical system.

Portugal — the 2001 decriminalisation arc

Often misreported as legalisation. It isn't. Portugal's 2001 reform decriminalised personal-use possession of all illicit drugs — heroin, cocaine, cannabis, MDMA — making it an administrative offence akin to a parking ticket. Anyone caught is referred to a Comissão para a Dissuasão da Toxicodependência (CDT) — a three-person panel (lawyer, doctor or social worker, plus a chair) that can warn, fine, or — most often — route into treatment.

The carrot. Treatment is voluntary, free at point of use, and was — at peak funding — generously resourced. The canonical paper is Hughes & Stevens, British Journal of Criminology, 2010. Findings:

• No major increase in overall drug use post-reform (small rises in some categories, falls in others)
• Reductions in problematic use, drug-related harms and criminal-justice overcrowding
• HIV diagnoses among injecting drug users fell dramatically
• Cost savings on enforcement offset increased spend on treatment and harm reduction

The same authors' 2012 re-examination in BJC was more nuanced — neither resounding success nor disastrous failure, but a meaningful net positive when the system was funded. A 20-year retrospective in SATPP, 2021 tracks the same arc.

The lesson is not that decriminalisation failed — Portuguese drug deaths remain a small fraction of US per-capita rates — but that decriminalisation without funded treatment is a brittle policy. Transform 2024 and UNSW 2024 make the same point. The carrot needs to be funded. The stick on its own is just the war on drugs in a softer voice.

Alcohol — RCT evidence

Project MATCH (NIAAA, 1997). 1,726 clients randomised to CBT, Motivational Enhancement Therapy, or Twelve-Step Facilitation across two arms. Headline finding: all three produced significant, sustained improvement, with few clinically significant differences between them. The matching hypotheses — that specific patient types would respond better to specific modalities — largely failed. NIAAA release; posttreatment paper.

Naltrexone and acamprosate. Maisel et al., Addiction, 2013 — 64 RCTs, ~11,000 participants. Both significantly outperform placebo with modest effect sizes. Number needed to treat for an additional abstinent case is approximately 8 for acamprosate and 9 for naltrexone. Acamprosate better for abstinence; naltrexone better for heavy-drinking reduction and craving.

The Sinclair Method — targeted naltrexone taken 1–2 hours before drinking, building pharmacological extinction. Mechanism plausible, several smaller targeted-naltrexone RCTs supportive, but no large dedicated RCT of TSM as a complete protocol has been published.

AA and 12-Step Facilitation — the surprise. Kelly et al., Cochrane Review, 2020 — 27 studies, 10,565 participants, 21 RCT or quasi-RCT. Manualised TSF produced higher rates of continuous abstinence than CBT at follow-up, with high-certainty evidence on the abstinence outcome and equivalence on drinking-intensity outcomes — at substantially lower cost. This genuinely surprised the field. The cheap, lay-led, peer-driven model outperformed the professional one on the primary outcome.

Opioids — agonist therapy is the strongest evidence base in addiction medicine

Methadone and buprenorphine. Sordo et al., BMJ, 2017 pooled 19 cohorts (122,885 methadone-treated, 15,831 buprenorphine-treated). All-cause mortality rate: 11.3 versus 36.1 per 1,000 person-years in versus out of methadone treatment — rate ratio ~3.2. Buprenorphine: 4.3 versus 9.5 — rate ratio ~2.2. Overdose mortality drops are similar magnitude. Treatment retention saves lives. The EMCDDA grades this as best-practice, evidence-graded A.

Extended-release naltrexone (Vivitrol). Lee et al., NEJM, 2016 — XR-NTX reduced opioid relapse among criminal-justice-involved adults vs usual care. The comparative-effectiveness X:BOT trial found XR-NTX harder to initiate than buprenorphine-naloxone, equivalent in those who started. In practice it under-performs agonist therapy on retention and on mortality.

Heroin-Assisted Treatment (HAT). Six RCTs across Switzerland, the Netherlands, Germany, Spain, the UK, Canada (NAOMI). Supervised diacetylmorphine outperforms methadone for the methadone-refractory subgroup on illicit use, illegal income, retention. The Canadian SALOME trial showed injectable hydromorphone non-inferior to diacetylmorphine — clinically useful where pharmaceutical heroin is unavailable. EMCDDA HAT Insights monograph.

Stimulants — contingency management

No approved pharmacotherapy for cocaine or methamphetamine. Decades of trials, no agent has cleared regulatory bars.

Contingency management — incentives (vouchers, prize draws, cash) for drug-negative urine tests. Pioneered by Stephen Higgins (cocaine, 1990s) and Nancy Petry (prize-based, 2000s). Moderate-to-large effect sizes — the strongest behavioural evidence in stimulant treatment. AJP 2024 "Call to Action". Adoption is throttled by cultural distaste for "paying people to be sober" rather than weak evidence. The policy refusal to deploy CM at scale is one of the cleanest cases on this site of the willpower-frame producing demonstrably worse outcomes.

Obesity — surgery, CBT, drugs

Bariatric surgery delivers the largest sustained weight losses available. Roux-en-Y gastric bypass: weighted mean ~57% excess weight loss at 10+ years; sleeve gastrectomy ~58%; biliopancreatic diversion / duodenal switch ~74%. O'Brien et al. systematic review, 2018.

CBT for obesity — medium short-term effect sizes; most gains attenuate by 12 months (Behaviour Research & Therapy meta-analysis, 2020). Behavioural therapy alone is not a long-run obesity solution. The published long-run sustained-loss rates of behavioural-only interventions are in single digits at five years.

GLP-1 / GIP receptor agonists — the full case is on the Costs page. Roughly 20% weight loss at one year on tirzepatide; 19% all-cause mortality reduction extrapolated from SELECT; chronic suppressive therapy, regain on discontinuation. Closes about two-thirds of the gap to bariatric surgery without anaesthesia, scars or reoperation risk.

Compulsive Sexual Behaviour Disorder — what we don't yet know

ICD-11 impulse-control disorder, not addictive disorder. Treatment evidence base is thin. Briken et al.'s 2022 preregistered systematic review identified 24 studies, only 4 RCTs. Tentative support for CBT, very weak evidence overall. Journal of Behavioral Addictions, 2022. Honest answer: we don't yet know what works. Any treatment programme claiming high efficacy in this space is selling rather than evidencing.

The American rehab industry — loudest, weakest

Maia Szalavitz and Marc Lewis line up with much of the public-health community on this: most US 28-day residential rehab is not evidence-based; many therapeutic communities lack oversight; the "hit bottom" trope is fallacious; what consistently helps is treating people kindly, with dignity, with medication where indicated, and with access to housing, employment and community.

The Recovery Research Institute residential treatment review finds mixed and often weak evidence for the canonical 28-day abstinence-only model. The model continues to dominate the US treatment market because it is the model insurance reimburses and the model the recovery industry was built around — not because it has the best outcomes. This is a textbook case of the willpower-and-education frame producing market dominance for an intervention with thin evidence.

What works, in one table

The through-line. Substitution works (opioids). Harm reduction in place works (alcohol, housing-first). Decriminalisation works when paired with funded treatment (Portugal pre-austerity). Incentives work (stimulants). Cheap, manualised peer-support works better than expected (AA via TSF). The aggressively-marketed 28-day abstinence-only model is the part of the field with the weakest evidence and the loudest marketing. That is the asymmetry policymakers need to see.

The next page is the load-bearing one. It is about what happens to policy when you assume the willpower-and-education frame is sufficient. →