What Is Harm Reduction, and Does It Work?

The short answer

Harm reduction is a term referring to a set of practices aimed at lowering the damage done by drug use without necessarily requiring the person to stop using. Examples include: Naloxone in a bag. Clean syringes. Methadone or buprenorphine. Fentanyl test strips.

Whether it works depends heavily on each specific intervention and its purpose, and the ranking is not objective or widely agreed upon. Medication for opioid addiction has the strongest evidence in the field: death rates during treatment run at roughly half the rate out of treatment, and the weeks right after someone stops are some of the most dangerous stretches in the whole timeline. In one Massachusetts study covering 2002 to 2009, communities that distributed take-home naloxone had lower opioid overdose death rates than communities that did not, though whether that community-level effect holds in the fentanyl era has not been tested at the same scale. Syringe programs combined with medication show weaker, but real evidence against hepatitis C. Supervised consumption sites reliably prevent the overdoses that happen inside them, while the newest reviews find no detectable signal at the province-wide level, a more complicated result than advocates usually concede. Harm reduction is also not always a substitute for other forms of treatment.

Finally: if you are making decisions about your own use or someone else's, this page is a starting point and not a clinical assessment. Talk to a clinician who specializes in addiction.

If someone is overdosing right now

This section is written for the worst ten minutes of someone's life. Read it before you need it.

  1. Call 911 first. Say the person is unresponsive and not breathing. Forty-eight states and DC have Good Samaritan laws covering people who call for an overdose, though what they protect varies and many do not cover outstanding warrants or probation violations. Call anyway. The alternative is a death.

  2. Give naloxone, even if you are not sure opioids are involved. Nasal spray into one nostril. It is harmless if you turn out to be wrong, and fentanyl now turns up in stimulants and counterfeit pills, so "it was just cocaine" is not a reason to skip it.

  3. If they are unresponsive and not breathing normally, start CPR. This is the part most overdose guidance leaves out. Current resuscitation guidance tells lay rescuers to begin chest compressions rather than stopping to check for a pulse. If you are trained, confident there is a pulse, and the problem is only that they are not breathing, rescue breaths alone are appropriate: head tilted back, nose pinched, one breath every five seconds. Oxygen is what is actually running out.

  4. No response after two to three minutes? Give a second dose in the other nostril. Fentanyl overdoses frequently need more than one.

  5. Do not leave, and do not let them use again. Naloxone wears off in roughly thirty to ninety minutes and the person can stop breathing a second time. They may also wake into sudden, brutal withdrawal and want to use immediately, which is one of the most common ways somebody dies after being saved. Stay until EMS arrives. Put them on their side.

  6. Know what naloxone will not do on its own. It acts on opioids. It will not by itself resolve a stimulant, benzodiazepine, or alcohol overdose, and it does not reverse xylazine or medetomidine, the veterinary sedatives now widely mixed into the illicit supply. So someone can stay sedated after a correct dose. That is not naloxone failure. Keep breathing for them and keep the ambulance coming.

Where to get naloxone: it is legally available without a prescription in every US state, but pharmacy stocking and out-of-pocket cost vary a great deal, so call ahead rather than assuming. Local health departments and syringe service programs frequently give it away free, and that is usually the faster route.

What the evidence actually says

Most arguments about harm reduction turn out to be arguments about the word, so start there. Harm reduction names a category of interventions rather than a position on whether people should use drugs: things that keep people who are using right now alive and out of the hospital, without making abstinence the price of entry. Some of those interventions are also treatments. Methadone is harm reduction and medicine. That overlap is where most of the public confusion lives.

The strongest evidence in the field is about staying alive on medication. The largest synthesis is Thomas Santo and colleagues' 2021 systematic review and meta-analysis in JAMA Psychiatry, pooling 15 randomized trials covering 3,852 participants and 36 cohort studies covering 749,634 participants (DOI: 10.1001/jamapsychiatry.2021.0976). Across the cohorts, the all-cause mortality rate during opioid agonist treatment was about half the rate during time off it (rate ratio 0.47, 95% CI 0.42 to 0.53), and the association held regardless of sex, age, geography, HIV or hepatitis C status, and whether people injected. Drug-related deaths specifically came in at 0.41, suicide at 0.48. Two numbers in that paper are the ones nobody in treatment gets told. During the first four weeks of methadone treatment, mortality ran at roughly double the rate seen later in treatment (2.01, 95% CI 1.55 to 5.09), with no evidence of a comparable spike for buprenorphine, though that estimate is imprecise (0.58, 95% CI 0.18 to 1.85). And in the four weeks after treatment stopped, all-cause mortality was six times higher than during treatment (6.01, 95% CI 4.32 to 8.36).

One of the most dangerous months in an opioid patient's year is the one right after they leave treatment - the thing that was keeping them alive - which is exactly the moment programs describe as graduation. The same mechanism, tolerance falling faster than circumstances change, makes release from jail or prison and discharge from detox comparably lethal windows. Santo's review found treatment associated with dramatically lower mortality both during incarceration and after release, which only makes sense against a post-release baseline risk that is itself extreme.

The earlier synthesis by Luis Sordo and colleagues in BMJ in 2017 pointed the same way with the raw rates attached: 11.3 all-cause deaths per 1,000 person-years in methadone treatment against 36.1 out of it, and overdose deaths of 2.6 against 12.7, across 19 cohorts and 122,885 people on methadone; buprenorphine ran 4.3 in treatment against 9.5 out (DOI: 10.1136/bmj.j1550). Both papers carry the same caveat, and I want to keep it attached to the numbers. These are overwhelmingly observational cohorts, not randomized trials, and the people who stay in treatment differ from the people who leave in ways statistical adjustment does not fully resolve. The size and consistency of the effect make confounding an unlikely full explanation. It does not make the comparison clean.

Naloxone in the hands of bystanders lowered community death rates. Alexander Walley and colleagues ran an interrupted time-series analysis in BMJ in 2013 across 19 Massachusetts communities, selected for having at least five fatal opioid overdoses a year, some of which implemented overdose education and nasal naloxone distribution and some of which did not (DOI: 10.1136/bmj.f174). Programs trained 2,912 potential bystanders, who went on to report 327 rescues. Compared with communities that implemented nothing, communities with modest program enrollment had an adjusted rate ratio for opioid overdose death of 0.73 (95% CI 0.57 to 0.91), and communities with high enrollment came in at 0.54 (0.39 to 0.76). Rates of acute care hospital use did not differ meaningfully. This is the finding people reach for when they argue naloxone "just enables," and it points the other way: the communities that put the antidote in more hands buried fewer people. It is a community-level comparison, so it tells you about places rather than individuals.

Syringe programs are where the evidence is genuinely mixed, and the reasons are worth stating precisely. Lucy Platt and colleagues did the Cochrane review in 2017, covering 28 studies of needle and syringe programs and opioid substitution therapy for preventing hepatitis C (DOI: 10.1002/14651858.CD012021.pub2). Current opioid substitution therapy was associated with a 50% reduction in hepatitis C acquisition risk (risk ratio 0.50, 95% CI 0.40 to 0.63; 12 studies, 6,361 people), with essentially no disagreement between studies. High syringe-program coverage on its own did not reach significance overall (0.79, 95% CI 0.39 to 1.61), and the studies disagreed sharply. Split by region, the two European studies (2,903 people) showed a large reduction (0.24, 95% CI 0.09 to 0.62), while the three North American studies (627 people) produced a point estimate of 1.25 (0.63 to 2.46), numerically pointing the wrong way. Combining both interventions was associated with a 74% reduction (0.26, 95% CI 0.07 to 0.89), from 3 studies. Cochrane graded the opioid substitution evidence low quality and the syringe-program evidence very low.

It would be convenient to explain the European–North American split as "America underfunds syringe access." The review does not read it that way. Platt and colleagues attribute the North American result to a mixture of causes without ranking them: confounding, since people who attend syringe programs tend to report higher injecting risk than those who don't, which biases against finding a benefit; differences in injecting patterns, with more stimulant injecting in North America, which removes the protective contribution that opioid substitution therapy would otherwise make; selection bias; and misclassification of exposure. That last one is worth spelling out. The European studies measured the proportion of a person's injections actually covered by a sterile syringe, while the North American ones mostly measured how often someone showed up. Those are different exposures wearing the same label. Thin US coverage appears on their list too, as one possibility among several rather than the explanation. I am laboring this because the subgroup finding is very-low-quality and post-hoc, and it gets quoted in both directions by people who have not read past the abstract. Note also that the well-established case for syringe programs preventing HIV is a separate literature from this hepatitis C review, and the two should not be quoted interchangeably.

Supervised consumption sites: strong evidence inside the room, weak evidence at the population level. Timothy Levengood and colleagues published a systematic review in the American Journal of Preventive Medicine in 2021, covering 22 studies, 16 of them on a single facility in Vancouver, with a literature search that closed in September 2019 (DOI: 10.1016/j.amepre.2021.04.017). They could not pool the results at all, because the studies measured different things. What they could report is the direction of findings: mostly significant reductions in overdose morbidity and mortality, improved injection practices, better access to addiction treatment, and no increase in crime or public nuisance in surrounding neighborhoods through that date. That last finding has since been contested by newer analyses of neighborhood crime around Canadian sites, so treat it as the state of the evidence up to 2019 rather than a closed question.

The most-quoted single study is Brandon Marshall's 2011 Lancet analysis of the Vancouver site, where the fatal overdose rate within 500 metres fell 35%, from 253.8 to 165.1 deaths per 100,000 person-years, against a 9.3% decline in the rest of the city (Lancet 2011;377:1429–37, PubMed 21497898). Read the comparator rates before you read the percentages: the rest of Vancouver went from 7.6 to 6.9 per 100,000 person-years. The neighborhood around the site had a baseline overdose death rate more than thirty times the rest of the city, so this is not a like-for-like contrast. The within-500-meter result was also marginal (p=0.048), and it drew published correspondence in Lancet from Drug Free Australia and the Southern Cross Bioethics Institute, advocacy organizations rather than independent methodologists, disputing the choice of baseline years and unaccounted-for policing changes in the same area over the same window. The authors published a reply. I mention the exchange because it is real, and the affiliation because it changes how much weight it carries.

The picture has moved since. Geneviève Gariépy and colleagues at the Public Health Agency of Canada published a systematic review in 2025 restricted to the post-2016 fentanyl era, asking specifically whether supervised consumption sites move population-level overdose mortality (DOI: 10.24095/hpcdp.45.9.02). Six studies met criteria, all Canadian, and the results split four ways. Of the four quasi-experimental studies, the two large province-wide analyses found no significant association between areas with and without sites and overdose mortality, while some analyses of smaller urban areas did find protective associations, inconsistently. Two further observational studies suggested lower mortality, with methodological limitations the authors flag. The review has since carried a corrigendum adding an important qualification about the Toronto analysis, whose geographically weighted models found associations between proximity to a site and lower overdose mortality that strengthened over time, adjusted for neighborhood characteristics.

Their conclusion is careful, and I think correct: these sites reliably prevent the deaths that would otherwise happen in them and reliably connect people to services, and at the scale of a province, with the coverage actually in place, that has not yet shown up as a detectable mortality signal. Anyone claiming these sites are proven to lower a city's death rate is ahead of the evidence. So is anyone claiming they are useless.

Drug checking has moved fast and the evidence has not caught up with the deployment. Fentanyl test strips and, increasingly, xylazine strips and spectrometry-based checking services are now among the most widely distributed harm-reduction tools in North America. The mechanism is straightforward and the observational data on behaviour change is reasonably encouraging: people who get a positive result often use less at a time, use more slowly, or use with someone present. What does not yet exist is good evidence that drug checking lowers mortality at a population level, partly because when nearly the entire supply is contaminated a positive result stops carrying much information. Treat it as sensible and low-cost, not as proven protection.

And underneath all of it, an argument about what counts as success. A 2026 individual-participant meta-analysis in JAMA Psychiatry by Masoumeh Amin-Esmaeili and colleagues pooled 12 NIDA multisite randomized trials of medications for stimulant use disorder, 2,000 participants in total (DOI: 10.1001/jamapsychiatry.2026.1092). First, the medications mostly did not beat placebo, on either outcome. That is the honest finding, and I am not going to dress it up. Second, the result that matters here is a measurement one: 31.2% of participants met the trial's "reduced use" criterion while 13.3% reached abstinence. Three caveats have to travel with those numbers. The criterion was defined as moving from five or more using days a month down to one through four or to abstinence, so the larger figure contains the smaller one rather than describing a separate group. Reduced use was self-reported while abstinence was confirmed by urine testing, so part of the gap reflects how each outcome was ascertained rather than how much change occurred. And the constituent trials ran from 2001 to 2011, which means the data predate the current supply entirely. What survives all three is the authors' own conclusion: trials should evaluate a continuum of outcomes rather than abstinence alone.

Where the experts disagree

The abstinence-first critique, in its strongest form. The weak version is "harm reduction enables addiction," and it is easy to dismiss. The strong version is empirical and much harder: harm-reduction expansion has not reliably produced the population-level mortality improvements its advocates predicted. The Gariépy review above is the cleanest example, and it comes from inside the harm-reduction literature rather than from its opponents. A critic can fairly say: you promised fewer deaths at scale, and at scale the deaths have not clearly fallen.

Denying that is the worst available answer. Get specific instead about what each intervention was ever going to do. Naloxone reverses an overdose in progress. A consumption site prevents the deaths that happen in it. Neither was ever a plan for moving a province's mortality on its own, and defending them as though they were is how you end up committed to a claim the evidence will not carry. It is also worth separating harm reduction from drug policy, which the public argument fuses constantly. Decriminalization is a legal regime, not a clinical intervention, and studies of it are testing something else entirely. The budget version of the critique stands on its own too: a person maintained on methadone for a decade with no counseling and no housing help, and nowhere to go next, has been kept alive by a system that then declared victory. The answer to that is not to withdraw the methadone.

The evidence-centrist position. A large group of addiction researchers strongly support the interventions with the best data, medication and naloxone, while pushing back on advocacy that outruns the evidence, particularly around consumption sites. That is a defensible place to stand, and it is where the Gariépy review's willingness to publish null findings becomes a feature rather than an embarrassment.

Carl Hart goes considerably further than any of them. In Drug Use for Grown-Ups and across his research, Hart argues that most drug use is not addiction, that adults have a right to alter their own consciousness, and that much of what we attribute to drugs is produced by prohibition and poverty. His endpoint is not decriminalization, which he has argued does not go far enough, but legal regulated markets. Whether you find that persuasive or alarming, it is a different kind of claim from the public-health one, and the two get conflated constantly.

Gabor Maté and the compassion frame. Maté was arguing that people who use drugs deserve care rather than punishment long before that was a mainstream clinical position, and he did it working in Vancouver's Downtown Eastside while the debate was still theoretical for most of the field. His written position on causes is more careful than his critics usually allow. In In the Realm of Hungry Ghosts he writes: "Not all addictions are rooted in abuse or trauma, but I do believe they can all be traced to painful experience." His public claims tend to run stronger than that sentence, which is what his critics are usually responding to, so it is only fair to make the criticism against the stronger version. And it lands there: "painful experience," defined that broadly, is difficult to test against any evidence that could come out the other way. On harm reduction specifically, his moral case and the epidemiological case point in the same direction.

And a disagreement that is mostly about words. A meaningful share of this fight is people using "harm reduction" to mean different things: some meaning naloxone and syringes, others meaning a stance that abstinence should never be proposed as a goal. Few clinicians hold the second position, though some drug-user advocacy organizations come close to it, and critics tend to argue against that version regardless of who is in front of them.

My take: harm reduction isn't a philosophy, it's a stage

I spent years using in ways that could easily have killed me. I did not stop because someone finally found the right argument. I stopped, eventually, after a lot of things had to go right, and every one of them required me to still be here.

That is the whole case, and it is less ideological than the debate makes it sound. In the SPARO lens I use - Stimulus, Perception, Activation, Response, Outcome - what happens before the use is actually the most important intervention target. The deepest treatment works upstream: change what the Stimulus means to you, change how you Perceive it, learn to control and redirect your somatic and emotional Activation, build a different Response. Good work. Slow work. Harm reduction operates almost entirely at the far end, on the Outcome, and it makes one modest promise: this particular use episode will not be the one that ends the story. It buys the time that upstream work needs.

Which is why "harm reduction versus abstinence" is a badly posed question. It sets a stage against a destination. Somebody carrying naloxone this month can be six months into a real change process by spring, and the naloxone is part of why they got the chance.

The EAT process I use runs Explore, Accept, Transform, and Accept is the stage most people skip. People hear "accept" and think surrender. Accepting where you actually are, including that you are still using, is what gives the Transform step real information to work with. A person who has to perform abstinence to stay in a program is a person whose program is now running on bad data.

I want to be disciplined about what I take from the newest evidence, because the temptation to overclaim runs in my direction too. The JAMA Psychiatry reduced-use paper shows one thing: a scoring rule can make a large group of people who changed disappear from a results table. It says nothing about whether cutting down produces better health than not cutting down. That is narrower than I would like, and it is still worth having. I have spent a decade arguing that our definition of success in this field is too small to describe what actually happens to people. Watching that argument turn up as a methodological finding in a psychiatry journal is a strange kind of vindication.

And the thing the harm-reduction movement itself sometimes underplays. Staying alive is necessary. It is not the goal. If reduction is the most anyone ever offers you, go and ask someone else.

What to actually do

✅ If anyone in your house uses opioids, get naloxone and read the response steps above. It is legally available without a prescription in every US state, and usually free from local health departments and syringe programs. Keep it where someone else can find it, not locked in your room.

✅ If opioids are the problem, ask about buprenorphine or methadone before anything else. This is the intervention with the strongest survival data in the field. Two practicalities. In the US, buprenorphine can be prescribed by any DEA-registered clinician, including your regular doctor, while methadone for opioid use disorder can only be dispensed through a federally certified opioid treatment program, usually with daily in-person dosing. And starting buprenorphine while fentanyl is still in your system can trigger sudden, severe precipitated withdrawal. There are established low-dose and timed-induction protocols for exactly this, so make it a conversation with a prescriber rather than something you improvise. If a program tells you medication is "trading one addiction for another," that program is decades behind the evidence.

✅ Treat leaving treatment as the dangerous moment it is. Mortality in the four weeks after stopping opioid agonist treatment runs six times higher than during it. If you are tapering off or being discharged, plan the exit with a clinician, and make sure naloxone is in the house before the last dose, not after.

✅ Never use alone. Most fatal overdoses happen with nobody there to respond. Have someone with you, or call a phone-based overdose response line such as Never Use Alone (1-800-484-3731) in the US, which stays on the line and sends help if you stop responding. Supervised consumption sites are not a realistic option for most US readers; only a handful operate and they remain federally contested.

✅ Assume everything is contaminated, and never mix downers. Fentanyl, xylazine, and medetomidine turn up across the illicit supply, including in pills and in stimulants. Test strips are cheap and worth using. And combining opioids with benzodiazepines or alcohol is one of the most reliable ways to stop breathing, which makes "just one drink on top" a genuinely different risk category.

✅ If you drink heavily or use benzodiazepines, do not stop suddenly on your own. Alcohol and benzodiazepine withdrawal can cause seizures and can be fatal. This is one of the few places where the abstinence instinct is actively dangerous. Ask a clinician about a supervised taper.

✅ Set a reduction goal you can actually measure, and count it as progress. Fewer days, smaller amounts, never before noon, never alone. Write it down. Fewer using days mean fewer chances to overdose, which is a straightforward arithmetic benefit. My clinical view, and I want to be clear that it is a view rather than a trial result, is that people who are allowed to count reduction as progress stay in contact with help instead of vanishing until the next crisis.

✅ If a program requires you to be abstinent before it will help you, keep looking. Options that meet people mid-use exist, including medication-based care, moderation-oriented programs, and non-abstinence coaching programs such as the one I founded at IGNTD. None of these is right for everyone. The point is that a door you have to already be well to walk through is not a door.

FAQ

Does harm reduction encourage more drug use? Be precise about what has actually been measured. The specific harms people predict have not materialised: seven studies in the supervised-consumption review examined crime and public nuisance and found no increase through 2019 (DOI: 10.1016/j.amepre.2021.04.017), and Massachusetts communities that distributed naloxone had lower overdose death rates than those that did not (DOI: 10.1136/bmj.f174). Neither of those outcomes is drug consumption, though, and whether these programs change how much or how often people use is much less well studied than the argument about them would suggest. What can be said is that the predicted catastrophes have not shown up.

Is methadone or buprenorphine really "recovery," or just substituting one drug for another? Pharmacologically these medications act on the same receptor system, which is where the intuition comes from. Functionally they are nothing alike: stable dosing does not produce the intoxication, compulsive seeking, or escalating harm that defines addiction, and mortality during treatment runs at roughly half the rate seen off it, with drug-related deaths lower still (DOI: 10.1001/jamapsychiatry.2021.0976). Insulin is also a daily drug you do not stop taking.

Do supervised consumption sites actually save lives? Inside the room, the record is strong: no overdose death has been reported at a sanctioned Canadian supervised consumption site, though that figure depends on facility reporting. At the scale of a city or province the evidence is genuinely unsettled. A 2025 Public Health Agency of Canada review of the post-2016 period found large province-wide analyses detecting no significant association with overdose mortality, some smaller urban analyses finding protective associations, and two observational studies suggesting lower mortality with methodological limitations (DOI: 10.24095/hpcdp.45.9.02). Both halves are true at once. Most public arguments only quote the half they like.

Do syringe exchange programs actually reduce disease? For hepatitis C, the Cochrane review found opioid substitution therapy reduced acquisition risk by 50% and the two interventions combined by 74%, while syringe programs alone gave inconsistent results across regions (DOI: 10.1002/14651858.CD012021.pub2). The review attributes much of that inconsistency to how "coverage" was measured rather than to real geographic differences. The separate evidence that syringe access prevents HIV is stronger and should not be confused with this.

If I am not ready to quit, is there any point in trying to cut back? Yes, for two different reasons that are worth keeping apart. The practical one: fewer using days and smaller amounts mean fewer chances to overdose, and people who set reduction goals tend to stay connected to care instead of vanishing until a crisis. The research one is narrower than it is usually reported. A 2026 JAMA Psychiatry meta-analysis showed that scoring trials by abstinence alone erases most of the movement that occurs, with 31.2% meeting a reduced-use criterion against 13.3% reaching abstinence (DOI: 10.1001/jamapsychiatry.2026.1092). That is a finding about how we measure change, not proof that cutting back beats quitting for any particular person.

Related in the Behavior Change Atlas

If you need help right now

If you or someone you love is struggling with drug use, help is available: call or text the SAMHSA National Helpline at 1-800-662-4357 (free, confidential, 24/7), or find treatment at findtreatment.gov. If you are in emotional crisis or having thoughts of harming yourself, reach the 988 Suicide & Crisis Lifeline (call or text 988). If you are about to use and there is nobody with you, the Never Use Alone line is 1-800-484-3731. If you are witnessing an overdose, call emergency services first, then follow the response steps near the top of this page.

References

All citations verified against PubMed via the PubMed MCP tools at draft time (2026-08-03).

  1. Santo T, Clark B, Hickman M, Grebely J, Campbell G, Sordo L, Chen A, Tran LT, Bharat C, Padmanathan P, Cousins G, Dupouy J, Kelty E, Muga R, Nosyk B, Min J, Pavarin R, Farrell M, Degenhardt L. Association of opioid agonist treatment with all-cause mortality and specific causes of death among people with opioid dependence: a systematic review and meta-analysis. JAMA Psychiatry. 2021;78(9):979–993. DOI: 10.1001/jamapsychiatry.2021.0976 (PMID: 34076676)

  2. Sordo L, Barrio G, Bravo MJ, Indave BI, Degenhardt L, Wiessing L, Ferri M, Pastor-Barriuso R. Mortality risk during and after opioid substitution treatment: systematic review and meta-analysis of cohort studies. BMJ. 2017;357:j1550. DOI: 10.1136/bmj.j1550 (PMID: 28446428)

  3. Walley AY, Xuan Z, Hackman HH, Quinn E, Doe-Simkins M, Sorensen-Alawad A, Ruiz S, Ozonoff A. Opioid overdose rates and implementation of overdose education and nasal naloxone distribution in Massachusetts: interrupted time series analysis. BMJ. 2013;346:f174. DOI: 10.1136/bmj.f174 (PMID: 23372174)

  4. Platt L, Minozzi S, Reed J, Vickerman P, Hagan H, French C, Jordan A, Degenhardt L, Hope V, Hutchinson S, Maher L, Palmateer N, Taylor A, Bruneau J, Hickman M. Needle syringe programmes and opioid substitution therapy for preventing hepatitis C transmission in people who inject drugs. Cochrane Database Syst Rev. 2017;9(9):CD012021. DOI: 10.1002/14651858.CD012021.pub2 (PMID: 28922449)

  5. Levengood TW, Yoon GH, Davoust MJ, Ogden SN, Marshall BDL, Cahill SR, Bazzi AR. Supervised injection facilities as harm reduction: a systematic review. Am J Prev Med. 2021;61(5):738–749. DOI: 10.1016/j.amepre.2021.04.017 (PMID: 34218964)

  6. Marshall BDL, Milloy MJ, Wood E, Montaner JSG, Kerr T. Reduction in overdose mortality after the opening of North America's first medically supervised safer injecting facility: a retrospective population-based study. Lancet. 2011;377(9775):1429–1437. DOI: 10.1016/S0140-6736(10)62353-7 — PubMed 21497898

  7. Gari-py G, Prowse RKM, Plouffe R, Graham E. Supervised consumption sites and population-level overdose mortality: a systematic review of recent evidence, 2016–2024. Health Promot Chronic Dis Prev Can. 2025;45(9):357–366. DOI: 10.24095/hpcdp.45.9.02 (PMID: 40960731). See corrigendum, Health Promot Chronic Dis Prev Can. 2026;46(2):73, doi 10.24095/hpcdp.46.2.04 (PMID 41671413), adding a clarification about the Toronto analysis.

  8. Amin-Esmaeili M, Farokhnia M, Mojtabai R, Leggio L, Johnson RM, Susukida R. Evaluating reduced use and abstinence as outcomes in pharmacotherapy trials for stimulant use disorder: a meta-analysis of 12 randomized controlled trials. JAMA Psychiatry. 2026. DOI: 10.1001/jamapsychiatry.2026.1092 (PMID: 42234418)

Limitations stated honestly: Almost none of the core harm-reduction evidence comes from randomized trials, because randomizing people to withheld naloxone or withheld medication is not ethically appropriate. The mortality findings therefore rest on cohort designs, where people who stay in treatment differ from those who leave in ways adjustment cannot fully resolve. Cochrane graded the opioid substitution evidence for hepatitis C as low quality and the syringe-program evidence as very low, and the regional split within it is a post-hoc subgroup analysis built on two and three studies respectively. The supervised consumption literature could not be pooled at all, is dominated by one Vancouver facility, and the widely quoted Marshall analysis is a before-and-after comparison in a single neighbourhood that drew published methodological criticism. The 2025 review finding no province-wide mortality signal is limited to six Canadian studies. Drug-checking is discussed without a citation because the deployment has outrun the peer-reviewed outcome literature, and it is labelled as such. Rate ratios describe populations rather than individuals. None of this substitutes for an individual clinical assessment. Where the evidence is weakest, I have tried to say so rather than round up.

Disclosure: I am the founder of IGNTD, which offers non-abstinence-based recovery programs and which I mention above as one option among several. I have a financial interest in it. I have no financial relationship with any of the researchers cited or with any other intervention discussed on this page.

Written by Dr. Adi Jaffe, PhD (UCLA), author of The Abstinence Myth and Unhooked. Last medically reviewed: 8-8-26