Why Do People Relapse?

The short answer

Two terms get used interchangeably and they shouldn't be. Being clear about terminology can help us make better choices. A lapse can describe a single episode of use after a period without it. A relapse is typically a more substantial return to the old behavioral pattern. What happens after the first drink matters more than most people expect: the reaction to a lapse is among the factors that push many toward a full relapse. People return to a compulsive behavior because it still works, faster and more reliably than anything else they have learned to do, and because the moments that call for it - stress, low mood, craving, exhaustion - keep showing up in their lives. The research points to those triggers, to treatments that measurably shift the odds, including medication, and to a well-described trap in which self-blame after a lapse collapses a person's self-confidence and creates shame that drives them to continue down the path. If you are reading this in the hours after a slip, skip to What to actually do. If you use opioids and have been off them for any stretch, read the safety box first: that gap has lowered your tolerance, and this is the most dangerous moment in the cycle.

What the evidence actually says

The studies below were checked on PubMed at draft time; DOIs and PubMed links go to each one.

Returning to use is common across chronic conditions. Thomas McLellan and colleagues set drug dependence side by side with type 2 diabetes, hypertension, and asthma in a 2000 review in JAMA. Heritability, personal choice, and environmental influenes were found to be comparably involved in driving relapse across all four, and medication adherence and relapse rates are similar across them (DOI: 10.1001/jama.284.13.1689). Their practical point still hasn't landed in most of the treatment world: addiction gets treated as an acute illness, a broken bone you fix once, when long-term monitoring and management produce more lasting benefit. Nobody says a person with asthma failed when they need their inhaler again in March.

The lapse-to-relapse trap has a name. In the cognitive-behavioral relapse-prevention model developed by Alan Marlatt, summarized by Mary Larimer, Rebekka Palmer, and Marlatt in Alcohol Research & Health. Four immediate determinants sit at the front of a relapse:

  1. high-risk situations,

  2. whether the person has a coping response ready,

  3. what they expect the substance to do for them, and

  4. the abstinence violation effect (PubMed: 10890810).

That last factor describes what happens when a lapse triggers self-blame and a collapse in the person's belief that they can do this at all, which pushes a single episode toward a full return. Lifestyle drift and unattended cravings set the stage for the downfall beforehand.

Two honest notes about this. The abstinence violation effect is one of four determinants, not the master switch, and the source is a narrative overview of a model rather than a trial, so treat it as a well-supported clinical framework rather than a settled empirical law. And the time windows I use below, the first hour, the first day, are my clinical rules of thumb. No study I know of has tested them.

Still. The dangerous thing about the first drink is not the alcohol in it. It's the internal dialogue that often follows it: well, I've blown it now. Might as well keep going.

Marlatt's model was expanded later by Marlatt and his team. In 2004, Katie Witkiewitz and Marlatt published a reconceptualization in American Psychologist responding to criticism that the original was too linear (DOI: 10.1037/0003-066X.59.4.224). They proposed treating relapse as a dynamic system, with multiple risk factors and situational triggers interacting and sometimes tipping suddenly rather than sliding down a ramp. It's a proposed model, and the authors close by discussing its limitations. But it fits what people report. "It came out of nowhere" is what an unstable system looks like from the inside.

Stress, mood, and craving show up in the prospective research. Rajita Sinha's 2011 review in Current Psychiatry Reports compiled prospective studies on predictors of return to use (DOI: 10.1007/s11920-011-0224-0). Depressive symptoms, stress, and drug craving predicted future relapse risk. So did biological measures: cortisol and the cortisol/ACTH ratio, serum brain-derived neurotrophic factor, atrophy in medial frontal regions, and hyperreactivity of the anterior cingulate during withdrawal. Sinha's argument, as of that 2011 review, was that they should be developed into clinical markers, which tells you they weren't being used as tests. The field is generally limited to literally measuring use, by self-report, as the only nuber that matters, which is an important hurdle to progress.

Aftercare shifts the odds, and the advantage shifts with it. Sarah Bowen, Witkiewitz and colleagues randomized 286 people who had completed treatment at a private facility to mindfulness-based relapse prevention, standard cognitive-behavioral relapse prevention, or treatment as usual, and followed them twelve months, in a trial published in JAMA Psychiatry (DOI: 10.1001/jamapsychiatry.2013.4546). At six months, both structured conditions beat treatment as usual on relapse risk, and among those who did use, on days of use and heavy drinking. Standard relapse prevention delayed time to first drug use relative to the mindfulness arm. By twelve months, only the mindfulness condition still held an advantage over both the other arms. Outcomes leaned on self-report alongside urinalysis, but they revealed that what you do after treatment is incrediby important for continued or sustained recovery.

Medication is the part most articles about relapse leave out

If you take one thing from this page and you or someone you love uses opioids, take this - Thomas Santo and colleagues published a systematic review and meta-analysis in JAMA Psychiatry in 2021 (DOI: 10.1001/jamapsychiatry.2021.0976). Pooling the 36 cohort studies in that review, the all-cause mortality rate while people were receiving opioid agonist treatment, methadone or buprenorphine, was less than half the rate during their time off it (RR = 0.47, 95% CI 0.42 to 0.53). Drug-related death, suicide, and cardiovascular death were all lower during treatment. This is a within-person comparison of time on versus time off, drawn from observational cohorts rather than the randomized trials the review also included, so people in treatment may differ from those out of it in ways the analysis can't fully account for.

Two numbers from the same paper belong beside that one. In the first four weeks of methadone treatment, all-cause mortality and drug-related poisoning ran about double the rate during the rest of treatment, an early-risk spike that did not appear for buprenorphine. And in the four weeks after stopping treatment, all-cause mortality was six times the rate during it. Starting is a risky transition. Stopping is a riskier one. Neither is an argument against the medication; both are arguments for staying close to whoever prescribes it.

Starting medication is a risky transition. Stopping it is a riskier one. But medication substantially reduces mortality risk while on board.

For alcohol, Melissa McPheeters and colleagues meta-analyzed 118 trials and 20,976 participants in JAMA in 2023 (DOI: 10.1001/jama.2023.19761). One of eleven patients treated with acamprosate remained sober beacuse of the medication (95% CI 1 to 32) and one out of eighteen (18) did for oral naltrexone at 50 mg/day (95% CI 4 to 32). One of eleven (11) for oral naltrexone were able to avoid a return to heavy drinking (95% CI 5 to 41), which is actually the primary claim for nalterxone utilization. Those intervals are wide, and the review's own conclusion is measured: alongside psychosocial treatment, oral naltrexone and acamprosate are the first-line options. Modest effects. Real ones. And almost nobody prescribes them.

Incentives move behavior, and then they stop. Contingency management pays people, in vouchers or prize draws, for drug-free samples. Lois Benishek and colleagues meta-analyzed 19 randomized studies of the prize-based version in Addiction and found an average end-of-treatment effect of Cohen d = 0.46 (95% CI 0.37 to 0.54) (DOI: 10.1111/add.12589). The effect shrank to d = 0.33 at follow-ups within three months, and at six months the pooled estimate was 0.09 in the opposite direction (95% CI −0.28 to 0.10, from six studies). The authors concluded the effect does not appear to persist to six months. The interval is wide, but its upper bound still rules out anything like the end-of-treatment effect.

Larger syntheses are more positive about the treatment itself. Brandon Bentzley and colleagues compared eleven categories of treatment for cocaine use disorder across 157 studies and 15,842 participants in JAMA Network Open, and contingency management was the only category significantly associated with a negative test for cocaine at end of treatment (OR = 2.13, 95% CI 1.62 to 2.80) (DOI: 10.1001/jamanetworkopen.2021.8049). Tom Ainscough and colleagues reviewed 22 studies of contingency management during treatment for opiate addiction and found it efficacious for most drug use, though notably not for opiate use itself, with follow-up data available in only three of the studies (DOI: 10.1016/j.drugalcdep.2017.05.028).

Put together: behavior responds to what the environment pays, far more than the "they just don't want it enough" story allows, and what happens after the payments stop is genuinely understudied.

The scoreboard itself distorts the picture. A 2026 individual-participant meta-analysis in JAMA Psychiatry by Masoumeh Amin-Esmaeili and colleagues pooled 12 NIDA-sponsored multisite randomized trials of medications for stimulant use disorder, run between 2001 and 2011, with 2,000 participants who all also received CBT (DOI: 10.1001/jamapsychiatry.2026.1092). The medications mostly did not beat placebo on either outcome, so nobody should read this as a drug endorsement. Now the numbers, and they need care, because the two categories are nested rather than side by side. 13.3% of participants achieved abstinence, while 31.2% managed to substantially reduce their use. Importantly, reduced use was defined as dropping from five or more days a month to one to four days or reaching abstinence, so the 13.3% who achieved abstinence sit inside the 31.2% who achieved reduced use. Roughly eighteen percent of participants improved without getting to abstinence - 50% more than those who were able to achieve abstinence. One caution the authors' design makes necessary: reduced use was self-reported, and abstinence was urine-verified, so some of that gap is measurement method rather than hidden progress. Even allowing for that, the authors' conclusion stands. Trials should evaluate a continuum of outcomes rather than a binary, and a person who went from daily to twice a month has moved, whatever the abstinence column says.

Many people do resolve a substance problem, and the timeline is long and uneven. John Kelly and colleagues surveyed a probability-based sample of US adults, reporting in Drug and Alcohol Dependence in 2017 that 9.1% of adults said they used to have a problem with alcohol or drugs and no longer do, which scales to tens of millions of people (DOI: 10.1016/j.drugalcdep.2017.09.028). That is a population prevalence, not a cure rate. Of those who resolved a problem, 46% identified as being "in recovery" and 53.9% had used an assisted pathway, mutual help most often, then treatment, then recovery support services. Gene Heyman's analysis of national epidemiological surveys in Annual Review of Clinical Psychology adds the time dimension: remission rates differ sharply by drug, with a half-life of about four years for cocaine dependence and about sixteen years for alcohol dependence, and each year a roughly constant proportion of those still addicted remit (DOI: 10.1146/annurev-clinpsy-032511-143041). Slow, uneven, and real.

⚠️ Safety box: the return is the dangerous part

Start here: the strongest protection is medication. Buprenorphine or methadone can often be started the same day at a low-threshold clinic, and they are associated with less than half the death rate compared with time off treatment. Call SAMHSA at 1-800-662-4357 (free, 24/7) or search findtreatment.gov. If you have been using fentanyl, say so to the prescriber, because starting buprenorphine too early can trigger severe withdrawal, and there are protocols for that.

Any gap in use, jail, treatment, a hospital stay, a determined dry month, lowers your tolerance while the memory of your old dose stays intact. Ingrid Binswanger and colleagues followed 30,237 people released from Washington State prisons and reported in The New England Journal of Medicine that in the first two weeks after release the risk of death was 12.7 times that of other state residents, with a relative risk of overdose death of 129 (DOI: 10.1056/NEJMsa064115). That cohort is specific to incarceration and predates the fentanyl era. But the physiology generalizes to any post-abstinence return, and today's supply makes it worse: with fentanyl and nitazenes present in unpredictable amounts, there is no dose you can titrate to safety on your own.

What an opioid overdose looks like: unresponsive to a hard sternal rub or shouting; breathing very slow, gurgling, snoring-like, or stopped; lips and fingertips turning blue or grey; pupils shrunk to pinpoints. Do not assume someone is just nodding off.

If you are returning to opioid use, or you are with someone who is:

  • Don't use alone. If nobody can be present, the Never Use Alone line (1-800-484-3731) will stay on the phone with you.

  • Have naloxone in the room and make sure the other person knows how to use it. It is available over the counter at pharmacies and free from most harm-reduction programs.

  • If someone overdoses: call 911 first, give naloxone, give rescue breaths, and repeat naloxone every two to three minutes if there is no response. Fentanyl and nitazene overdoses often need more than one dose.

  • Stay with them even after they wake up. Naloxone wears off before the opioid does, and people go back down. Do not let them refuse an ambulance because they feel fine.

  • Naloxone does not reverse xylazine or medetomidine, which are now common in the supply. If breathing does not come back after naloxone, that may be why, which is another reason 911 is not optional.

  • A small test dose lowers risk. It does not make a dose safe. It only counts alongside naloxone and someone present.

  • Fentanyl test strips help, but a negative result is not a clean result: they miss most nitazenes and can miss fentanyl that is unevenly mixed. Treat every negative as if it were positive.

  • Most US states have Good Samaritan laws, but they vary, and many do not cover parole or probation violations or outstanding warrants. Call anyway, and look up your own state's law before you need it.

This article is not enough if: you are physically dependent on alcohol or benzodiazepines (stopping abruptly can cause seizures and can be fatal, so a supervised taper is a medical matter), you have overdosed before, you are pregnant, or you are the parent or partner of someone using and don't know what to do. Each of those needs a person, not a page. If you are having thoughts of harming yourself, call or text 988 (Suicide & Crisis Lifeline) now, or go to an emergency room if you have a plan or means or can't keep yourself safe.

Where the experts disagree

Inside the chronic-care tradition. Researchers who treat addiction as a chronic condition needing long-term management, McLellan among them, removed a great deal of blame from the field. The criticism that "chronic relapsing brain disease" can become a self-fulfilling story is real, and it comes from two directions. Some of it is internal: McLellan and Kelly both sit within this tradition and have themselves pushed back against the fatalistic reading. Some of it is external and sharper, from Hart, Heyman, Marc Lewis, and others who reject the disease framing outright. Both critiques exist, and the internal one does not dispose of the external one.

The choice and behavioral-economics view. Carl Hart's laboratory work, with six experienced cocaine smokers choosing between smoked cocaine and a $5 voucher, found that they often took the voucher, and that the effect was dose-dependent: as the cocaine dose rose, choosing the drug rose with it, and money vouchers competed better than merchandise vouchers (DOI: 10.1097/00008877-200002000-00010). Here’s the takeaway: Drug use responds to the alternatives offered, and it responds less flexibly as the dose gets bigger. Gene Heyman's position is distinct from, and often conflated with, Hart's. His argument concerns the pattern of remission over time and the local-versus-global structure of choice, not incentives in a lab. Critics of both note that choosing a voucher in a controlled setting says little about choice at three in the morning in withdrawal, and that this framing has been misused to justify punishment. Both objections are fair. What survives is an argument for building better alternatives and making those alternatives more available.

The learning and neuroplasticity view. Marc Lewis and others argue addiction is better understood as deeply entrenched learning, motivated repetition that carves a groove, rather than as pathology. On that reading, a return to use is a well-learned response reasserting itself under familiar conditions, which is closer to how relapse actually feels and further from the language of disease.

The trauma and dysphoria views. Gabor Maté locates the driver in early pain, and Anna Lembke emphasizes a homeostatic swing toward dysphoria after heavy reward, which makes the period after quitting feel unbearable and the return predictable. Both name something real about why the moment is so hard. The disagreements concern how much of the whole picture each one explains, which I address in detail in the trauma piece linked below.

Twelve-step practice. The counting convention, in which a slip resets a sobriety date, is the aspect most often criticized for making a lapse feel like erasure. It’s worth being precise here: that's a fellowship custom and how it gets interpreted, not doctrine, and AA's own literature does not teach that a slip destroys the work. And the outcome evidence for twelve-step facilitation is stronger than its critics usually acknowledge. A 2020 Cochrane review by John Kelly, Keith Humphreys, and Marica Ferri covering 27 studies and 10,565 participants found manualized AA/twelve-step facilitation more effective than other established treatments, including CBT, for continuous abstinence at twelve months. (DOI: 10.1002/14651858.CD012880.pub2). What varies by person is whether the surrounding culture lands as steadying or corrosive.

My take: relapse lives in one link of the chain

I relapsed. More than once. And what I remember is the twenty minutes after, when I decided what the slip meant about me. That decision did more damage than the substance did.

Here is how I hold it now, using the SPARO lens I use with clients. Every episode runs Stimulus, Perception, Activation, Response, Outcome. A Stimulus arrives: a fight, a bonus, a song, three bad nights. Perception decides what it means. Activation is the feeling in the body. Response is what you reach for. Outcome is the relief, which teaches the chain to run faster next time.

Almost everyone who has relapsed tries to prevent the next one by attacking the Stimulus. Avoid the bar. Delete the number. Move cities. That works until life delivers a stimulus you can't avoid, which it will, because the list includes grief, promotions, and Tuesdays. My reading of the prospective predictors, stress, low mood, and craving, is that they point at Activation, the feeling. That's an interpretation of correlational findings, not something Sinha or anyone else set out to test. But it locates the work somewhere useful, because the link you fully own is Response.

So the question stops being "how do I never feel this again" and becomes "what else can answer this feeling, quickly enough to compete." A substance delivers almost immediately. A twenty-minute breathing practice loses that race unless you've practiced it while calm. That's what my EAT process is for at the break point: Explore what the moment actually was, Accept the feeling without arguing with it, Transform the Response. Run it until the new Response is quick. Once through a workbook won't get you there. But that’s why we have many exercises and versions of running through that EAT loop.

Then the part I care most about, because it's where the abstinence violation effect lives. A lapse is data. It's the most specific information you'll ever get about which link in your chain is weak, and it arrives with the details still warm. People who treat it as data tend to get better. People who treat it as a verdict on their worth reach for the thing that turns off the feeling of being a person who just proved they're worthless. That loop is why shame accelerates the exact thing everyone thinks it prevents, and it's the argument at the center of The Abstinence Myth.

One change in how you keep score, offered as clinical opinion rather than evidence: instead of asking whether you stayed perfect, watch four things. Are the episodes further apart? Shorter? Less severe? Do you come back faster? In my practice those move well before "days sober" looks impressive. I'm not aware of trial data validating them as predictors, and I'd like someone to test them.

What to actually do

(A note on format: the ✅ items below are the Atlas house style for action lists.)

✅ Ask a clinician about medication, especially with opioids or alcohol. Buprenorphine and methadone are associated with less than half the death rate compared with time off treatment. For alcohol, acamprosate and oral naltrexone are the first-line options, with modest but real effects; injectable naltrexone has a thinner evidence base for preventing a return to drinking. If you've been using fentanyl, tell the prescriber before starting buprenorphine. This is the most evidence-backed item on the list and the one most often skipped.

✅ Write your lapse plan before you need it. One page, written on a good day: who you call, what you say, where you go, what you do first. Nobody makes a plan well while the self-blame is running.

✅ Debrief a lapse instead of confessing it. Within a day, answer four questions on paper. What was the stimulus? What was I feeling right before? What did I expect it to do for me? What was missing that could have competed?

✅ Track four things besides the streak. Frequency, duration, severity, and how fast you come back. If those are improving, you are improving, whatever the abstinence column says.

✅ Build one quick tool for your most common feeling. Pick the emotion that shows up before you reach, build a response that works in under two minutes, and rehearse it when you're calm. It has to be faster than the thing it's replacing.

✅ Watch stress and mood as leading indicators, and sleep alongside them. Stress and low mood predict return in the prospective research. Sleep is my own clinical addition, not Sinha's finding: three bad nights reads to me as a risk factor rather than just a bad week.

✅ Get real aftercare and keep it past the point it feels necessary. Structured aftercare beat treatment as usual at six months in the Bowen trial, with the mindfulness arm still ahead at a year. Ongoing support is not evidence that you're fragile.

✅ If you love someone who is using, learn CRAFT rather than staging a confrontation. In the original randomized trial of 130 family members of problem drinkers, the CRAFT approach engaged 64% of initially unmotivated drinkers in treatment, compared with 30% for a Johnson Institute confrontation and 13% for Al-Anon facilitation. That 30% is mostly about uptake: 70% of families couldn't bring themselves to hold the confrontation, and among those who did, most succeeded (DOI: 10.1037//0022-006x.67.5.688). Ask for a CRAFT-trained clinician, and note that all three approaches improved the family member's own functioning.

✅ Read the safety box above before any return to opioid use, and don't stop alcohol or benzodiazepines abruptly on your own. Both are medical situations, not willpower situations.

FAQ

What is the difference between a lapse and a relapse? A lapse is a single episode of use after a period without it. A relapse is a return to the old pattern. The distinction matters because the reaction to a lapse is one of four immediate determinants that push it toward becoming a relapse, alongside the situation itself, whether you have a coping response ready, and what you expect the substance to do for you (Larimer, Palmer & Marlatt, 1999, PubMed: 10890810).

Is relapse a normal part of recovery? It's common enough that any honest plan should account for it, and medication adherence and relapse rates in addiction look similar to those in diabetes, hypertension, and asthma (McLellan et al., 2000, DOI: 10.1001/jama.284.13.1689). Plenty of people resolve a substance problem without a dramatic cycle of returns. Plan for a lapse anyway. A plan doesn't summon one.

Does anything actually reduce relapse risk? Yes, and medication belongs at the top of the list. Opioid agonist treatment was associated with less than half the all-cause mortality rate compared with time off treatment (Santo et al., 2021, DOI: 10.1001/jamapsychiatry.2021.0976), and for alcohol the number needed to treat to prevent a return to any drinking was 11 for acamprosate and 18 for oral naltrexone (McPheeters et al., 2023, DOI: 10.1001/jama.2023.19761). Structured aftercare beat treatment as usual in a 286-person trial (Bowen et al., 2014, DOI: 10.1001/jamapsychiatry.2013.4546), and contingency management was the only treatment category significantly associated with reduced cocaine-positive tests relative to baseline across 157 studies (Bentzley et al., 2021, DOI: 10.1001/jamanetworkopen.2021.8049).

Why is returning to use so dangerous after time away? Tolerance falls during any break while the memory of your old dose doesn't. In a cohort of 30,237 people released from prison, the relative risk of overdose death in the first two weeks after release was 129 (Binswanger et al., 2007, DOI: 10.1056/NEJMsa064115). With fentanyl and nitazenes in the supply there is no dose you can safely titrate on your own. Don't use alone, keep naloxone present, and see the safety box above.

Does a lapse mean I have to start over? Not in any sense that matters. The abstinence violation effect describes how self-blame after a single lapse collapses confidence and pushes one episode toward a full return (Larimer, Palmer & Marlatt, 1999). The skills you built are still there the next morning. Only the counter resets.

If I cut down but didn't quit, did I fail? Not by the measure the trialists themselves argue we should be using. In a 2026 pooled re-analysis of 12 trials run between 2001 and 2011, 31.2% of participants achieved reduced use, a category that includes the 13.3% who reached verified abstinence, meaning roughly eighteen percentage points improved without becoming abstinent. The authors concluded that trials should evaluate a continuum of outcomes rather than abstinence alone (Amin-Esmaeili et al., 2026, DOI: 10.1001/jamapsychiatry.2026.1092). Those trials measured use, not health outcomes, so read this as an argument about scorekeeping.

Related in the Behavior Change Atlas

References

All citations verified against PubMed at draft time (2026-07-30).

  1. McLellan AT, Lewis DC, O'Brien CP, Kleber HD. Drug dependence, a chronic medical illness: implications for treatment, insurance, and outcomes evaluation. JAMA. 2000;284(13):1689–1695. DOI: 10.1001/jama.284.13.1689

  2. Larimer ME, Palmer RS, Marlatt GA. Relapse prevention. An overview of Marlatt's cognitive-behavioral model. Alcohol Res Health. 1999;23(2):151–160. PMID: 10890810. PubMed

  3. Witkiewitz K, Marlatt GA. Relapse prevention for alcohol and drug problems: that was Zen, this is Tao. Am Psychol. 2004;59(4):224–235. DOI: 10.1037/0003-066X.59.4.224

  4. Sinha R. New findings on biological factors predicting addiction relapse vulnerability. Curr Psychiatry Rep. 2011;13(5):398–405. DOI: 10.1007/s11920-011-0224-0

  5. Bowen S, Witkiewitz K, Clifasefi SL, Grow J, Chawla N, Hsu SH, Carroll HA, Harrop E, Collins SE, Lustyk MK, Larimer ME. Relative efficacy of mindfulness-based relapse prevention, standard relapse prevention, and treatment as usual for substance use disorders: a randomized clinical trial. JAMA Psychiatry. 2014;71(5):547–556. DOI: 10.1001/jamapsychiatry.2013.4546

  6. Santo T Jr, 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

  7. McPheeters M, O'Connor EA, Riley S, Kennedy SM, Voisin C, Kuznacic K, Coffey CP, Edlund MD, Bobashev G, Jonas DE. Pharmacotherapy for alcohol use disorder: a systematic review and meta-analysis. JAMA. 2023;330(17):1653–1665. DOI: 10.1001/jama.2023.19761

  8. Benishek LA, Dugosh KL, Kirby KC, Matejkowski J, Clements NT, Seymour BL, Festinger DS. Prize-based contingency management for the treatment of substance abusers: a meta-analysis. Addiction. 2014;109(9):1426–1436. DOI: 10.1111/add.12589

  9. Bentzley BS, Han SS, Neuner S, Humphreys K, Kampman KM, Halpern CH. Comparison of treatments for cocaine use disorder among adults: a systematic review and meta-analysis. JAMA Netw Open. 2021;4(5):e218049. DOI: 10.1001/jamanetworkopen.2021.8049

  10. Ainscough TS, McNeill A, Strang J, Calder R, Brose LS. Contingency management interventions for non-prescribed drug use during treatment for opiate addiction: a systematic review and meta-analysis. Drug Alcohol Depend. 2017;178:318–339. DOI: 10.1016/j.drugalcdep.2017.05.028

  11. 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

  12. Kelly JF, Bergman B, Hoeppner BB, Vilsaint C, White WL. Prevalence and pathways of recovery from drug and alcohol problems in the United States population: implications for practice, research, and policy. Drug Alcohol Depend. 2017;181:162–169. DOI: 10.1016/j.drugalcdep.2017.09.028

  13. Heyman GM. Quitting drugs: quantitative and qualitative features. Annu Rev Clin Psychol. 2013;9:29–59. DOI: 10.1146/annurev-clinpsy-032511-143041

  14. Binswanger IA, Stern MF, Deyo RA, Heagerty PJ, Cheadle A, Elmore JG, Koepsell TD. Release from prison — a high risk of death for former inmates. N Engl J Med. 2007;356(2):157–165. DOI: 10.1056/NEJMsa064115

  15. Hart CL, Haney M, Foltin RW, Fischman MW. Alternative reinforcers differentially modify cocaine self-administration by humans. Behav Pharmacol. 2000;11(1):87–91. DOI: 10.1097/00008877-200002000-00010

  16. Kelly JF, Humphreys K, Ferri M. Alcoholics Anonymous and other 12-step programs for alcohol use disorder. Cochrane Database Syst Rev. 2020;3(3):CD012880. DOI: 10.1002/14651858.CD012880.pub2

  17. Miller WR, Meyers RJ, Tonigan JS. Engaging the unmotivated in treatment for alcohol problems: a comparison of three strategies for intervention through family members. J Consult Clin Psychol. 1999;67(5):688–697. DOI: 10.1037//0022-006x.67.5.688

Limitations stated honestly: The chronic-illness comparison (McLellan 2000) is a narrative review making a health-policy argument, not a head-to-head study of relapse rates. The abstinence violation effect comes from a narrative overview of a clinical model, is one of four named immediate determinants rather than a sole mechanism, and the time windows used in this article are the author's clinical heuristics with no trial behind them. The dynamic reconceptualization (Witkiewitz & Marlatt 2004) is a proposed model whose authors discuss its own limitations. Sinha's predictors are group-level findings from prospective studies, not validated clinical tests, and vary by drug and phase. The Bowen trial studied aftercare among people who had completed treatment at a private facility, leaned partly on self-report, and by twelve months only the mindfulness arm retained an advantage. The opioid agonist mortality findings (Santo 2021) come overwhelmingly from observational cohorts, where people in treatment may differ systematically from people out of it. The alcohol pharmacotherapy effects (Jonas 2014) are modest, and numbers needed to treat describe populations, not individuals. The contingency-management six-month null (Benishek 2014) rests on six studies with a confidence interval spanning zero, so it is an underpowered null rather than a demonstration of no effect; Ainscough 2017 had follow-up data for only three of 22 studies and found no benefit for opiate use specifically. The 2026 stimulant meta-analysis found the medications tested mostly did not outperform placebo, measured drug use rather than health outcomes, and compared a self-reported reduced-use rate against a urine-verified abstinence rate. Kelly's 9.1% is a population prevalence of resolved problems, not a resolution rate among people who ever had one. Heyman's remission half-lives come from cross-sectional national surveys with retrospective recall. The Binswanger cohort is specific to people released from prison in one state between 1999 and 2003, before fentanyl dominated the illicit supply. Hart's choice study had six participants. None of this replaces an individual clinical assessment.

If you or someone you love is struggling, help is available: call or text the SAMHSA National Helpline at 1-800-662-4357 (free, confidential, 24/7), find treatment at findtreatment.gov, or reach the 988 Suicide & Crisis Lifeline (call or text 988) if you are in emotional crisis. If you use opioids, keep naloxone available, never use alone, and consider the Never Use Alone line at 1-800-484-3731.

Written by Dr. Adi Jaffe, PhD (UCLA), author of The Abstinence Myth and Unhooked. Last medically reviewed: 8-18-26