It is late, and you have run out of ideas. You have asked. You have argued. Someone has told you to go to Al-Anon and detach with love, and someone else has told you to get the family in a room and confront him before it is too late. Both pieces of advice came from people who meant well, and they contradict each other. Underneath the question of which one to follow is a harder one: does anything a family does actually change whether a person walks into treatment, or is that entirely up to him?
That question has an unusual property. It has actually been tested. In 1999, a research team put all three of the standard approaches side by side in a randomized trial and counted who got into treatment. The families trained in the third approach -- the one you have probably never heard of -- got their loved one into treatment roughly five times as often as the families sent to Al-Anon. This article is about that trial, what it does and does not prove, and where a family can learn the skills it points to.
---
For most of the modern history of addiction treatment, families were offered two scripts. The first was the twelve-step family program: you did not cause it, you cannot control it, you cannot cure it. Its purpose was to help the family survive, and it was never designed to get the drinker into a clinic. The second was the planned confrontation, popularized as the Johnson Institute intervention -- the surprise meeting, the letters read aloud, the ultimatum.
A third approach was developed in the late 1970s by Robert Meyers and colleagues out of the community reinforcement tradition. Community Reinforcement and Family Training, CRAFT, starts from a different premise: the family is not a bystander to be supported, and not a weapon to be aimed, but the person with the most contact and therefore the most leverage. It teaches specific skills rather than a posture. The American Psychological Association's summary lists the components as "motivation building, functional analysis, contingency management training, communication skills training, treatment entry training, immediate treatment entry, life enrichment, and safety training."
Three approaches, three theories, decades of confident advice. In 1999 someone finally measured them against each other.
This article has three parts: what the head-to-head trials found; what the newer and more equivocal trials actually compared, which is not what it first appears; and the limits, including what to do this week. What turns on it for you is whether the time between tonight and the day he agrees to go is time you can use.
---
The central study is Miller, Meyers and Tonigan, published in the Journal of Consulting and Clinical Psychology in 1999. It randomized "130 concerned significant others (CSOs)" to one of three manual-guided approaches -- Al-Anon facilitation, a Johnson Institute confrontation, or CRAFT -- each with twelve hours of contact. Follow-up ran a full year: "Follow-up interviews continued for 12 months, with 94% completed."
The result: "The CRAFT approach was more effective in engaging initially unmotivated problem drinkers in treatment (64%) as compared with the more commonly practiced Al-Anon (13%) and Johnson interventions (30%)."
Two details in that same abstract matter as much as the headline. First, on the confrontation: "most CSOs decide not to go through with the family confrontation (70% in this study) and ... among those who do, most (75%) succeed in getting the drinker into treatment." The intervention works when it happens; it mostly does not happen, because families cannot bring themselves to do it. Second, and this is the finding most often left out when CRAFT is promoted: "All 3 approaches were associated with similar improvement in CSO functioning and relationship quality." Al-Anon helped the family as much as CRAFT did. It simply did not get the drinker in the door. Those are different jobs, and a family may legitimately want both.
The trial also reported that "treatment engagement occurred after 4 to 6 sessions" on average, and that "Overall treatment engagement rates were higher for CSOs who were parents than for spouses."
A second 1999 trial, by Kirby and colleagues in Drug and Alcohol Dependence, extended the approach from alcohol to other drugs. It "randomly assigned 32 concerned family members and significant others (FSOs) of drug users" to community reinforcement training or a twelve-step self-help group, and found that "the community reinforcement intervention was significantly better at retaining FSOs in treatment and inducing treatment entry of the DUs."
The original study, Sisson and Azrin in 1986, is the one usually cited first and it is the weakest of the three. Its abstract is worth quoting exactly, because the widely repeated "86 percent" figure does not appear in it: "Twelve concerned family members were given either community-reinforcement counseling or a traditional type of counseling (control group)," and "The reinforcement counseling resulted in more alcoholic persons obtaining treatment than did the traditional type." Twelve people, direction reported, no percentage in the record. Treat it as the origin of the idea, not as evidence for it.
---
If you go looking, you will find recent, larger, better-funded CRAFT trials that reported no significant effect, and you deserve to know about them before you spend money on this.
In a 2012 pilot published by Manuel and colleagues, forty concerned significant others were randomized between group-delivered and self-directed CRAFT. The result: "Although results indicated no significant between-group difference in engaging treatment-refusing substance-using individuals ... into treatment," and "For the intent-to-treat analysis, 60% of Group CRAFT CSOs engaged their loved one into treatment, as compared with 40% in Self-Directed CRAFT."
In 2022, Hellum and colleagues published a Danish cluster-randomized trial in which "Eighteen public treatment centers for alcohol use disorders were randomly assigned to deliver CRAFT in one of the three formats as part of their daily clinical routine." "At three months follow-up, 29% (n = 32) of the CSOs who received group/individual CRAFT reported that their IP had engaged in treatment. The corresponding rate for the CSOs who received self-administered CRAFT was lower (15%; n = 5) but did not differ significantly from the other group of CSOs (Odds ratio (OR) = 2.27 (95% CI: 0.80, 6.41))."
Read the comparison groups. In both trials, every arm received CRAFT. The 2012 study compared group delivery against a workbook. In the Danish trial, "self-administered CRAFT was limited to handing out a self-help book." Neither study asked whether CRAFT beats no CRAFT. They asked how it should be delivered -- and found that the format did not make a statistically detectable difference.
That is a real result, and it cuts in an unexpected direction. It is not evidence that CRAFT fails. It is evidence that a book in a motivated family's hands did not perform detectably worse than sessions with a trained clinician. For a family without money for a specialist, that is the most useful sentence in this article.
Two honest caveats on the Danish trial: "The three-month follow-up rate was 60%," and the outcome was what the family member reported, not a verified clinic record.
---
Here is the case against everything above, and it is not weak.
The one large trial showing a dramatic advantage is now more than twenty-five years old, and Robert Meyers, who co-developed CRAFT, is an author on it. The most recent, largest, independent trial found a difference that did not reach statistical significance. A skeptic could reasonably say: this is an aging result, produced in part by the people who invented the intervention, and the modern independent evidence is equivocal.
The fair answer is a concession and a distinction. The concession: no one has repeated the 1999 head-to-head comparison against Al-Anon or a planned confrontation. The 64 percent figure stands unreplicated -- not overturned, not confirmed. Anyone who quotes it to you as settled science, including anyone selling a program, is overstating it.
The distinction: the newer trials did not fail to replicate that comparison, because they did not attempt it. You cannot read a null result from a format comparison as a null result for the intervention. What we have is one strong direct comparison, two smaller supporting trials, and a body of later work that has moved on to asking implementation questions.
The broader literature sits in the same posture. A 2026 systematic review in the Indian Journal of Psychological Medicine screened 3,864 records, found that "15 trials met the inclusion criteria," and reported "11 studies demonstrating significant positive effects of family-based interventions" -- while concluding that "further high-quality RCTs are necessary to strengthen these conclusions and provide more definitive evidence." That is the accurate summary: the best-supported option available to a family, resting on an evidence base thinner than its reputation.
---
CRAFT assumes it is safe for you to change how you respond. Safety training is a formal component of the model for a reason. If there is violence in your home, this is not a self-help project; it is a conversation to have with a professional who knows your situation.
And entry is not recovery. Getting someone through the door is the outcome these trials measured, and it is not the same as getting well. What happens next still matters -- a 2021 systematic review of social network support in medication treatment for opioid use disorder screened 5,193 articles, included eight studies, and found that "Five studies indicated that social network support had a statistically significant effect on improved MOUD treatment outcomes," with "the strongest support for the positive impact of family social network support." Five of eight is a modest base. It points the same direction.
---
The most common failure of articles like this one is that they name the thing that might help and never tell you where to get it. So:
Read the book. Robert Meyers and Brenda Wolfe wrote Get Your Loved One Sober: Alternatives to Nagging, Pleading, and Threatening as the consumer version of CRAFT. In the Danish trial, the self-administered arm was essentially this -- a book -- and it held its own against clinician-delivered sessions.
Work through a structured program. Allies in Recovery offers web-based CRAFT training for families. CMC: Foundation for Change, a nonprofit, teaches a related evidence-based approach for families. The Partnership to End Addiction publishes free CRAFT-based guidance and runs a helpline for parents.
Find a trained clinician if you can. Ask directly whether they are trained in CRAFT; it is specific enough that a yes or no is meaningful.
And if you need a place to start tonight, SAMHSA's National Helpline is 1-800-662-HELP (4357) -- "a free, confidential, 24/7 treatment referral and information service for individuals and families facing mental and/or substance use disorders."
You cannot decide this for him. That was never on offer. What the research says is narrower and more useful: how you respond in the months before he agrees is not neutral, the specific skills that shift the odds can be learned, and in the one trial that compared the options directly, families who learned them were far more likely to see the door open. That is not certainty. It is leverage, and it is more than you were told you had.