Dr. Bankole Johnson, a justly respected addiction researcher, published an op-ed in the Washington Post last week that lambasted 12-step mutual help organizations such as AA. In the Washington Post today, I correct Johnson’s assertion that there are no randomized clinical trials supporting 12-step interventions. The amount of benefit to addicted patients in the trials I mention (One published by Christine Timko and colleagues in Addiction, the other by Leonard Jason and colleagues in American Journal of Public Health) not incidentally surpasses that found in most studies of the alcoholism medications for which Johnson advocates. We have some promising developments in terms of medications, but we clearly don’t have a blockbuster drug yet and certainly don’t have a basis for saying that we don’t need the 12-step approach anymore.
What I didn’t have space to say in a 200-word letter I will say here: I agree with Dr. Johnson’s questioning of the value for money from $50,000/month day-spa-with-massage-rehabs for addicted movie stars in Southern California, but am mystified that he lumped a free, non-profit mutual help organization together with such boondoggles. My research with Professor Rudolf Moos on these organizations shows that they take an enormous financial burden off of society because they substantially lower health care utilization. That means lower tax burden and reduced insurance premiums for the rest of us while at the same time saving lives…..which as a cost-benefit arrangement is the other end of the world from Malibu rehab inc.
I've never understood the antipathy towards AA. I know the good it does, but even if one doesn't believe that, it certainly does no harm. And it costs next to nothing. The groups collect non-compulsory contributions from members and uses them to pay the church where they meet a small rental fee, to buy some literature and usually a small percentage to AA HQ to help fund its operations. What's so threatening about a group of drunks sitting around talking to one another that in no way troubles or costs anyone else?
It's a fair question, You Don't Say. I have spent almost a quarter century in hospitals and medical schools, and one thing I have observed about medical culture is that anything that doctors don't do themselves tends to be less valued. The idea that a "mere alcoholic patient" might have unique talents and ability to help goes against the grain and hits some medical professionals both in their personal ego and in their professional vanity. Thus they deny the evidence that in fact, sometimes a "group of drunks" really is the best medicine. I am no better than my colleagues on this, I was contemptuous when I first heard of AA early in my career…but the data kept piling up and the patients kept getting better and it was time for me to ask myself Keynes question: "When the facts change, I change my mind. What do you do, sir?".
@You Don't Say: Much of the antipathy toward AA comes from former AAers, individuals whose first-hand knowledge would contradict your claim that AA "certainly does no harm." A typical narrative might run like this: somebody with a drinking problem turns to AA, thinking that they can help him. He sticks with the program for years, trying very hard to make it work, but he keeps relapsing. Eventually, he realizes that all this powerlessness stuff was a lot of baloney, he stops going to meetings, and he realizes that he can stop drinking on his own. If you have a sincere desire to understand where this "antipathy" comes from, here's a good place to get started:
http://stinkin-thinkin.com/2010/03/16/tell-us-you…
This stuff comes from real people's experiences, not out of thin air.
Dr. Humphreys:
You gave your post a title of "Science supports the 12 steps for addiction," but wouldn't it be more accurate to say that some science supports the 12 steps, and some science doesn't? We've got one scientist who cites a few papers that show one thing, and then another scientist cites some that show the opposite. But let's just discuss the ones you brought up. You mention of "a 12-step-oriented sober-living home," but a close reading of the article shows it was no such thing.
From the article itself:
http://www.ncbi.nlm.nih.gov/pmc/articles/PMC15861…
"Similar to those who participate in Alcoholics Anonymous, members of an Oxford House receive abstinence support from peers; however, unlike Alcoholics Anonymous, there is no single, set course for recovery that all members must follow. In fact, residents of Oxford House are free to decide personally whether to seek psychological or substance abuse treatment by professionals or a 12-step organization. In short, Oxford House offers residents the freedom to decide whether to seek and choose which (if any) treatment they desire while receiving constant support and guidance within an abstinent community."
I don't understand why you characterize the Oxford House as being oriented toward the twelve steps. If I was to draw one inference from that study, it would be that those living together with a group of people sharing a commitment to sobriety are more likely to stay sober than those participating in an outpatient program while living in the community at large.
I'm afraid that the study you did with Dr. Moos doesn't offer much support for the 12-step approach, either. To simplify, you've got two groups, one of which was assigned to a 12-step program, and the other assigned to a cognitive-behavioral (CB) program. The outcomes for most metrics were comparable, but the 12-step group showed a significantly higher level of abstinence.
All this is fine, but there was no control group, i.e., no group that got no treatment at all. Without that there's no way to say whether either of these treatment programs did any good. It's conceivable that a control group would have done better than either of the groups that some form of treatment or another.
You mentioned another study conducted at the Veterans Affairs Palo Alto Health Care System, but you didn't provide a link. I couldn't find any reference to it on line based on the few clues you gave.
As I find that my professional license prohibits my using the "randomized clinical trial" approach because I find it to have deceptive aspects, I use a somewhat contrasting method to unriddle the mechanisms of addictive behaviors. I talk with people who have encountered addiction and I talk with people who have found ways to escape from, evade, or elude addiction.
As I have observed, AA (and Narcotics Anonymous, Sex Addicts Anonymous, Gamblers Anonymous, Schizophrenics Anonymous, Emotions Anonymous, and the list of other 12 Step groups who have sincere adherents is long indeed) are effective to very effective for some people and are sometimes even damaging to others.
If addiction is a condition of formidable complexity, perhaps a variety of methodologies may be required to effectively address the whole gamut of addiction, and seeking the "one answer" to a problem having a multiplicity of "answers" may impede "solving the problem."
There is also Abraham Low Self-Help Systems — formerly Recovery, Inc. — which has adherents no less sincere than many AA members are, and there is also a third approach, that of Grow, Inc. (started, as I recall, in Australia). What works for one person works for that one person.
As for the need, in valid science, for adequate controls, I use the method of longitudinal control, wherein the "research subject" is his/her own control, with the state of the person prior to an "intervention" being the control, and the state of the person after said "intervention" providing the data which contrasts with the control data.
With a sufficiently large group of people who are, individually their own controls, pattern-recognition methodologies bring forth what I find to be usefully valid data. Until someone comes up with a verifiably-real panacea, perhaps a multi-faceted effort to understand humanity will continue?
I seem to have encountered the "but its not a panacea" argument enough times to have come to regard it as plausibly specious.
Thanks for posting Rachelrachel. Leonard Jason's research on Oxford House shows that almost every member of the homes is actively involved in 12-step self-help groups, and as the founder of OH (whom I know well) would tell you and is evident in OH materials, AA's philosophy had an enormous influence on how OH operates. I spoke an OH's annual meeting this past year and met dozens of members. The influence of the 12 steps on OH residents is omnipresent.
Note also that I mentioned two clinical trials in my letter for space, but there have been others, including by Dr. Mark Litt and Dr. Kim Walitzer. One would have to deny a lot of science to say that the 12-steps don't work.
On my cost study, remember that the purpose of a control group is to make a comparison relevant to the question of the study. If our study was asking, "What are the cost impacts of treating versus not treating addiction?", we would have had a group of untreated individuals to compare to. But the answer to that question is well known and was not what we were studying. Rather, we were looking at the question "Among those who get treatment, what are the cost implications of actively promoting AA/NA involvement versus not doing so?" for which the appropriate control group was the one we used, people in treatment programs that did not engage in such encouragement.
Finally, because some people think that if AA works, it must be the only way, I would note there are other options out there. I served on the advisory board of an organization called SMART, which is based on cognitive-behavioral principles and has no spiritual content. A colleague and I have been working to expand SMART in the U.K. with some success.
Perhaps I was not fully awake when I wrote my comment of "August 13, 2020 at 6:25 am". The end of the "third from the end paragraph" was intended to be, "…providing the data set which contrasts with the control data set." My mother, prior to marriage, was an English teacher. The word "data" is plural, so, were "data" my intended term, the verb would properly be "contrast." "Misteaks" happen? Sorry about that.
The Cochrane review which was mentioned in the Johnson editorial does conclude that there is a lack of experimental evidence on the effectiveness of 12 step programs. The review also concludes that the interventions appeared to improve at least some of the outcomes. Cochrane places a lot of emphasis on risks of bias in randomized trials, and notes that the 12 step program studies did not describe the method of randomization or make it clear that there was adequate concealment of allocation. The Cochrane review also had problems with some of the statistical analyses, especially the inflation of Type I error that can occur with multiple hypothesis testing.
So the question for Keith is how to interpret the Cochrane review. I think that there is soon to be a meeting of leaders from both the Cochrane and the Campbell collaboratives (in Colorado in October if I am not mistaken), where some of the issues of how to study social interventions may be discussed. Randomized assignment to AA raises problems, since the program is set up as one that is selected and sought after by alcoholics, often with referrals from doctors who think they will be likely to succeed. Since Keith mentioned randomized trials as if they are the preferred study design for AA, I would be interested in his thoughts on this issue. If Bankole Johnson is wrong to place too high a value on these studies, what is a better study design?
To rachelrachel: It does not help everybody, a smaller percentage than it does helps. It is entirely voluntary so if it doesn't work it's up to the person to move on, nobody is holding anybody hostage. The antipathy I've encountered and speak of, though, is from pundits who have no real personal experience with AA, they only study or comment on it. And from people who are vehemently opposed to anything that smacks of religion. I admit that grappling with the issue of "God" at all can be burdensome when you're just trying to get sober, but AA is clear that it is a higher power of your choosing. If that still seems ridiculous to someone who is trying to get sober, then move on. Sobriety is the goal, not adherence to AA.
Dear Ed
Thanks for your well-informed post — did not know about the upcoming meeting. The Cochrane review was published in 2006. The Jason, Timko, Litt, Kaskutas and Walitzer trials were all published after that date. I have some reservations about Cochrane reviews in general, particularly that they often don't say anything that helps in patient care (Too often it's "We looked at a thousand studies and think more research is needed" as if patients could just sit there for years waiting for care). On clinical trials in general, I am not one of those people who think they are the only path to knowledge, for a variety of reasons. If you want to see a short discussion of some of the weaknesses of trials, Stanford Medical Magazine did an accessible article on it a few years back
http://stanmed.stanford.edu/2008summer/fixing_tri…
But however one feels about Cochrane and trials, one thing is certain: When Cochrane updates its review it will get much more positive about AA because (1) Cochrane focuses almost exclusively on trials and (2) Since the last review a number of trials have been done and they have all been positive.
One thing about Cochrane is that you do not necessarily have to wait for the next review to be published. They make the RevMan (for Review Manager) software free on their website. They did not try to do a meta-analysis on the trials they had because of the heterogeneity of the effects that were measured. But if you have more recent trials and think that some of them are similar enough to pool with the existing Cochrane trials, you can get the software and plug in the data from the new trials and see what you come up with.
But that is a minor issue. The major one is random assignment to an intervention that requires major participation and commitment on the part of the participant. For example, there is actually some pretty good basic science to suggest that abstinence prevents teen pregnancy and STDs. Suppose you went in to a high school and randomized the kids to abstinence or to usual adolescent behavior. Over the course of the study, you could anticipate that some of those randomized to abstinence would cross over into the control group. And not everyone in the control group will get lucky. When you did your intention to treat analysis of the results, you would find a treatment effect that was more conservative than the one you would get if you only analyzed the kids who adhered to their assigned protocols. This is why there is animosity between the advocates of abstinence alone and their opponents. The former are looking at data from those who adhere to the program; the latter are looking at all participants.
This does not mean that abstinence does not “work.” It does. But it should inform the decisions that are made concerning what to teach in the schools. Maybe you want to have a program that goes to the classrooms and teaches teenage girls to say, “Not tonight dear; I have a headache.” But if you do, you owe it to everyone to be honest about what analysis you used and what assumptions you made in using it.
From a response posted this week (http://addictionandrecoverynews.blogspot.com/2010/08/rehab-doesnt-work.html)
I’ve posted about Cochrane before, the most germane is a summary of a Sara Zemore presentation last year:
She very effectively rebutted the Cochrane Review from a few years ago by making the following points. (These are based on notes I took and are incomplete. Hopefully they post video so that you can see her complete rebuttal for yourself.)
* It was limited only to randomized trials and ignored the overwhelming observational evidence.
* It included one of Zemore’s studies which was NOT a randomized study of AA.
* She acknowledged that the randomized evidence is ambiguous.
* Randomized trials of AA are hard to do because some subjects in other groups end up participating in AA. This happened in Project MATCH.
* The Cochrane Review did not find Twelve-step Facilitation ineffective. It found it no more effective that CBT and MET. (The summary from the abstract says, “The available experimental studies did not demonstrate the effectiveness of AA or other 12-step approaches in reducing alcohol use and achieving abstinence compared with other treatments, but there were some limitations with these studies.”)
* Finally, she cited 4 randomized studies of Twelve-step Facilitation: The outpatient arm of Project MATCH, a study by her colleague Kaskutas, and two others that I missed.
Jason points out the restriction of Cochrane review on AA to randomized trials. Interestingly enough, although only a small percentage of Cochrane reviews included non-randomized trials in its first couple of decades, an increasing percentage of reviews in the past few years incorporate observational studies. The old "hierarchy" of study designs, with homogeneous meta-analyses and large randomized trials at the top of a pyramid of designs, has been debated by researchers and has been supplanted by a more flexible approach to study designs. They cite several examples of the need for this flexibility.
For example, Reye syndrome, once a fairly common disease of children, was linked to the use of aspirin in febrile illness. On the basis of a few case-control studies, the use of aspirin was discouraged for febrile children, and Reye syndrome became very rare. Sudden Infant Death Syndrome (SIDS) was similarly linked by case-control studies to placing infants on their stomachs to sleep; when recommendations to place babies to sleep on their backs, the incidence of SIDS sharply declined. These examples are cited by the Cochrane leaders themselves in emphasizing the need to find studies whose methods are well fitted to the purpose of the study, rather than emphasizing one design only.
There is a tendency in some domains of science for flexible approaches to thought to morph into rigid systems. Sir Ronald Fisher, when asked how great a chance was acceptable to make the error of concluding that there was a difference between groups when in reality there was none, said, "Oh, I should think about one in twenty." This morphed into a rigid "statistical significance" level of p<0.05. Bradford Hill suggested a variety of "viewpoints" for thinking about causal relationships between different variables; these morphed into "The Bradford Hill Criteria" for causality. The principle that some kinds of study lend themselves to more straightforward interpretation than others morphed into a rigid Hierarchy of Study Designs.
So it seems that those who dislike uncertainty have a way of taking power and making strict sets of rules where there were originally adaptable approaches to thinking. Any of you sociologists have any insights into how this happens?
If ever I am on a crashing airplane with one other person and only one parachute, I hope the other person is a rigid trialist. That way I can take the parachute and say "Don't worry, no one has done a randomized clinical trial of these things!"
The parachute example is a great conversation starter, since it illustrates one time where having a mechanistic explanation (gravity accelerating at 32 feet per second per second) makes probabilistic inference (odds ratios, relative risks with confidence intervals) unnecessary. Similarly, we think that dropping a nuclear bomb on a large city would have adverse consequences for public health, in spite of the fact that Truman only dropped a real bomb on Hiroshima and did not at the same time drop a placebo bomb on Kyoto. The laws of physics were robust enough to make a comparison unnecessary.
However, gravity was also the plausible mechanistic reason that Dr. Spock recommended to millions of mothers that they put baby to sleep on the tummy; that meant that the risk of accidental aspiration would decrease since gravity would direct the potential aspirated contents down onto the mattress and not into the lungs. The mechanism seemed sound, but the approach probably led to the deaths of many infants from SIDS.
The same kind of reasoning that succeeded in helping people survive a jump out of an airplane failed terribly in protecting babies from sudden death. An explanatory principle that works in one place fails in another. That is why evidence-based health care is hard and not easy. If anyone claims that interpretation of evidence is straightforward, you should regard them with the same horror that you would feel if a nominee to be Chief Justice of the United States got up and said that interpreting the Constitution was just like an umpire calling balls and strikes. Thank God such a thing could never happen here!
Maybe I need to clarify the big problem with randomizing people to AA. You cannot really assign anyone to surrender to a higher power; the impulse to join AA has to come from within, not from without. If you try to allocate someone to AA, you can try, but it is not really AA to which you are assigning them; it is something else altogether. That is why AA is a great example of when not to even try to randomize, since anything based on a fundamental contradiction is going to be flawed, regardless of methodological rigor elsewhere.
Here is an articulate response to Dr. Johnson's op-ed by Dr. Deni Carise of Phoenix House
http://www.phoenixhouse.org/blog/johnson/
Ed
I have articulated almost the same argument you make in my writings and agree with you, we can't randomly assign people directly to full blown AA membership because part of whether AA "takes" is not controllable by a research procedure. However, I think this is less serious a problem than at first might appear.
Randomized trials are never trials of full compliance with treatment. For example, although we say for short hand that a trial of medication A versus medication B is a comparison of the two medications, it isn't really. It is a comparison of *being given* one medication or the other by the physician. Many people will not take their meds, or will take them the wrong way etc.. Thus all the intent to treat analysis difference in outcome tells us is what is the effect of a doctor prescribing one or the other. By the same token, a "randomized study of AA" is not a study of accepting a higher power/getting a sponsor/loving the program etc., but of being introduced to AA or not (which some will comply with and some will not). This applies to many things about which we make inferences, for example we study the effect of exercise, diet etc., all of which are really studies of being encouraged to do something by a health care provider, only some of which people will follow through with.
In a trial, there is a way to get an estimate of the portion of compliance that is exogenous, namely by using randomization itself as an instrumental variable. This allows nice secondary analysis of the effect of the "treatment" (whatever it is) on the outcome apart from self-selection. But even then, that is not what the intent to treat main analysis is about, its about being introduced to the treatment or not.
No disagreement about what intention to treat analysis does. It preserves the randomization, but at a small price: substituting one research question for another. The effectiveness of prescribing an intervention is not the same question as the effectiveness of the intervention itself.
But it still seems that a physician referral to AA contains an element of paradox if there is something odd about being referred by someone else to a program whose very heart is a realization that comes from within. Self-contradiction may be stating it too strongly. Sure, you can prescribe an introduction to the idea that "we were powerless over alcohol and our lives had become unmanageable." That is what the intention to treat analysis tests: the average effectiveness of the prescription.
However, you can get better decision-making information from good observational studies giving the characteristics that distinguish people who succeed at AA from those who fail. A handful of those may be more suited to the clinician's purpose than an equal number of randomized trials. Making appropriate referrals could be better guided by knowing who is likely to succeed than by a comparison of average success rates in people referred or not referred at random. I am not saying "down with randomized trials." I am saying "up with observational studies."
I agree that Dr. Bankole Johnson’s recent opinion piece (August 8th) that criticized professional 12-step-based treatment and Alcoholics Anonymous participation overlooked important contrary evidence. It should have been noted that there have been a number of experiments conducted testing 12-step-oriented treatment against state-of-the-art, alternative treatments, such as cognitive-behavioral therapy. Three recent trials conducted by the University of Connecticut, the Research Institute on Addictions in Buffalo, NY, and the Alcohol Research Group in Berkeley, CA, showed clinically superior outcomes for up to two years for those receiving 12-step, compared to other, treatments. It was demonstrated also in these studies that the reason for the 12-step advantage was due to AA participation during the follow-up period.
Because of the burden of disease attributable to alcohol and drugs together with high and increasing healthcare costs, our societal approach should be comprised of a combination of professional and community interventions, such as AA, whenever possible. The broader evidence shows that when this done, such as in 12-step-oriented treatment, the combination results in reduced health care costs and improved outcomes. Finally, a noteworthy absence in Dr. Johnson's opinion piece was a mention of specific better alternatives. I presume he means a medication is the answer but we don't have the magic bullet yet.
Excellent summary of the evidence by Dr. Kelly (who was too modest to note that he is a Harvard Medical School professor and one of the country's leading experts in this area)