Alcoholics Anonymous is the most widely used treatment for alcoholism in the world, yet it continues to come under attack by popular media ignorant of the science behind its success. A recent high profile attack appeared in the April 2015 issue of The Atlantic, in the form of an article by Gabrielle Glaser titled, “The Irrationality of Alcoholics Anonymous.” In this article, Ms. Glaser boldly states that 12-step programs lack a scientific foundation and that most professional treatment programs fail to provide scientifically supported treatment, largely because they are 12-step oriented. Ms. Glaser writes, “The problem is that nothing about the 12-step approach draws on modern science: not the character building, not the tough love, not even the standard 28-day rehab stay.”
Contrary to Ms. Glaser’s sweeping statements about the lack of science concerning AA, a significant body of research has been conducted on this organization and its impact on drinking and other variables. For example, three colleagues and one of the authors (CDE) published a meta-analysis of the scientific literature on Alcoholics Anonymous in 1993, incorporating a grand total of 107 data sets in the overall analysis. The findings of this meta-analysis were correlational due to the fact that most of the available data at that time were correlational in nature. These results showed positive correlations between AA membership and drinking outcome, as well as other outcome measures such as psychological health. Of course, correlation does not mean causation. Thus, the data at that time offered promising evidence for the effectiveness of AA but could not support the conclusion that involvement in AA causes better outcomes with respect to drinking and other variables.
But science moves on. A more recent publication (Vaillant 2012) offers yet another example of high-quality research on AA. Dr. George Vaillant of Harvard University reported his analysis of two male cohorts (Harvard undergraduates and inner-city Boston youth) who were studied in depth for 60 years (from the time they were 20 until they were 80)! Over the course of the study, 39 men in the college cohort and 101 men in the inner city cohort were identified as alcoholics. When the lives of these men were studied at age 80 (some of the men were deceased but information was obtained on the status of these individuals at the time of their death), 9 of the college cohort who became alcoholic had been abstinent an average of 15 years and 57 of the inner city cohort alcoholics had been abstinent for an average of 16 years. The remaining men in both cohorts had been abstinent an average of only 1 year over the course of their lives. It is important to note that the duration of active alcoholism did not differ between those who developed long-term abstinence and those who did not. Of relevance to the present article is that those who achieved long-term abstinence in the college cohort attended an average of 137 AA meetings compared to just 2 meetings among those with only short-term abstinence, while those in the inner-city cohort who maintained long-term abstinence attended 143 AA meetings, on average, compared to just 8 meetings among those who did not acquire long-term abstinence. In answer to the question, “is recovery through AA the exception or the rule?” Dr. Vaillant concludes, “In both cohorts, the men who were stably abstinent attended about twenty times as many AA meetings as the chronically alcoholic.” These data, while remarkable, are, as with the Emrick et al. findings, plagued with the issue of self-selection bias. It could be that individuals who go to AA are more motivated to stop drinking than those who don’t become AA involved, with the result that AA members have better drinking outcome, not because they are participating in AA, but rather because they were a more motivated group of alcoholics to begin with. The possibility of self-selection bias thus prevents Vaillant’s (as well as Emrick et al.’s) data from offering evidence that AA involvement causes better drinking outcome.
Fortunately, scientific investigations of AA have continued to advance. From 1993 to 2010, five randomized clinical trials were conducted in which AA Facilitation Interventions (AAFI) was one of the treatments studied. A general finding of these studies is that patients who received some form of AAFI had better drinking outcome than patients receiving alternative treatment(s), with the better outcome appearing to be mediated by AA involvement. Unfortunately, even with these and other clinical trials on AA, selection bias continues to be a thorny problem. That is, some patients assigned to AAFIs do not become involved in AA and patients assigned to alternative interventions become involved despite their being in treatment that does not encourage participation in AA. Given this situation (known as crossover) if AA participation in these studies is found to lead to better outcome than non-participation, we cannot be sure that involvement in AA per se is causing the better outcome. This is because the better outcome seen in AA members may be due, at least in part, to their having stronger motivation to recover from alcohol problems than do non-AA participants. Thus, selection bias is not fully eliminated even when using a randomized clinical trial research design.
In order to address this nagging issue pertaining to the aforementioned randomized clinical trials, Dr. Keith Humphreys, a professor at Stanford University, and colleagues employed an innovative statistical analytic method that controls for selection bias--a procedure called instrumental variables modelling. This analysis enabled the researchers to determine if increased AA involvement due to AAFIs made a difference in drinking outcomes when the role of the participants’ motivation to recover from alcoholism was taken out of the comparison between patients receiving AAFIs and those getting alternative treatments. The results of this study were published in the prestigious peer-reviewed journal, Alcoholism: Clinical and Experimental Research in November of 2014. Humphreys et al. used the number of days abstinent as the outcome measure. The main finding was that at both three and 15-month follow-ups, those who increased AA attendance due to the effects of AAFIs (not personal motivation) had significantly more days of abstinence than those getting alternative treatments who did not go to AA. To clarify, involvement in AA was the variable that led to better drinking outcome, not receipt of AAFIs per se. The scientists conclude, “For most individuals seeking help for alcohol problems, increasing AA attendance leads to short- and long-term decreases in alcohol consumptions that cannot be attributed to self-selection.”