Authors' objectives

To examine the effectiveness of the Alcoholics Anonymous (AA) programme.

Searching

The following computerised databases were searched (from 1953-1995): PsycLIT, MEDLINE and Dissertation Abstracts International. In addition the reference sections of key review articles and numerous individual articles were checked for further studies. Both published and unpublished studies were included.

Study selection

Assessment of study quality

The main criteria were study design, whether participants volunteered or were coerced into participation, whether the alcohol status was verified by collateral reports, the type of outcome measure and other study limitations/unresolved issues. A narrative description of overall study quality was reported and a summary table of methodological features for the individual studies was presented (study site/setting, data source and study limitations/unresolved issues). Not stated who or how many individuals assessed the studies.

Data extraction

It was not stated who or how many individuals were involved in extracting the data. Data tables reported in the review include: bibliographic details, whether participants were coerced or volunteered, the outcome measures, whether the outcome was collateral, length of follow-up, interventions and the number of participants included in each study group at baseline and follow-up. For each study the outcome measures were converted to a standardised mean difference statistic (d) for continuous data or an odds ratio (OR) for dichotomous data (this was subsequently converted to a d statistic to facilitate study pooling).

Methods of synthesis

Results of the review

Twenty-one controlled studies including approximately 7000 participants (10 randomised controlled trials (RCTs) and 11 non-RCTs).

Effect sizes:

1. RCTs of conventional AA meetings.

Overall (n=5) -0.54 (95% CI: -0.98, -0.11; p<0.05).

AA vs. no treatment (n=2) -0.25 (95% CI: -0.54, 0.05).

AA vs. alternative treatment (n=3) -0.61 (95% CI: -1.31, 0.10).

2. Non-randomised studies of conventional AA meetings.

Overall (n=10) 0.40 (95% CI: 0.19, 0.61; p<0.05).

Attended vs. not (n=7) 0.39 (95% CI: 0.11, 0.66; p<0.05).

Attended more vs. less (n=2) 0.55 (95% CI: 0.27, 0.82; p<0.05).

Attended vs attended plus treatment (n=1) 0.27 (95% CI: -0.18, 0.71).

3. RCTs of AA residential treatments.

Overall (n=2) 0.39 (95% CI: -0.16, 0.95).

4. Non-randomised studies of AA residential treatments.

Overall (n=2) 1.33 (95% CI: 0.28, 2.39; p<0.05).

5. RCTs of components of AA programme (i.e. dismantling studies).

Overall (n=5) 0.38 (95% CI: 0.13, 0.63; p<0.05).

Three studies looked at the use of recovered alcoholic counsellors (vs. non-alcoholic counsellors) 0.31 (95% CI: 0.03, 0.59); and two studies at training participants in the AA twelve steps or the AA honest inventory (vs. discussion groups, hypnotherapy, or pharmacotherapies) 0.54 (95% CI: 0.16, 0.92).

Study quality and methodological issues:

The results from the non-randomised studies (vs. RCTs) of conventional AA meetings showed significantly larger more positive results (chi-squared=14.56, df=1, p=0.0001). Studies that used collateral reports to check outcome status yielded non-significantly smaller effect sizes (0.10, 95% CI: -0.32, 0.53) than when no collateral checks were used (0.49, 95% CI: 0.18, 0.81). However, the pooled effects sizes were not significantly different when the analyses were repeated using only collateral reports. For pooled effects sizes of conventional AA meeting data all of the RCTs used coerced participants while 8/9 of the non-randomised studies used volunteers. Therefore volunteer-coercion status was nearly completely confounded with non-random-random assignment for conventional AA meetings. Some of the other studies reported in the review also used a mixture of volunteers and coerced participants, which may bias their findings.

Authors' conclusions

Given the widespread use of AA as a treatment for alcoholism, the findings from this meta-analysis are noteworthy in two respects. First is the fact that the randomised trials suggest that AA at best does no better than alternatives and in some cases may do significantly worse. Second is the fact that so little good research on AA exists.

CRD commentary

This is a clearly written review which, although it lacks certain methodological detail (e.g. the number of reviewers involved in the selection, assessment and data extraction of studies), presents a discussion of the problems and biases encountered within this field of work. The review is based on a reasonable search of the literature, and includes both published and unpublished work. The possibility of publication bias however still exists as no specific attempts were made to locate unpublished work. The studies were pooled and the level of heterogeneity assessed, although chi-squared results were not presented for all of the group comparisons. Therefore caution is warranted before interpreting the results of the review.

Implications of the review for practice and research

Practice: The authors state that 'the present results lend at least some weak support to AA-based inpatient treatment, although so few randomised studies of such treatment exist that the results are preliminary at best. Results more clearly support some components of AA, including the use of recovered alcoholics as counsellors, and training in specific AA procedures such as the Twelve-Step program and the honest inventory'. The authors also state that from their results 'it is probably a bad idea to coerce individuals to attend conventional AA meetings' as 'coercion apparently yields significantly worse results than treatment alternatives and non-significantly worse than doing nothing at all'.

Research: The authors recommend that more methodologically sound randomised controlled trials need to be carried out in this area, although they acknowledge and discuss the possible ethical problems associated with this type of study design. The authors also state that 'future studies of AA effectiveness will also benefit by using some holistic measures of outcome'.

Funding

W T Grant Foundation, grant number #94158394.

Bibliographic details

Kownacki R J, Shadish W R. Does Alcoholics Anonymous work: the results from a meta-analysis of controlled experiments. Substance Use and Misuse 1999; 34(13): 1897-1916. [PubMed: 10540977]

Indexing Status

Subject indexing assigned by NLM

MeSH

Alcoholics Anonymous; Alcoholism /rehabilitation; Bias (Epidemiology); Humans; Randomized Controlled Trials as Topic; Temperance; Treatment Outcome

AccessionNumber

11999002138

Database entry date

30/09/2000

Record Status

This is a critical abstract of a systematic review that meets the criteria for inclusion on DARE. Each critical abstract contains a brief summary of the review methods, results and conclusions followed by a detailed critical assessment on the reliability of the review and the conclusions drawn.