- Ditman 1967. This is one of the earliest studies on AA, done before we knew how to properly study AA. It was a RCT -- a "randomized controlled trial." A RCT doesn't work with AA because 1) The people assigned to AA do not necessarily work the steps as suggested by oldtimers in AA meetings 2) The people not assigned to AA who are serious about getting sober end up going to the program anyway.
- Brandsma 1980. Like Ditman 1967, this was a RCT, with the same problems. In addition, there is an opt-repeated claim that this study shows AA results in increased binge drinking. However 1) The study looked at about three dozen different findings at multiple follow-up periods, but this result was only seen in this one finding at one follow-up period 2) This result has not been replicated in recent studies  even though 3) The study is 36 years old.
- The 1990 AA Triennial survey. This was not a scientific survey; this was an informal survey AA sent to members. Of the people in their first year of recovery, about 5% were in their 12th month of recovery, which is lower than the 8.3% rate we expect to see if AA had a 100% retention rate. Statistical extrapolations show that this survey shows a 29% retention rate; the question I ask is this: Is it AA's fault that 71% of newcomers decide to not come back to an AA meeting?
Now, this study showing that it's not just self selection is just one study, and we need to have further studies to confirm that AA itself (and not just self-selection) helps keep alcoholics sober, but that's the direction the evidence points to right now. I would worry about these results being in just one study if the study was 36 years old, but this study is not even two years old.
Bottom line: If you have a drinking problem, and want to get and stay sober, AA is an effective way to do just that.
 Edit: I need to point out Ogborne 1982, but this study is most likely just observing the fact that AA members are more likely to be honest about the negative consequences of their drinking.