“Why do large observational studies such as that of Fiorentine and Moos seem to suggest that AA is effective, while smaller controlled studies like those of the Brandsma, Walsh, and others included in the Cochrane Review do not? The likely explanation is simple: people stay in AA if they’re getting better and leave if they aren’t. This is understandable. If you are able to stop drinking, then continuing to attend AA is a comfortable and affirming choice. If you struggle with drinking and can’t make use of the AA approach, then you are less likely to keep attending. Over the long term, the people who remain in AA are, by definition, the success stories. But they represent a very small slice of the people who start there; as we will see shortly, the dropout rate from AA is extremely high. These facts—that AA works for the diehards and fails for the dropouts—are perennially misunderstood by the press and even by some researchers. Proponents of the program proudly point to the fact that people who stay in AA tend to be sober, ignoring the tautological nature of this claim. Reviewing this logic, Harvard biostatistics professor Richard Gelber said, “The main problem is the self-fulfilling prophesy: the longer people stick with AA the better they are, hence AA must be working. It is like saying the longer you live, the older you will be when you die.” As we will soon see, this fundamental error in logic undergirds nearly every claim of AA’s efficacy.”