To the Editor New treatments—both pharmacologic and nonpharmacologic—for one of the most devastating substance abuse disorders, alcoholism, are direly needed. We were therefore delighted to read an article reporting on a trial of gabapentin as a treatment option for alcohol dependence.1 Unfortunately, this article does not clarify whether gabapentin increases the proportion of abstinent patients or those without heavy drinking. While the research protocol had stipulated a sample size of 150 patients randomized to 1 treatment and 1 placebo control group, the same number of patients was split among 3 groups: 2 active treatment and 1 placebo control group (clinicaltrials.gov identifier: NCT00391716). The authors calculated their sample size based on an acamprosate trial from a different setting.2 Consequently, it seems that the odds ratio for one of the co-primary end points, abstinence, compares the higher dosage group, gabapentin 1800 mg/d, with placebo and was not statistically different (odds ratio, 4.8; 95% CI, 0.9-35.0). This fact is obscured by the article’s focus on a linear dose effect, with a P value below .05 from an extended Mantel-Haenszel χ2 test for linear association. In contrast, no P values were reported for the prespecified analysis. We believe that outcomes should not be selectively reported, and clinical trials should either adhere to their preregistered protocol or explain why they differed from it. Therefore, this present trial should be regarded as a dose-ranging study but not as a pivotal trial, which provides proof with traditional frequentist inference.