To the Editor We read with interest the work of Anton and colleagues1 evaluating gabapentin in the treatment of alcohol use disorder. We are concerned that the authors report a positive primary outcome even though the original preregistered primary outcome was nonsignificant, suggesting that gabapentin is not effective in the way the authors assert.
The authors report a statistically significant decrease in percentage of participants with no heavy drinking days. They define this by 2 criteria: (1) self-report of no heavy drinking days and (2) no reading on the percentage of disialo carbohydrate-deficient transferrin (%dCDT) greater than 1.7% at any point in the trial. However, in both the original preregistered outcome and the supplementary protocol, a %dCDT threshold is not part of the primary outcome definition. There was no significant difference between gabapentin and placebo using the self-report criterion. Only when the %dCDT threshold was applied did the difference between the groups become statistically significant.