Gambling disorder creates a significant public health burden. Despite decades of clinical trials, there are no licensed pharmacological treatments for gambling disorder. Contributing factors to this are the high placebo response rates seen in clinical trials, the heterogeneity of the disorder and high rates of psychiatric comorbidities. Indeed, a number of demographic and clinical variables have previously been associated with altered responses to pharmacotherapy, psychotherapy and placebo. Which variables are likely to predict response to one modality over another remains uncertain. We carried out multiple linear regression analyses in a pooled dataset from six treatment studies in gambling disorder with the aim of identifying predictors of treatment response. Potential predictors were identified a priori through hypothesis and entered into models including all patients, and subsequently for those randomized to active medication or placebo separately. We found that baseline severity of gambling symptoms and number of weeks completed in a trial were predictors of active medication response, while decreased baseline symptoms of anxiety, increased baseline symptoms of depression, and non-Caucasian ethnicity were associated with placebo response. Sensitivity analyses showed that these associations were robust to choices made during the analysis. Further research is required to understand whether controlling for these variables, or using enriched samples, improves assay sensitivity in placebo-controlled clinical trials for gambling disorder.