difficult to know whether this bias over- or underestimates the prevalence of gambling problems. People who are currently gambling might be more likely to be interested in the research and complete the questionnaire; conversely, they might prefer not to disclose their gambling behaviours and thus not respond. However, the PGSI was included in a mail-shot with a number of other questionnaires, and responders completed all questionnaires, which reduces the likelihood that the decision to respond was particularly influenced by the inclusion of the PGSI. Third, the study was limited by the size of the sample at severe risk of problem gambling in the bipolar disorder group (n = 17), which was insufficient for further analysis. The sample size for individuals with major depression was also small (n = 115), so we can have less confidence in the estimated prevalence rates of gambling problems in this group. Fourth, given the exploratory nature of the study we did not control for multiple statistical tests across variables. Therefore, our findings require independent replication. However, some of our statistically significant findings would stand up to correction for multiple comparisons; for example, the associations of moderate and severe risk of problem gambling with suicidal ideation
