Deriving low-risk gambling limits from longitudinal data collected in two independent Canadian studies.

AIMS: To derive low-risk gambling limits using the method developed by Currie et al. (2006) applied to longitudinal data.

DESIGN: Secondary analysis of data from the Quinte Longitudinal Study (n = 3054) and Leisure, Lifestyle and Lifecycle Project (n = 809), two independently conducted cohort studies of the natural progression of gambling in Canadian adults.

SETTING: Community-dwelling adults in Southeastern Ontario and Alberta, Canada.

PARTICIPANTS: A total of 3863 adults (50% male; median age = 44) who reported gambling in the past year.

MEASUREMENTS: Gambling behaviours (typical monthly frequency, total expenditure and percentage of income spent on gambling) and harm (experiencing two or more consequences of gambling in the past 12 months) were assessed with the Canadian Problem Gambling Index.

FINDINGS: The dose-response relationship was comparable in both studies for frequency of gambling (days per month), total expenditure and percentage of household income spent on gambling (area under the curve values ranged from 0.66 to 0.74). Based on the optimal sensitivity and specificity values, the low-risk gambling cut-offs were eight times per month, $75CAN total per month and 1.7% of income spent on gambling. Gamblers who exceeded any of these limits at time 1 were approximately four times more likely to report harm at time 2 [95% confidence interval (CI) = 2.9-6.6].

CONCLUSIONS: Longitudinal data in Canada suggest low-risk gambling thresholds of eight times per month, $75CAN total per month and 1.7% of income spent on gambling, all of which are higher than previously derived limits from cross-sectional data. Gamblers who exceed any of the three low-risk limits are four times more likely to experience future harm than those who do not.