Background and aims: Flaws in previous studies mean that findings of J-shaped risk curves for gambling should be disregarded. The current study aims to estimate the shape of risk curves for gambling losses and risk of gambling-related harm (a) for total gambling losses and (b) disaggregated by gambling activity.
Design: Four cross-sectional surveys.
Setting: Nationally representative surveys of adults in Australia (1999), Canada (2000), Finland (2011) and Norway (2002).
Participants: A total of 10 632 Australian adults, 3120 Canadian adults, 4484 people aged 15-74 years in Finland and 5235 people aged 15-74 years in Norway.
Measurements: Problem gambling risk was measured using the modified South Oaks Gambling Screen, the NORC DSM Screen for Gambling Problems and the Problem Gambling Severity Index.
Findings: Risk curves for total gambling losses were estimated to be r-shaped in Australia {β losses = 4.7 [95% confidence interval (CI) = 3.8, 6.5], β losses(2 =) -7.6 (95% CI = -17.5, -4.5)}, Canada [β losses = 2.0 (95% CI = 1.3, 3.9), β losses(2 =) -3.9 (95% CI = -15.4, -2.2)] and Finland [β losses = 3.6 (95% CI = 2.5, 7.5), β losses(2 =) -4.4 (95% CI = -34.9, -2.4)] and linear in Norway [β losses = 1.6 (95% CI = 0.6, 3.1), β losses(2 =) -2.6 (95% CI = -12.6, 1.4)]. Risk curves for different gambling activities showed either linear, r-shaped or non-significant relationships.
Conclusions: Player loss-risk curves for total gambling losses and for different gambling activities are likely to be linear or r-shaped. For total losses and electronic gaming machines, there is no evidence of a threshold below which increasing losses does not increase the risk of harm.
Keywords: Electronic gaming machines; dose-response; gambling; gambling expenditure; gambling losses; gambling-related harm; problem gambling; public health; risk curves.
© 2015 Society for the Study of Addiction.