Commentary on Currie <i>et al</i> . (2017): Low‐risk gambling limits—a bridge too far?

Currie et al. conclude that low-risk gambling limits can be developed. The challenges, however, are underestimated and may be unsurmountable. This in no way negates the value of increasing research on the largely neglected topic of linkages between gambling participation and gambling-related harm. Currie et al.'s study 1 draws upon data from large prospective studies and advances previous research on this topic. They develop guidelines using general participation measures (times participated per month, monthly gambling expenditure, percentage of income spent on gambling), but caution against premature application. Unlike alcohol, there is no standard gambling unit. Gambling forms and settings differ in ways that elude simple classification. Of note is the wide variation in gambling type ‘toxicity’. A recent prospective study found that more than a third of monthly electronic gaming machine participants became at-risk or problem gamblers 12 months later 2. They were 10.6 times more likely to become so than people who did not take part this often. Bingo [odds ratio (OR) = 9.4] was the only other form coming close to this. ORs for other forms ranged from 1.7 to 4.9. ORs for Currie et al.'s 1 risk factors ranged from 1.4 to 3.1, and together accounted for less than 20% of harm variance. Enhanced prediction will almost certainly require inclusion of particular gambling forms and, perhaps, settings. Some other measures not considered by Currie et al. are also strong predictors, e.g. length of gambling sessions and number of activities engaged in 2. Additionally, measures are inter-related, often in complex ways. Multivariate analyses have found that both general and form-specific measures uniquely predict harmful gambling 2, 3. While participation measures are the strongest harm predictors, additional factors are implicated 2, 4-7. It seems likely that different thresholds will apply to different groups, especially high-risk population sectors such as youth and some indigenous, ethnic and migrant groups 2. Currie et al. 1 acknowledge that non-gambling factors may need to be considered with ‘additional guidelines and cautionary statements for more vulnerable populations’. Past problem gamblers come into this category. In some jurisdictions, relapses constitute a third to two-thirds of problem gambling incident cases 2, 6, 7, and research suggests that they will require different thresholds 6, 8. There is a paucity of research linking participation parameters to gambling-related harm. Even in the alcohol field, where there is a substantial body of relevant literature, the identification of limits is not straightforward and they vary widely across jurisdictions. Studies seeking to identity low-risk gambling limits have used problem gambling screens rather than comprehensive harm measures. Currie et al. 1 used seven Problem Gambling Severity Index (PGSI) items that cover only three harm domains. Recent research has identified seven domains 9. Only the latest Currie et al. study used prospective data, and that was limited to 12–19 months. Currie et al. 1 include people who report one harm item in the reference group. Had all 10 PGSI items been administered, substantially more would have been in the PGSI low-risk category (score of 1 or 2) and some would have been moderate-risk gamblers. Browne et al. 10 found that 48% of harm was attributed to low-risk gamblers. If harm is to be reduced appreciably, measures will need to be taken to prevent people from moving into this category. In the unlikely event that limits can be specified clearly, it is difficult to predict what effect they might have. In the case of alcohol, where units are expressed simply as standard drinks, studies show that people have difficulty understanding and applying them 11. Many jurisdictions have introduced measures to reduce gambling-related harm. Reviews conclude that the most widely implemented are least likely to be effective 12. Hancock & Smith 13 claim that the Reno Model, a widely promulgated responsible gambling (RG) framework, provides justifications for implementing largely ineffective policies. The model has an emphasis on individual responsibility. The pursuit of low-risk limits coheres strongly with principles underlying the Reno Model, including the belief that gambling can be engaged in at safe levels. As mentioned, Currie et al. focus upon generic measures rather than on ‘toxicities’ of individual gambling forms. They report J-shaped risk curves 1. Not only does this suggest relatively safe participation levels, it implies that moderate participation is protective. Markham et al. 14 also examined risk curves associated with gambling expenditure (losses) overall, as well as separately for five gambling forms in four countries. Curves were mainly either linear or R-shaped and varied somewhat across forms and countries. In the case of R-shaped curves, risk increases more at low to moderate expenditure levels and then attenuates. There may be no safe limits for many gambling forms. This does not mean that the research agenda proposed by Currie et al. 1 should be abandoned. On the contrary, it addresses a neglected area and could inform people of risks associated with different gambling activities and intensities. A meaningful reduction in gambling-related harm will, however, probably require a multi-faceted approach that addresses supply and demand reduction and includes the full range of modifiable risk and protective factors at individual, community and wider social levels. The author received financial assistance from the Canadian Centre on Substance Abuse to participate in a meeting of the Low-Risk Gambling Guidelines Working Group in Canada. The author anticipates working with members of this group to construct risk curves using data from a number of different countries. He does not believe that this compromises his capacity to reflect critically and impartially on matters related to the topic of this commentary. The author acknowledges Dr Komathi Kolandai-Matchett for assistance in formatting and proofing this article.

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.