**3. Results**

#### *3.1. Matched Gamblers*

The matched gamblers were aged 31.5 (*sd* = 9.5) on average and predominantly male (87%). Money/time spent by the matched gamblers in the preceding 4 weeks amounted to 12.9€ (*sd* = 145.5)/ 3.5h (*sd* = 12.9) on average before the self-exclusion day of their matched self-excluders, and 4.2€ (*sd* = 72.5)/ 1.3h (*sd* = 7.2) at12 months after the end of the self-exclusion period of their matched self-excluders. Account age was 322.33 days on average (*sd* = 445.83).

#### *3.2. Self-Excluders*

The characteristics of the first-time self-excluders and short-duration first-time self-excluders are presented in Table 1. The self-excluders were aged 31 on average and predominantly male. The short-duration first-time self-excluders amounted to 30% of all first-time self-excluders over the 7 years. Money/time spent by self-excluders in the last 4 weeks was 398.5€ (*sd* = 1221.4)/ 32.8h (*sd* = 40.1) before self-exclusion, and 32.3€ (*sd*= 386.6)/6.3h (*sd* = 20.0) at12 months after the end of the self-exclusion period. The mean length of self-exclusion was 614 days (*sd* = 499). Short-duration first-time self-excluders were younger and had a greater financial and time involvement in gambling; their account was one month older on average than in the overall sample.

**Table 1.** Characteristics of first-time self-excluders, and short-duration first-time self-excluders subgroup.


### *3.3. E*ff*ects of Self-Exclusion over 12 Months after the End of the Self-Exclusion Period*

A significant effect of self-exclusion was found for money and time spent over the 12 months after the end of the self-exclusion period (*p*-value for both models < 2.2e−16) using mixed models with a subject random effect (Figure 1).

### *3.4. E*ff*ect of Self-Exclusion over 12 Months after the End of the Self-Exclusion Period among the Most Heavily Involved Gamblers*

The average amount of money spent in the four weeks before and after the self-exclusion period among the gamblers who were the most heavily involved in terms of money, among self-excluders (*n* = 2255) and in the matched group (*n* = 79) are shown in Figure 2. No significant effect of self-exclusions was found on the amounts of money spent (*p* = 0.072) and the effect size was very small (*d* = 0.18).

The average amount of time spent in the 4 weeks before and after the self-exclusion period among the gamblers who were the most heavily involved in terms of time among self-excluders (*n* = 2150) and in the matched group (*n* = 185) are shown in Figure 3. A significant effect of self-exclusion was found for time spent (*p* < 2.2e−16) and the effect size was small (*d* = 0.34).

**Figure 1.** Evolution of money/time spent in the last 4 weeks (€/hours) at baseline and after the end of self-exclusion period (*n* = 4887 and *n* = 4451). (\* = *p*-value < 0.05—ANOVA between the mixed model with and the null model without the interaction of self-exclusion X time).

**Figure 2.** Evolution of money spent (net loss) in the last 4 weeks before and after the self-exclusion period among the gamblers who were the most heavily involved in terms of money (*n* = 2255 and 79 respectively for the self-excluders and the control group of matched gamblers) and time (*n* = 2150 and 185 respectively for the self-excluders and the control group of matched gamblers) (\* = *p-*value < 0.05 —ANOVA between the mixed model with and the null model without the interaction of self-exclusion X time). (\* = *p-value* < *0.05* - ANOVA between the mixed model with and the null model without the interaction of self-exclusion X time).

#### *3.5. Short Self-Exclusions*

Significant effect of short self-exclusion was found for money and time spent over the 12 months after a short self-exclusion (*p*-value in both models <2.2e−16) using mixed models with a subject random effect (Figure 3).

No significant effect of short self-exclusions was found on money/ time spent gambling among the gamblers who were the most heavily involved in terms of money/time (respective *p*-values = 0.873 and 0.491) (Figure 4) but the sizes of the control groups were very small (respectively *n* = 683 vs 18, and *n* = 665 vs 35). The effect size was very small for money spent (*d* = 0.17), and negative and below the very small level for time spent (*d* = −0.09).

**Figure 3.** Evolution of money / time spent in the last 4 weeks (€/hours) before and after a short self-exclusion (*n* = 1460 and 1333). (\* = *p*-value < 0.05—ANOVA between the mixed model with and the null model without the interaction of self-exclusion X time).

**Figure 4.** Evolution of money spent (net loss) in the last 4 weeks before and after a short self-exclusion among the gamblers who were the most heavily involved in terms of money (*n* = 683 and 18 respectively for the self-excluders and the control group of matched gamblers) and in terms of time (*n* = 665 and 35 respectively for the self-excluders and the control group of matched gamblers). (\* = *p*-value < 0.05—ANOVA between the mixed model with and the null model without the interaction of self-exclusion X time).

#### **4. Discussion**

This is the first real life study, reporting comparative follow-up data on voluntary self-exclusion on the initiative of gamblers and including non-self-selected gamblers. This retrospective study analyzed prospectively registered account -based gambling data. The aim was to assess the efficacy of self-exclusion in the long term in term of time and money involvement.

The analysis of account-based gambling data for all first-time self-excluders on a website over 7 years confirmed the efficacy of self-exclusion on gambling outcomes in the long term. The exhaustiveness of this data is a strength that ensures representativeness and power for the statistical analyzes. However, the effect of self-exclusion among the most heavily involved gamblers was found only for the time spent, and not for the money spent, despite a very high level of expenditure before self-exclusion in this subgroup [14]. One important piece of information here is the spontaneous decrease in gambling involvement among gamblers who were the most heavily involved and who did not self-exclude. This result is congruent with a high rate of spontaneous remissions observed in gambling disorder [21]. This result shows the need to provide comparative data, more informative than a tool that is de facto considered to be efficient and promoted by the regulatory authorities [17]. Another interpretation of this decrease among heavy gamblers who did not self-exclude is that the gamblers were not randomized here, and could have chosen to self-exclude if they lacked confidence in their ability to bring about a change in their gambling without an external constraint such as self-exclusion, the reverse being true for non-self-excluders. The efficacy of short self-exclusions among the most heavily involved gamblers was not supported by our data. This is in line with recent experimental data among problem gamblers suggesting no efficacy of very short self-exclusions on gambling outcomes [10]. Another qualitative study reported a preferred duration to ensure efficacy of 12 months from the perspective of problem gamblers who self-excluded [22].

This study presents some important limitations. First, we included only poker gamblers. As discussed in the introduction, poker gamblers present particular cognitive profiles. Moreover, online poker gamblers are younger than other gamblers [4,12], and their history of gambling and associated damages could differ, as well as their motivation to change. The presented results could reflect some of these particularities and not be true in other gambling activities. No data was available on gambling on other online or offline gambling service providers. Gamblers could have just switched from one website to another during the exclusion period. However, all follow-up data reported here concerns gambling after the end of the exclusion period. Gamblers can gamble back on the website after this period and are commercially encouraged and sometimes offered incentives to do so. Moreover, we have already documented in another study that most gamblers return to the initial website to gamble after a self-exclusion [1]. On the other hand, as gambling is regulated in France, gamblers have to provide their Identity Card when opening an account; this measure theoretically prevents from gambling from an account opened under a false identity. The gambling profiles observed are still informative as such, even if not representative of all gambling activities. The use of account-based gambling data is a strength of the study because it enables objective data to be reported. However, it would be interesting to document the effect of self-exclusion on non-gambling outcomes, such as quality of life. No formal diagnosis of gambling disorder and no information on mental disorders or comorbidities were available. We could not, for technical reasons, match the gamblers for the level of gambling involvement in terms of time and money. Our statistical analysis allows for comparisons between the groups by adjusting the mixed model on the subject, which takes into account all subject characteristics including gambling involvement; however, it does not replace a control group with similar involvement in gambling in term of time and money spent. In addition, self-excluders could be different from non-self-excluders in term of the degree of motivation to change, as self-exclusion is a voluntary process in France. Finally, the naturalistic and ecological design of this study of course prevented any randomization process. We therefore report here results on the effect of self-exclusion rather than on efficacy. Further studies could inform on possible response factors to self-exclusion. No information was available on the health care resources used by the gamblers included.

#### **5. Conclusions**

Self-exclusion seems efficient in the long term (i.e., 12 months after the end of the self-exclusion period). However, the effects on money spent as a result of self-exclusion or short self-exclusion should be further explored among the online poker gamblers who are the most heavily involved. A spontaneous, clinically-relevant decrease in gambling activities was demonstrated among most involved gamblers who did not self-exclude. Further study with a randomized design and non-gambling outcomes should be conducted to conclude robustly on the efficacy of short and long self-exclusions in problem gambling, and on response factors.

**Author Contributions:** Conceptualization, A.L., A.D., H.P., S.G. and E.B.; Methodology, A.L., A.D., H.P., S.G. and E.B.; Software, A.L., A.D., H.P., S.G. and E.B.; Validation, A.L.; Formal Analysis, A.L.,A.D., H.P., S.G. and E.B.; Investigation, A.L.; Resources, A.L.; Data Curation, A.L.; Writing—Original Draft Preparation, A.L.; Writing—Review & Editing, A.L., A.B., A.D., H.P., S.G. and E.B.; Supervision, A.B., S.G. and E.B.; Project Administration, A.L.; Funding Acquisition, A.L.

**Funding:** This study received a grant from the "Poste d'Accueil" program exchange between the APHP and the Ecole polytechnique.

**Acknowledgments:** We would like to thank Winamax for allowing access to data through an agreement allowing free analysis and interpretation by our academic team.

**Conflicts of Interest:** A.L. has received sponsorship to attend scientific meetings, speaker honoraria and consultancy fees from Lundbeck, Indivior, and ARJEL. A.B. has received sponsorship to attend scientific meetings, speaker honoraria and consultancy fees from Lundbeck, Mylan, Gilead, Jansenn Cilag and Indivior. A.D., H.P., S.G. and E.B. have no conflict of interest to report.

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

International Journal of *Environmental Research and Public Health*
