Introduction

Originating in the alcohol treatment literature, the highly contested issue of abstinence versus moderation has also been debated in the gambling treatment literature. Traditionally, complete abstinence from gambling as a desired treatment goal and as an outcome measure of success has been championed as the only legitimate pathway to recovery (Blaszczynski et al. 1991; Ladouceur 2005; Ladouceur et al. 2009; Slutske et al. 2010). Indeed, Gambler’s Anonymous (GA) is a popular treatment option that is rooted in a zero tolerance, twelve-step model, and studies in the gambling literature that have evaluated other treatment modalities have often used abstinence from all types of gambling as the index of successful treatment (Blaszczynski et al. 1991; Ladouceur 2005; Petry 2005). Beyond the gambling literature, the endorsement of complete abstinence goals among clients in treatment for alcohol, opiate, nicotine, and cocaine dependence has been found to be associated with reduced risk for relapse compared to the endorsement of less restrictive goals (Hall et al. 1990, 1991).

Counter to the traditional view, it has been thought that offering the option of non-abstinence treatment goals might provide a more realistic and attractive option for problem gamblers, which might also lead to lower attrition rates by increasing self-efficacy and motivation early in treatment (Blaszczynski et al. 1991; Ladouceur et al. 2009; Rosecrance 1986, 1989). Research has demonstrated the viability of non-abstinence treatment goals and has supported the notion that moderated gambling is a popular road to recovery among both naturally recovered and treatment-assisted problem gamblers (Blaszczynski et al. 1991; Dickerson et al. 1990; Dickerson and Weeks 1979; Dowling et al. 2009; Ladouceur et al. 2009; Rankin 1982; Robson et al. 2002; Slutske et al. 2010; Weinstock et al. 2007).

Despite the high stakes and theoretical controversy about abstinence versus moderation goals in gambling treatment, very few empirical studies have closely examined the nature and impact of goal selection in the treatment of gambling disorders. Investigations into this issue are burdened by the fact that there is no standardized operational definition of what exactly constitutes a moderation goal with respect to gambling (see Ladouceur 2005; Currie et al. 2008a, b), and terms such as “moderated gambling”, “controlled gambling”, and “reduced gambling” are often used interchangeably in both research and practice. Nevertheless, a handful of studies have helped to shed light on this topic.

For instance, studies have shown that a number of problem gamblers select non-abstinence goals when they are available (Blaszczynski et al. 2005; Dowling and Smith 2007; Toneatto and Dragonetti 2008), and Dowling and Smith (2007) have investigated the reasons behind goal selection. Dowling and Smith concluded that among treatment-seeking female pathological gamblers in Australia, the most common reason for selecting moderated gambling was that abstinence was perceived as unrealistic or overwhelming, whereas the most common reason for selecting abstinence was the belief that gambling behaviour cannot be controlled. Dowling and Smith found few differences in demographic, gambling, and psychosocial client characteristics between those that selected abstinence and those that selected moderation. There were only two differences found between the groups; namely, that those who endorsed moderated gambling goals were significantly older and more likely to hold the belief that recovery does not require abstinence. It is noteworthy that in contrast to Dowling and Smith’s finding that older age is associated with the selection of moderated gambling goals, the alcohol treatment literature has generally found that the selection of moderated drinking is associated with younger age (Booth et al. 1984; Hodgins et al. 1997). Interestingly, Dowling and Smith’s findings also did not support the association between moderated gambling and lower gambling problem severity, which is in contrast to the alcohol treatment literature (Adamson and Sellman 2001; Hodgins et al. 1997; Ogborne 1987) as well as a gambling treatment study by Toneatto and Dragonetti (2008). Toneatto and Dragonetti reported that those who selected abstinence at the beginning of treatment had greater number of baseline Diagnostic and Statistical Manual of Mental Disorders (DSM-IV; American Psychiatric Association 1994) symptoms for pathological gambling, attended more treatment sessions, and had higher abstinence rates at the end of treatment and 12-month follow-up versus those who selected moderation at pretreatment.

Only one study in the literature has examined the dynamic nature of participant goal selection in gambling disorders over time (Ladouceur et al. 2009). This study demonstrated that goal selection is fluid, which is consistent with research demonstrating that goal selection in alcohol treatment programs is also often fluid (Cox et al. 2004; Hodgins et al. 1997). Specifically, Ladouceur et al. recruited participants for a moderated gambling cognitive-behavioural treatment and found that 52 % of program completing participants kept their initial goal of moderated gambling and 48 % shifted to abstinence at least once during the 12-months follow-up. Compared to participants who did not shift their goal from moderated gambling, participants who shifted their goal to abstinence at least once were more likely to have consulted or received help regarding their gambling activities in the past, and they were more likely to have heard about the program via information available from health care professionals and gambling treatment centres. No other socio-demographic differences between shifters and non-shifters were found. In terms of outcomes, among program completers, participants who kept their initial goal of moderated gambling were more likely to have gambled at least once at post-treatment and at 6-months follow-up; and participants who shifted their goal to abstinence were more likely to have reported having often succeeded in respecting their own gambling limits (gambling sessions and money spent) and had higher mean self-efficacy perception scores.

In light of the dearth of research on this topic, and to advance the literature with a view towards informing clinical practice, more research is required to elucidate the nature and impact of participant goal selection in the treatment of gambling disorders. In this paper, we examined the nature and impact of goal selection in archival data from a randomized controlled trial (RCT) of a brief motivational intervention for pathological gambling (Hodgins et al. 2009). This paper is only the second in the literature to report the dynamics of goal selection in pathological gambling over the course of treatment. Specifically, the purpose of this paper was to answer three research questions: (1) What are the correlates of goal selection at pretreatment? (2) What is the temporal stability of goal selection over the course of follow-up? (3) How does goal selection influence gambling outcomes over the course of follow-up?

Methods

Participants

Participants were recruited for participation in an intervention study via media announcements soliciting individuals who were concerned about their gambling and wanted to cut down or quit on their own. Inclusion criteria were as follows: 18 years of age or older, perception of a gambling problem, met criteria to be considered a pathological gambler, gambled in the past month, not involved in treatment at present, willingness to read a short book written in English (to ensure reading ability), willingness to provide follow-up data, willingness to provide the name of a collateral to help locate them for follow-up interviews and the name of the same or a different collateral for data validation, and willingness to have telephone contacts recorded. In total, 314 participants met the inclusion criteria and participated in the study. The sample had a mean age of 47.5 years (SD = 11.0), and 55.4 % were women. Culture group was identified as 85.0 % Canadian and 15.0 % other. The mean education level including years of grade school and higher education was 15.1 years (SD = 1.8) and 71.0 % were employed. In terms of marital status, 46.2 % were married or living in common-law marriages. The mean SOGS score was 11.1 (SD = 3.0). The types of gambling causing major problems were video lottery terminal play (76.1 %), slot machines (50.6 %), casino games (12.7 %), bingo (6.1 %), lotteries (4.8 %), horse racing (3.5 %), Nevada tickets (2.2 %), sports pools (1.6 %), Keno (1.6 %), card games (1.0 %), and speculative investments (1.0 %). Less than half of the sample (41.7 %) reported previous treatment, mostly Gamblers Anonymous (39.5 %). In the month before the study, participants reported gambling a mean of 7.6 days (SD = 6.7) and losing a mean of $1,460.03 Canadian.

Procedure

Of the 314 eligible participants, 83 (26.4 %) were randomly assigned to a combined telephone motivational intervention and self-help workbook condition, 84 (26.8 %) were assigned to a combined telephone motivational intervention and self-help workbook plus telephone booster sessions condition, 82 (26.1 %) were assigned to a self-help workbook-only condition, and 65 (20.7 %) were assigned to a wait-list. The telephone motivational intervention was based on MI principles and the self-help workbook (Becoming a Winner: Defeating Problem Gambling; Hodgins and Makarchuk 2002) was based on a cognitive-behavioural model and was designed to be brief and nontechnical.

Participants assigned to the combined telephone motivational intervention and self-help workbook condition were initially assessed by a research assistant over the telephone, called by a therapist for the motivational interview, and then mailed a personalized handwritten note of encouragement along with the self-help workbook. Participants assigned to the wait-list were informed by a research assistant at the initial telephone assessment that there would be a 6-week delay before the self-help workbook would be mailed, and these participants did not receive any therapist contact. All participants were followed-up 3-, 6-, 9-, and 12-months after their initial interview at pretreatment. Of the 314 participants assessed at pretreatment, 279 (89 %) were followed-up at 3-months, 267 (85 %) at 6-months, 263 (84 %) at 9-months, and 267 (85 %) at 12-months.

Measures

Demographic and Descriptive Information

A demographic profile and gambling history were obtained, including past gambling treatment and a timeline interview of types of gambling, frequency, and money spent for the past month (Hodgins and Makarchuk 2003; Sobell and Sobell 1992). At pretreatment, gambling problem severity was assessed with the South Oaks Gambling Screen (SOGS; Lesieur and Blume 1987); participants were asked to predict how likely they will succeed in overcoming their gambling problem in the next 6 months and 12 months ranging from 0 (not at all successful) to 10 (extremely successful); participants were asked to predict the difficulty in overcoming their gambling problem ranging from 0 (not very difficult) to 10 (very difficult); and participants were asked to describe their control over their gambling habits ranging from 0 (no control) to 10 (complete control). Additionally, participants were asked to rate their motivation to overcome their gambling problem ranging from 0 (not at all motivated) to 10 (extremely motivated) at pretreatment, 3-, 6-, 9-, and 12-months follow-up.

Participant Goal Selection

Two qualitatively different goal selection questions were posed to participants, and thus, two types of goal selection variables (dichotomous vs. nuanced) are reported in this study. Participants were asked for a dichotomous treatment goal (i.e., to quit problem gambling vs. to cut back on problem gambling) at pretreatment, and were then asked to provide a more nuanced treatment goal (to quit all types of gambling vs. to quit problem gambling type vs. to gamble in a controlled manner) for the remaining follow-up periods (at 3-, 6-, 9-, and 12-months follow-up).

Gambling Outcomes

In this paper, the following gambling outcomes were examined at pretreatment, and at 3-, 6-, 9-, and 12-months follow-up: average number of days, dollars and dollars per day gambled, and perceived goal achievement. In terms of the first three outcomes, it is noteworthy that average number of days gambled and average number of dollars gambled might be regarded as confounded with the goal selection variables. For example, for a participant that endorses the moderation-based goals (‘to cut back on problem gambling’ and ‘to gamble in a controlled manner’), it is difficult to ascertain whether the goal is focused towards the number of days or dollars gambled, or both. However, the outcome of average number of dollars per day gambled is less confounded with goal selection, since it is more clear that the desire to cut back or control gambling would represent a desire to decrease expenditures.

In terms of perceived goal achievement, participants were asked at each follow-up time period to rate the degree to which they met their goal since the last assessment period using four qualifiers: not at all, partially, mostly, completely.

Results

What is the Temporal Stability of Goal Selection Over the Course of Treatment?

Of all 314 participants, complete goal selection data were available for 247 (78.7 %) participants over all five time periods (pretreatment, 3-, 6-, 9-, and 12-months follow-up). Table 1 presents participant goal selection over the time periods. At pretreatment, the majority (86.0 %) of participants endorsed an abstinence goal (i.e., to quit problem gambling) as compared to a minority of participants (14.0 %) who endorsed a moderation goal (i.e., to cut back on problem gambling). However, since participants were asked for more nuanced treatment goals at 3-, 6-, 9-, and 12-months follow-up, the pattern of goal selection beyond pretreatment was different, whereby the most popular treatment goal at each follow-up period was to ‘quit problem gambling type’.

Table 1 Participant goal selection over time
Full size table

Nuanced goal selection was fluid over the follow-up periods for many participants. That is, over the follow-up periods—with data missing from 23 (7.9 %) participants over all follow-up periods—almost half of participants (45.7 %) switched their goal at least one time, whereby 23.4 % switched their goal once and an additional 22.3 % switched their goal more than once. At the same time, some temporal stability was observed, whereby 26.1 % of participants endorsed a continuous goal of ‘quit problem gambling type’, 17.9 % endorsed a continuous goal of ‘quit all types of gambling’, and 10.3 % endorsed a continuous goal of ‘gamble in a controlled manner’.

What are the Correlates of Goal Selection at Pretreatment?

Table 2 displays the correlates of the dichotomous goal selection variable at pretreatment. With respect to gambling types, only video lottery terminal play, slot machines, casino games, and bingo are displayed in the table, since these were among the top four most frequent gambling types identified as problematic in the study. As can be gleaned from the table, nine variables were significantly associated with pretreatment goal. Specifically, compared to participants who selected the goal ‘cut back on problem gambling’, participants who selected the goal ‘quit problem gambling’ were more likely to be younger, have higher SOGS scores, identify video lottery terminal play as problematic, have lower number of days gambled, have greater number of dollars per day gambled, have greater motivation to overcome their gambling problem, have greater levels of predicted success at 6 and 12 months, and have lower levels of perceived control over gambling habits. These nine variables were included as a predictors of dichotomous pretreatment goal selection in a simultaneous entry binary logistic regression analysis. The overall model was statistically significant (−2 Log likelihood = 196.3, χ 2(9) = 57.9, p < .001) and had a classification percentage of 85.3 %, which did not differ from the base rate classification accuracy because of the highly unequal number of participants who selected to ‘quit problem gambling’ at pretreatment (n = 270) versus those who selected to ‘cut back on problem gambling’ (n = 44). Only three variables were found to uniquely predict pretreatment dichotomous goal selection: SOGS score (b = −0.2, Wald = 10.1, p < .01, Exp(b) = 0.80), problematic video lottery terminal play (b = 0.9, Wald = 5.6, p < .05, Exp(b) = 2.50), and motivation to overcome the gambling problem (b = −0.3, Wald = 10.0, p < .01, Exp(b) = 0.72). As indicated by the odds ratio, for each one point increase in SOGS scores and motivation to overcome the gambling problem, the chances of participants at pretreatment endorsing the ‘quit problem gambling’ goal increased by 0.80 and 0.72 times, respectively; and participants were 2.5 times more likely to endorse the ‘quit problem gambling’ goal if they had identified video lottery terminal play as problematic at pretreatment.

Table 2 Predictors of pretreatment goal selection (n = 314)
Full size table

How Does Goal Selection Influence Gambling Outcomes Over the Course of Treatment?

Days Gambled

To test the influence of dichotomous goal selection at pretreatment on average number of days gambled at 3-months follow-up, an ANCOVA was conducted with dichotomous goal selection entered as the predictor; to control for potential confounding variables, we examined pretreatment past month days gambled, SOGS scores, problematic video lottery terminal play, treatment condition, and pretreatment motivation to overcome the gambling problem as covariates. After controlling for the effects of the confounding variables, the analysis did not reveal a main effect of goal selection. However, significant covariate effects were found for pretreatment past month days gambled, F (1, 277) = 69.6, p < .001, and motivation to overcome the gambling problem, F (1, 277) = 6.1, p < .05, indicating that fewer pretreatment past month days gambled and greater motivation to overcome the gambling problem were related to fewer average number of days gambled at 3-months follow-up.

To test the influence of nuanced goal selection at 3-, 6-, and 9-months as a predictor of average number of days gambled at 6-, 9-, and 12-months, respectively, mixed-model random regression analyses using Type III sum of squares were conducted. A test of the reliability of the unconditional mean model on the data set revealed an intra-class correlation coefficient of 0.58, which is above 0.25, indicating that individual growth curve analyses are warranted over traditional methods (Shek and Ma 2011). A test of the unconditional linear growth curve model revealed a significant linear decrease in the average number of days gambled over 6-, 9-, and 12-month follow-up periods, t (266.0) = 2.8, p < .01, indicating a time effect. When predictors were added to the model, the mixed model regression analysis revealed that a first-order autoregressive covariance structure yielded the best-fitting model. To test the predictor effect of nuanced participant goal selection on average number of days of gambled over 6-, 9-, and 12-months follow-up, nuanced participant goal selection was examined as a time-varying covariate at 3-, 6-, and 9-months follow-up. Additionally, to control for potential confounding variables, we examined pretreatment past month days gambled, pretreatment SOGS scores, problematic video lottery terminal play, and treatment condition as time-invariant covariates, as well as motivation to overcome the gambling problem as a time-varying covariate at 3-, 6-, 9-months follow-up. After controlling for the effects of the confounding variables, the analysis revealed a significant main effect of time, F (2, 519.3) = 6.4, p < .01, and goal selection, F (3, 752.5) = 12.6, p < .001. The estimated marginal means of the average number of days gambled over the course of the follow-up periods were 2.4 (SEM = 0.3, df = 513.7) for participants who endorsed ‘quit all types of gambling’, 2.8 (SEM = 0.2, df = 408.4) for participants who endorsed ‘quit problem type of gambling’, and 4.2 (SEM = 0.3, df = 606.9) for participants who endorsed ‘gamble in a controlled manner’. Follow-up pairwise comparisons on the estimated marginal means revealed that participants who endorsed ‘quit all types of gambling’ gambled significantly fewer days over the follow-up periods compared to participants who endorsed ‘gamble in a controlled manner’ (p < .001), and did not statistically differ with participants who endorsed ‘quit problem gambling type.’ Participants who endorsed ‘quit problem gambling type’ gambled significantly fewer days over the follow-up periods compared to participants who endorsed ‘gamble in a controlled manner’ (p < .001). The time by participant goal selection interaction was non-significant. In addition, there were main effects of pretreatment past month days gambled, F (1, 210.5) = 40.0, p < .001, pretreatment SOGS scores, F (1, 307.2) = 6.2, p < .05, and motivation to overcome the gambling problem over time, F (1, 751.6) = 13.9, p < .001, indicating that greater pretreatment past month days gambled, greater pretreatment SOGS scores, and lower motivation scores over 3-, 6-, and 9-months follow-up were related to greater average number of days gambled at 6-, 9-, and 12-months follow-up.

Dollars Gambled

The same analytic approach used for days gambled was used for dollars and dollars per day gambled. The ANCOVA to test the influence of dichotomous goal selection at pretreatment on dollars gambled at 3-months follow-up did not reveal a main effect of goal selection. However, significant covariate effects were found for pretreatment SOGS scores, F (1, 275) = 9.9, p < .01, and past month dollars gambled, F (1, 275) = 26.9, p < .001, indicating that greater pretreatment SOGS scores and greater past month dollars gambled were related to greater average number of dollars gambled at 3-months follow-up.

Regarding the influence of nuanced goal selection at 6-, 9-, and 12-months, the test of the reliability of the unconditional mean model revealed an intra-class correlation coefficient of 0.30, indicating that individual growth curve analyses are warranted. The test of the unconditional linear growth curve model did not reveal a significant linear change in the average number of dollars gambled over the follow-up periods, indicating that no time effect was observed. When predictors were added to the model, the mixed model regression analysis revealed that a first-order autoregressive covariance structure yielded the best-fitting model. After controlling for the effects of the confounding variables, the analysis revealed no main effects for time or time varying goal selection. However, there were main effects of pretreatment past month dollars gambled, F (1, 249.5) = 65.2, p < .001, and motivation to overcome the gambling problem, F (1, 598.9) = 20.9, p < .001, indicating that greater pretreatment past month dollars gambled and lower motivation scores over the follow-up periods were related to greater average number of dollars gambled.

Dollars Per Day Gambled

The ANCOVA to test the influence of dichotomous goal selection at pretreatment on dollars per day gambled at 3-months follow-up did not reveal a main effect of goal selection but significant covariate effects were found for pretreatment past month dollars per day gambled, F (1, 238) = 36.5, p < .001, pretreatment SOGS scores, F (1, 238) = 9.8, p < .01, and problematic video lottery terminal play, F (1, 238) = 5.2, p < .05, indicating that greater pretreatment past month dollars per day gambled and greater pretreatment SOGS scores were both related to greater average number of dollars per day gambled at 3-months follow-up; and indicating that participants who identified video lottery terminal play as problematic gambled significantly fewer dollars per day (M = $120.50) at 3-months follow-up than participants who did not identify video lottery terminal play as problematic (M = $213.01).

Regarding the influence of nuanced goal selection at 3-, 6-, and 9-months as a predictor of dollars per day gambled at 6-, 9-, and 12-months, the test of the reliability of the unconditional mean model revealed an intra-class correlation coefficient of 0.30, indicating that individual growth curve analyses are warranted. The test of the unconditional linear growth curve model did not reveal a significant linear change in the average dollars per day gambled over the follow-up periods. When predictors were added to the model, the mixed model regression analysis revealed that a first-order autoregressive covariance structure yielded the best-fitting model. After controlling for the effects of the confounding variables, the analysis revealed no main effects for time or goal selection. There was a main effect of pretreatment past month dollars per day gambled, F (3, 170.8) = 4.1, p < .001, indicating that greater pretreatment past month dollars per day gambled was related to greater average number of dollars per day gambled over the follow-up periods.

Perceived Goal Achievement

Table 3 displays participants’ perceived goal achievement across the follow-up periods and shows that approximately half of participants reported that they mostly and completely met their goal across each follow-up period. To test the influence of dichotomous goal selection at pretreatment on perceived goal achievement at 3-months follow-up, an ordinal logistic regression was conducted. To control for potential confounding variables, we examined pretreatment SOGS scores and pretreatment motivation to overcome the gambling problem as covariates, and problematic video lottery terminal play, and treatment condition, as factors. While the model fitting information was non-significant, suggesting that the model was not improved by adding the predictors, the goodness-of-fit index was also non-significant, suggesting a good fit of the data to the model. The Nagelkerke R 2 indicated that all predictors accounted for only 4.2 % of the variance in the model. After controlling for the effects of the confounding variables, the analysis did not reveal a main effect of goal selection. However, a significant main effect was found for pretreatment motivation to overcome the gambling problem, Wald (1) = 5.1, p < .05, indicating that greater pretreatment motivation to overcome the gambling problem was related to participants perceiving greater goal achievement at 3-months follow-up.

Table 3 Perceived goal achievement across all the follow-up periods
Full size table

To test the influence of nuanced goal selection at 3-, 6-, and 9-months as a predictor of perceived goal achievement at 6-, 9-, and 12-months, respectively, a generalized estimating equation (GEE) ordinal logistic (multinomial distribution and cumulative logit link function) analysis was conducted, specifying an independent working correlation matrix structure. Goal selection was examined as a time-varying covariate at 3-, 6-, and 9-months. Additionally, to control for potential confounding variables, we examined pretreatment SOGS scores, problematic video lottery terminal play, and treatment condition as time-invariant covariates, as well as motivation to overcome the gambling problem as a time-varying covariate. After controlling for the effects of the confounding variables, the analysis revealed a significant main effect of time, Wald χ2 (2) = 9.5, p < .01, indicating that participants perceived greater goal achievement over the follow-up periods. No main effect for goal selection was observed. In addition, there were main effects of pretreatment SOGS scores, Wald χ2 (1) = 4.4, p < .05, and motivation to overcome the gambling problem over time, Wald χ2 (1) = 84.6, p < .001, indicating that greater pretreatment SOGS scores were related to participants perceiving less goal achievement, and indicating that greater motivation to overcome the gambling problem over time was related to participants perceiving greater goal achievement.

Finally, while participants’ actual moderation-based goals were not measured in the present study, Table 4 displays the mean days, dollars, and dollars per day gambled for participants with moderation-based goals who reported that they mostly and completely met their goals across the follow-up periods. Presumably, these values might reflect participants’ actual moderation-based goals. As can be seen in the table, for those participants who endorsed moderation-based goals and reported that they mostly achieved their goal, the average days gambled between assessment periods ranged from 3.3 to 7.1 days, the average dollars gambled ranged from $109.92 to $286.07, and the average dollars per day gambled ranged from $24.11 to $73.86. For those participants who endorsed moderation-based goals and reported that they completely achieved their goal, the average days gambled between assessment periods ranged from 1.4 to 2.8 days, the average dollars gambled ranged from $86.29 to $229.18 (not including the net win of $2.11 reported at 6-months), and the average dollars per day gambled ranged from $32.80 to $120.96.

Table 4 Mean days, dollars and dollars per day gambled for participants with moderation-based goals who mostly and completely met their goals across the follow-up periods
Full size table

Discussion

The present study examined the nature and impact of participant goal selection in the treatment of pathological gambling and was only the second in the literature to have reported on the dynamic nature of goal selection over time. This secondary analysis of a RCT of brief motivational gambling intervention demonstrated that the pattern of goal selection over time could be characterized by both fluidity and stability, and that goal selection at pretreatment could be uniquely predicted by gambling problem severity, problematic video lottery terminal play, and motivation to overcome the gambling problem. Moreover, the results demonstrated that goal selection over time had an impact on the average number of days gambled, but not the average number of dollars gambled, the average number of dollars per day gambled, or perceived goal achievement.

More specifically, a novel contribution of the present study was the examination of the temporal stability of goal selection over the course of treatment. Our results demonstrated that when seeking to overcome their gambling problem, most participants first considered an abstinence-based goal. Belief in the need for abstinence in overcoming gambling problems is also endorsed by the general public. Similar to beliefs about alcohol problems, Cunningham et al. (2011) found that 71 % of respondents in a general population survey believed that quitting was necessary for successfully overcoming a gambling problem.

Our results also demonstrated that goal selection was fluid for almost half of participants, which is consistent with the alcohol treatment literature (Cox et al. 2004; Hodgins et al. 1997) as well as the only other study in the gambling literature that reported on goal selection over time (Ladouceur et al. 2009). Some stability in goal selection was also observed, however, whereby over 25 % of participants selected a continuous goal of ‘quit problem gambling type’, almost 20 % selected a continuous goal of ‘quit all types of gambling’, and approximately 10 % selected a continuous goal of ‘gamble in a controlled manner.’ This pattern to the fluidity and stability of goal selection is similar to Ladouceur et al. insofar as approximately 50 % of their program completing participants switched to an abstinence-based goal at least once during the course of treatment, but also dissimilar insofar as over 50 % of participants in their sample selected a continuous moderation goal. The reason for this latter discrepancy between our results and that of Ladouceur et al. is likely due to the distinct nature of the treatment samples and protocols; that is, our treatment study included a brief motivational protocol that was designed to be largely silent with respect to the encouragement of specific treatment goals, whereas Ladouceur et al.’s cognitive-behavioural protocol was aimed at moderated gambling and recruited participants specifically interested in that goal. Thus, comparing our results to that of Ladouceur et al.’s suggests that the temporal stability of goal selection can be influenced by the agenda of a particular intervention program. It remains unclear to what extent the temporal stability of goal selection is influenced by the dynamic nature of motivation as well as individuals’ changing ideas about what is required to achieve their goals. It is possible that the pattern of goal selection over time observed in the present study might have been different if participants’ were attending weekly therapy sessions in which they were required to discuss and justify their goals.

The results also demonstrated that pretreatment goal was uniquely associated with three factors, whereby compared to participants who selected the goal to ‘cut back on problem gambling’, those who selected the goal to ‘quit problem gambling’ were more likely to have greater gambling problem severity, to have identified video lottery terminal play as problematic, and to have greater motivation to overcome the gambling problem. These results suggest that individuals who seek treatment with the aforementioned characteristics are more likely to select abstinence-based treatment goals at pretreatment, which might reflect good self-awareness and insight into their gambling problem. These findings are consistent with reports in the alcohol treatment literature (Adamson and Sellman 2001; Dunn and Strain 2013; Hodgins et al. 1997; Ogborne 1987), in which the selection of abstinence-based goals has been found to be related to greater alcohol problem severity; and are consistent with Toneatto and Dragonetti’s (2008) gambling treatment study, which demonstrated that individuals who selected abstinence-based goals had greater gambling problem severity and were more motivated to overcome their gambling problem as indicated by greater session attendance. However, it is not entirely clear as to why, in the present study, problematic video lottery terminal play, as opposed to other forms of gambling, emerged as a unique predictor of abstinence-based goals at pretreatment. This type of gambling was the most frequently reported as problematic in this sample and has also been cited as particularly “addictive” (Dickerson and Baron 2000). Nevertheless, the other form of electronic gaming machines locally available was not significantly associated with participant goal selection.

Another novel contribution of the present study was that it examined the influence of nuanced goal selection on gambling outcomes over the course of treatment. Our results demonstrated that while dichotomous goal selection at pretreatment did not predict the number of days gambled at 3-months follow-up, nuanced goal selection at 3-, 6-, and 9-months follow-up predicted the number of days gambled at 6-, 9-, and 12-months follow-up, respectively. Specifically, over the course of the follow-up periods, participants who selected the abstinence-based goals of ‘quit all types of gambling’ and ‘quit problem type of gambling’ gambled did not differ with each other with respect to days gambled, but participants who selected these goals both gambled significantly fewer days than those who selected the moderation-based goal of ‘gamble in a controlled manner’. This finding might be expected given that those with abstinence-based goals are aiming at zero days of gambling compared to those with non-abstinence-based goals. Importantly, however, neither dichotomous goal selection at pretreatment, nor nuanced goal selection at 3-, 6-, or 9-months was related to the other gambling outcomes over the course of study (i.e., average number of dollars gambled, average number of dollars per day gambled, or perceived goal achievement). Thus, the bulk of evidence from our results suggests that the endorsement of moderation-based goals does not appear to confer a disadvantage with respect to gambling outcomes. These findings then, are somewhat inconsistent with literature citing advantages (i.e., reduced risk for relapse) to the endorsement of complete abstinence goals among clients in treatment for alcohol, opiate, nicotine, and cocaine dependence (Hall et al. 1990, 1991). Moreover, our results are inconsistent with Ladouceur et al.’s (2009) finding that participants who switched from endorsing a moderation goal to abstinence were more likely to report that they had often succeeded in respecting their own gambling limits (gambling sessions and money spent) during the week preceding 6-months follow-up; instead, our results indicated that, irrespective of goal choice, participants perceived greater goal achievement over time. While we did not measure participants’ actual moderation-based goals, we were able to infer what outcomes might reflect these goals. Specifically, participants who endorsed moderation-based goals who completely achieved these goals reported overall that their goals, between the assessment periods, were found to gamble approximately between 1 and 3 days, between $86 and $258, and between $33 and $120 per day. Examination of self-reported gambling among community samples has identified that risk of harm from gambling is associated with gambling more than two to three times per month, more than $1,000 per year, and more than 1 % of gross income (Currie et al. 2006, 2008a, b). Weinstock et al. (2007) reported that individuals who were gambling without significant harm after abstinence-focused cognitive behavioural therapy reported gambling monthly or less and 1.9 % of income or less. We did not assess annual income in this sample but the frequency and expenditure figures are roughly equivalent to these risk guidelines. Future research is warranted that precisely measures participants’ actual moderation-based goals, the extent to which they achieve these goals, and the level of associated harm.

Several limitations and future directions in this paper are worthy of mention. Most importantly, this paper was a secondary analysis of a previously published RCT, and as such, it was not designed specifically to investigate participant goal selection. Consequently, our assessment of participant goal selection was not ideal. While a major strength of this paper was its unique documentation of nuanced participant goal selection, it was unable to make full use of this information due to incongruent measurement of goal selection between the pretreatment and follow-up periods. As well, we were unable to link specific types of gambling during the follow-up periods to participants’ specific abstinence- versus moderation-based goals. Future research that more rigorously examines nuanced goal selection variables, particularly in the context of other treatment modalities, would help to further refine our understanding of this process. Additionally, obtaining information directly from participants regarding their reasons for selecting a particular treatment goal (as was done at pretreatment by Dowling and Smith 2007), as well as the circumstances leading to changes in their goal selection, would also help to provide a better understanding of the nature of treatment goal setting and its influence on gambling outcomes.

Finally, it is important to underscore that our results do not address whether the imposition of treatment goals on treatment-seeking pathological gamblers is advantageous. The research protocol used in the present study required participants to select a treatment goal, and the self-help workbook also recommended that participants consider their goal carefully and make detailed frequency and expenditure goals for each specific type of gambling. This intervention was largely self-directed and did not require the discussion of goals with a therapist. In contrast, other protocols and clinical services that are more traditional only provide abstinence-based gambling treatment. There may be advantages to allowing individuals to come to their own determination with respect to treatment goal setting, as has been suggested in the alcohol treatment literature (Sobell and Sobell 2011).

To conclude, the present study has highlighted that pathological gamblers are not homogenous with respect to their therapy objectives, and has contributed to the wide gap in the empirical literature with respect to the lack of information about the nature and impact of goal selection in gambling treatment. It has demonstrated that goal selection is both fluid and stable over time for a substantial number of individuals, is predictable at pretreatment, and is related to certain aspects of treatment outcome (i.e., days gambled), but not others (i.e., dollars gambled, dollars per day gambled, perceived goal achievement). Overall, our findings do not suggest that there is an advantage to the endorsement of abstinence goals as opposed to moderation goals during the course of gambling treatment. Much more rigorous and refined research is needed to further advance the debate regarding abstinence versus moderation in the context of gambling treatment and addictions treatment more broadly.

Implications for Clinical Practice

Our findings, in conjunction with the broader literature, have several implications that can be used by clinicians to inform their clinical practice. Specifically, clinicians might consider incorporating the following psychoeducational information when discussing gambling therapy goals with their patients:

  • Gambling therapy goals have been found to be associated with specific client characteristics prior to the onset of treatment. Specifically, individuals who begin treatment with more severe gambling problems, more motivation to overcome their gambling problems, and who identify video lottery terminal play as particularly problematic, tend to aim towards quitting their problem gambling rather than cutting back.

  • Individuals who begin treatment vary in their initial gambling therapy goals. These goals have been found to shift over the course of treatment for some people, and remain unchanged for others.

  • It has been found that when individuals are free to choose their gambling therapy goals over the course of treatment, improvements in gambling treatment outcomes have been observed.

  • Gambling therapy goals have been found to be largely unrelated to gambling treatment outcomes, and therefore, there does not appear to be a therapeutic advantage for individuals to select an abstinence-based goal versus a moderation-based goal.

  • Individuals who successfully achieve their moderation-based gambling therapy goals tend to report low frequency and expenditure figures.