Introduction

There is increasing interest in the commonalities among addictive behaviours (Mudry et al. 2011; Potenza 2009; Shaffer 2012), including substance use disorders and behavioural addictions such as gambling. Because they share etiologic and phenomenological features, substance use disorders and gambling disorder are now classified together in the Substance-related and Other Addictive Disorders section of the Diagnostic and Statistical Manual (DSM-5) (American Psychiatric Association 2013). Disordered eating, such as binge eating disorder, is often conceptualized from an addiction perspective (Mudry et al. 2011; Cassin and von Ranson 2007; Wilson 2010). We previously explored the links between drinking and gambling among university students (Hodgins and Racicot 2013). Consistent with a problem syndrome model (Jessor and Jessor 1977), we found that drinking and gambling were most strongly linked through general dimensions of problematic involvement, specifically through coping motives. Students who drink and gamble were most likely to describe engaging in both these behaviours as ways of coping with negative affect. In this paper we attempt to replicate these findings and extend this examination to disordered eating in a new sample of university students.

Problematic drinking, gambling and eating behaviors are particularly frequent among university students (Dancyger and Garfinkel 1995; Wechsler et al. 2002; Neighbors et al. 2002; Martens et al. 2005; Ferrier and Martens 2008) and have been the focus of specially designed interventions (e.g. Cronce and Larimer 2011; Taylor et al. 2006; Manwaring et al. 2008)). However, data on whether these behaviors co-occur at rates greater than chance among university and college student populations are inconsistent, with some studies showing moderate links and others not (Dunn et al. 2002; Fischer et al. 2008; Piran and Robinson 2011). Understanding whether and how these behaviors are linked may have implications for effective prevention and treatment.

In our previous study we assessed students on a number of alcohol and gambling-related domains, including level of involvement, consequences, impairment of control, and motives (Hodgins and Racicot 2013). Impairment of control refers to a reduced ability to resist an urge to drink or gamble (Leeman et al. 2013, 2009). Alcohol impaired control typically emerges as an early indicator of student heavy drinking (Chung and Martin 2002, 2002) and is prospectively linked to alcohol dependence (Leeman et al. 2009). Similarly, gambling impaired control is linked to severity of gambling problems (Dickerson and O’Connor 2006), although prospective data are lacking. Drinking motives have been conceptualized in a four factor model as social, enhancement, coping and conformity motives (Cooper et al. 1995), and this model has been validated in describing gambling motives (Stewart and Zack 2008). Specific motives are differentially related to different aspects of drinking and gambling. For example, coping motives have generally been linked with heavier and more problematic drinking (Lyvers et al. 2010).

In the eating disorder literature, similar constructs exist, although specific labels and conceptualizations differ (Mudry et al. 2014). For example, the term “dietary restraint” is frequently used, with low restraint implying lack of control in some definitions (Stice et al. 2004). The reason that this concept differs from alcohol and gambling is that the desire to eat is biologically adaptive and recovery from anorexia nervosa and often bulimia nervosa, albeit not binge eating disorder, requires a reduction in dietary restraint (Wardle 1987). Whereas in the gambling and drinking domains, negative consequences of those behaviours are often measured, they are not typically assessed for disordered eating. In terms of eating motives, Cooper’s (1994) alcohol motives model has been applied to students with eating disorder symptoms (Jackson et al. 2003). As with alcohol and gambling, specific motives were found to be associated with different patterns of eating behaviour, such that coping motives were positively associated with restrictive eating, binge eating, and purging; and social motives were negatively associated with restrictive eating, but were positively associated with binge eating and purging. The relationship between eating disorders and drinking motives has been investigated, whereby coping drinking motives in female students were linked with probable bulimia nervosa or binge eating disorder (Luce et al. 2007) and eating pathology more generally (Anderson et al. 2006). These studies did not assess eating motives.

The goal of the present study was to assess links among drinking, gambling and eating among university students, using a variety of commonly assessed constructs. Based upon our earlier study, we hypothesized that drinking and gambling would be linked by overall problematic involvement but more strongly linked as a way of coping with negative affect. We expected similar relationships among eating pathology and drinking and gambling.

Methods

Participants

The sample (N = 301) was comprised of 212 women (70.4 %) and 89 men (29.6 %) with a mean age of 20.7 years (17–49, SD = 3.5). Of the sample, 78 % were born in Canada, with 39 % identifying as Caucasian, 24 % as Asian, 8 % as European, and the remaining 29 % as other. Almost all the participants (94 %) were single and 64 % currently lived at a family residence, 20 % lived in an independent residence and 13 % lived in a university residence. In terms of major, 38 % were enrolled in Psychology, 20 % in Sciences, 12 % in Business and 30 % in other fields.

Procedure

Ethics approval was obtained through the University of Calgary and students received course credit for participation. Students completed questionnaires in groups of 3 to 12, taking an average of 30 min.

Measures

Students completed a demographic questionnaire (Adlaf et al. 2005) to gather information on gender, age, ethnicity, income, place of residence, year of study, program of study, grade point average, and engagement in various campus activities.

Gambling Measures

Students rated their last six months involvement in gambling activities including lottery or raffle tickets, slot machines, betting on sports, cards, and dice, internet betting, casino gambling, etc. (Adlaf et al. 2005). An index of gambling involvement was provided from the summed total (α = 0.79). Additional questions (Adlaf and Ialomiteranu 2000; Neighbors et al. 2002) were used to measure overall gambling frequency and money spent/lost and won for descriptive purposes.

The Gambling Motives Questionnaire (GMQ) was used to measure gambling motives (Stewart and Zack 2008). The GMQ’s three scales, social, coping, and enhancement motives, have demonstrated high internal consistency and adequate test-retest reliability (Stewart and Zack 2008). In Hodgins and Racicot (2013), additional items designed to measure monetary and charitable gambling motives were developed. Cronbach’s alpha in the present sample were 0.86 for social; 0.78 for coping; 0.90 for enhancement; 0.95 for charity; and 0.85 for monetary.

The short version of the Scale of Gambling Choices (SGC) measured level of impaired control (O’Connor and Dickerson 2003; Dickerson and O’Connor 2006). The SGC has high internal reliability (α = 0.88 this sample), adequate test-retest reliability and has demonstrated modest to strong correlations with measures of gambling (O’Connor and Dickerson 2003). Gambling-related consequences were measured using the Gambling Problems Index (GPI), 20 items measuring consequences such as missing class, neglecting responsibilities, and arguing with friends or family (Neighbors et al. 2002). Six items measuring impairment of control were removed to provide a more conceptually pure measure of gambling-related consequences (α = 0.89 for the complete and α =0.87 for the revised scale).

Drinking Measures

Students completed the Timeline Followback (TLFB) to assess drinking over the past 30 days (Sobell et al. 1985). The self-report version has been shown to have good test-retest reliability with college students (Sobell et al. 1986). Mean number of drinks per drinking day and number of drinking days per month were multiplied to provide an Alcohol Involvement Index. The Drinking Motives Questionnaire (DMQ) measured coping, enhancement, social and conformity (Cooper et al. 1992; Cooper 1994). In the current sample α = 0.93 for social; α = 0.87 for coping; α = 0.90 for enhancement and α = 0.85 for conformity. The Impaired Control Scale (ICS) measured impaired control over drinking with three scales, attempted control (AC), failed control (FC) and predicted control (PC) over the past six months (Heather et al. 1998). The recommended substitution method of scoring the FC scale was used. The ICS has demonstrated high test-retest reliability and internal consistency (Marsh et al. 2002; Adamson et al. 2010; Nagoshi 1999; Heather et al. 1993). In the current sample ICS-AC, α = 0.90; ICS-FC, α = 0.71 and ICS-PC, α = 0.86. The Young Adult Alcohol Problems Screening Test (YAAPST) measured adverse drinking consequences (Devos-Comby and Lange 2008) (α = 0.85).

Disordered Eating Behavior Measures

The Eating Disorder Examination-Questionnaire (EDE-Q, version 6.0) assessed disordered eating attitudes and behaviours over the past 28 days (Fairburn and Beglin 1994). The EDE-Q provides a total and four subscale scores: Restraint, Eating Concern, Weight Concern, and Shape Concern (Restraint α = 0.90; Eating Concern: α = 0.74; Shape Concern: α = 92; Weight Concern: α = 0.83; Total: α = 0.89).

Eating Expectancies were measured with four of the Eating Expectancy Inventory (EEI) scales (Hohlstein et al. 1998), Eating Alleviates Negative Affect (α = 0.95), Eating is Pleasurable and Useful as a Reward (α = 0.75), Eating Alleviates Boredom (α = 0.83), and Eating Enhances Cognitive Competence (α = 0.82). These scales have been shown to have good internal reliability and validity in a number of adolescent samples (Simmons et al. 2002). An eight-item version of the Uncontrolled Eating Scale from the Three Factor Eating Questionnaire-R18 (TFEQ) measured the tendency to overeat with the feeling of being out of control (Angle et al. 2009). Cronbach’s alpha in this sample was 0.85.

Data Analysis

Following Hodgins and Racicot (2013), canonical correlation analysis was conducted to examine the relationship among the set of alcohol, set of gambling, and set of disordered eating variables. Canonical correlation assesses the relationship between two sets of variables in terms of dimensions (referred to as canonical variates) that are common between the sets. It is a useful method of exploring relationships among theoretically related domains. Rather than looking at a large number of pairs of variables separately, linear combinations of the variables within the sets are examined for correlation. Because canonical correlation is designed to compare two sets of variables, three separate analyses were conducted, comparing alcohol and gambling, alcohol and eating, and eating and gambling.

Finally, principal component analysis of the drinking, gambling and eating variables was conducted to examine further their patterns of associations and to determine whether these patterns varied according to gender. Number of factors was determined through examination of the eigenvalue distribution and parallel analysis (Horn 1965; O’Connor 2000) and gender differences in the factor structure were assessed using Tucker’s Test of Congruence of the Varimax rotated factor loadings (Lorenzo-Seva and Ten Berge 2006).

Results

Sample Characteristics

Table 1 displays the alcohol, gambling, and disordered eating variables separated by gender. Men gambled more frequently than women and consumed a greater mean number of drinks per month, but did not drink more frequently. Gambling activities in the past six months included lottery tickets (34.6 %), playing dice, cards or other games for money (27.6 %), casino table games (22.6 %), and electronic gaming machines (19.9 %). The majority of the total sample (81.4 %) indicated that they had lost less than $25 in the last year and the second largest group (10.6 %) reported losing $25 to $100, with 7.7 % reporting losing more than $100. Mean EDE-Q global and subscale scores for women in this sample fell within the 55th to 60th percentile of norms for US female undergraduate students (Luce et al. 2008) Raw scores for men were lower but fell at the 55th to 70th percentile compared with US norms (Lavender et al. 2010). Overall, two out of five participants reported overeating or excessive exercising in the last month, with a small proportion reporting self-induced vomiting and laxative use. There were no significant gender differences in frequency of disordered eating behaviours.

Table 1 Summary of Responses to alcohol, gambling, and disordered eating variables by gender (N = 301)
Full size table

Canonical Correlation

Seven outliers for alcohol involvement (2 SDs above the mean) were recoded to one greater than the next highest value. Log transformations improved the distribution of the following variables: alcohol involvement, alcohol impaired control (failed and predicted), gambling involvement, gambling impaired control, and gambling coping and winning motives. Table 2 presents the within set correlations for each of the drinking, gambling and eating variables and Table 3 presents the between set correlations. Bivariate correlations between the alcohol and gambling variables were generally stronger than correlations between the eating and gambling or eating and alcohol variables.

Table 2 Within set alcohol, gambling and eating pearson correlations
Full size table
Table 3 Between set alcohol, gambling and eating pearson correlations
Full size table

The canonical analysis for the alcohol and gambling variables yielded three significant canonical variates, R = 0.43, χ2 (72) = 167.7, p < 0.001, R = 0.35, χ2 (56) = 110.4, p < 0.001, R = 0.32, χ2 (42) = 73.5, p = 0.002. Table 4 displays results for the first pair of alcohol and gambling canonical variates, including standardized coefficients, loadings and cross loadings, within set variance accounted for and redundancies. Loadings, which are commonly emphasized in the interpretation of canonical analysis results (Tabachnick and Fidell 2007), represent the correlation of individual variables with the canonical variate. For the first variate of drinking, all variables loaded significantly (> 0.30) except for the failure of impaired control, and particularly high loadings were evident for drinking involvement, consequences, drinking to cope and for social reasons. The canonical variate accounted for 28 % of the variance of the drinking variables. The first variate of gambling showed moderate correlations with involvement, impairment of control, and social, enhancement and winning motives, with an especially strong correlation for gambling for social motives. The canonical variate accounted for 29 % of the variance of the gambling variables.

Table 4 Canonical coefficients and loadings of first canonical variate: alcohol and gambling
Full size table

Cross loadings show the correlation between specific variables and the opposite canonical variate (see Table 4). Drinking for social motives was related to the gambling variate and gambling for social reasons was related to the alcohol variate. The redundancy coefficient indicated that only 5 % of the variance in the first alcohol covariate was accounted for by the gambling variate and 5 % of the variance of the first gambling variate was accounted for by the alcohol variate.

Although the second and third canonical correlations were significant, the proportion of variance accounted for by the set of alcohol and gambling variables was small (0.07 and 0.10 for alcohol; 0.09 and 0.05 for gambling respectively), and therefore, were not interpreted.

The canonical analysis between alcohol and eating variables yielded one significant canonical variate, R = 0.40, χ2 (81) = 128.6, p < 0.001 (see Table 5). For the drinking variate, involvement, failure of impaired control, enhancement and social motives, and consequences correlated greater than 0.30. The canonical variate accounted for 20 % of the variance of the drinking variables. For the set of eating variables, eating to decrease negative affect and for the purpose of pleasure and cognitive control were moderately negatively correlated with the variate, which accounted for 15 % of the variance of the eating variables overall. Cross loadings suggested alcohol involvement was related to the disordered eating covariate and that eating to decrease negative affect was negatively related to the alcohol covariate. The redundancy coefficient indicated that only 3 % of the variance in the first alcohol covariate was accounted for by the disordered eating covariate and 2 % of the variance in the first disordered eating covariate was accounted for by the alcohol covariate.

Table 5 Canonical coefficients and loadings of first canonical variate: alcohol and eating
Full size table

Because the gambling variables were largely unrelated to the disordered eating variables, the canonical analysis did not yield a significant canonical variate, R = 0.36, χ2 (72) = 91.3, p = 0.064.

Principal Component Analysis

A principal component analysis (PCA) was conducted on the 25 alcohol, gambling and eating variables. The Kaiser-Meyer-Olkin measure of sampling adequacy was 0.83, which suggests good factorability. Not all variables loaded at 0.3 or greater on the first unrotated component, which suggests that more than one factor exists. Five eigenvalues were greater than one but distribution showed a sharp discontinuity after four factors. Parallel analysis also indicated four factors (O’Connor 2000). The four factors accounted for 62 % of the variance and varimax rotation revealed a clear factor loading pattern (see Table 6). The alcohol variables all loaded >0.6 on the first rotated factor and the gambling variables, with the exception of gambling for charity motives, loaded >0.67 on the second rotated factor. The disordered eating variables were split between third and fourth factors, with each variable loading strongly on only one factor. Examining the content of these scales, it appears that factor 3 tapped into symptoms of eating pathology, whereas factor 4 reflected more normative attitudes, expectancies and behaviours (e.g., overeating).

Table 6 Principal components analyses varimax rotated factor loadings for alcohol, gambling and eating variables
Full size table

The PCA was repeated for males and females separately and similar results were uncovered in terms of number of factors and rotated factor loadings. Tucker’s Test of Congruency indicated that the alcohol (CC = 0.98) and gambling factors (CC = 0.97) could be considered equal in men and women (on the basis of rotated factor loadings) and that the two eating factors were “fairly similar” (CC = 94. and 0.93).

Discussion

There is an increasing interest in understanding the commonalties and differences among addictive behaviours among university students and in the general population. Although there is some amount of maturing out of involvement in these behaviours as younger people move into different adult roles, involvement as a student does predict chronic difficulties for drinking and gambling. Our results replicate previous findings showing a significant but small link between alcohol and gambling problematic involvement (Hodgins and Racicot 2013; Fischer et al. 2008). At the bivariate level, alcohol and gambling frequency were correlated r = 0.21 as were negative consequences, r = 0.22 (Table 3). The multivariate analyses showed that a general dimension of problematic involvement had the strongest link (R = 0.43). Students who drank more heavily, had more negative consequences, and higher social and coping motives for drinking in particular also had greater gambling involvement, impaired control, negative consequences, and stronger motives for gambling. These results are very similar to our previous sample and are consistent with a general problem syndrome model (Jessor and Jessor 1977; Barnes et al. 2010).

In our previous report we found that alcohol and gambling were strongly linked in their use as ways of coping with negative affect. That finding was not replicated in the present sample, although motives for gambling involvement and alcohol use were particularly strong aspects of the alcohol-gambling link. In this case, drinking and gambling for social motives seemed to have a relatively strong connection. Given that both alcohol and gambling among young people are very socially focused activities, this connection is not surprising. However, it does suggest that environmental factors may be a larger role than individual factors in commonalities among these behaviours. Drinking for social reasons also tends to be less predictive of future drinking problems than coping and enhancement motives (Merrill et al. 2014).

There was also a small significant relationship between the alcohol and eating variables (R = 0.40) reflecting a negative relationship between problematic alcohol involvement and eating attitudes and behaviors. Results suggested that problematic alcohol involvement was inversely related to eating to decrease negative affect, and for the purpose of pleasure and cognitive control – heavier drinkers tended not to eat for these reasons. The latter results are inconsistent with limited previous research with college women that found a positive link between eating pathology and drinking to cope motives (Anderson et al. 2006; Luce et al. 2007). They instead suggest that eating and alcohol serve different purposes among students. Because affect regulation is a central concept in treatment of these disorders, future research clarifying these relationships is crucial. It is also important to note that within the drinking literature, a conceptual difference is recognized between drinking expectancies and drinking motives wherein expectancies are more distal predictors of drinking and motives are more proximal (Cooper et al. 1995). As far as we are aware, a distinction between eating expectancies and motives has not been as carefully drawn in the eating pathology literature. In this study we used an expectancy measure, and not a motive measure, which may have implications for our results.

In contrast to significant links between alcohol and both gambling and problematic eating, gambling and problematic eating were unrelated. This finding suggests that symptoms of and motives for problematic eating and gambling tend to be distinct addictive behaviours. These behaviours tend to be undertaken by different people, with individuals tending to endorse either gambling or disordered eating but not both. A recent study of a Spanish sample of individuals in treatment for eating disorders also did not find elevated rates of gambling disorder compared to the general population (Jimenez-Murcia et al. 2013). A study of university students (Fischer et al. 2008), however, found that although eating and gambling pathologies were not correlated, both were related to the tendency to react in a rash manner when upset (negative urgency). This finding is consistent with the model that underlying processes can have different behavioral expressions in different individuals (Shaffer 2012).

As expected, generally men reported more alcohol and gambling involvement and women more eating concerns. It is interesting that although men drank more than women and reported more negative consequences, they did not drink more frequently. Men and women also reported similar motives for drinking with the exception of drinking to conform to group norms, which was more frequently reported by men. Lack of gender differences in drinking motives and expectancies have been found in other student samples (e.g., Fischer et al., 2008; Merrill et al. 2014) and may reflect how socially embedded drinking is for undergraduate students generally. With gambling, however, men who gamble reported stronger coping, enhancement, social motives for gambling and more gambling to win money. Men and women were equally likely to report gambling to support charities. Men and women tended to endorse similar motives for eating, although women were more likely than men to endorse the use of eating to manage negative affect. Unexpectedly, men and women did not report statistically different frequencies of eating disorder behaviours such as overeating, self-induced vomiting, use of laxatives to manage weight or excessive exercise. Women, however, did report far greater dietary restraint and concerns about eating, weight, and shape than men.

Although men and women differed in their level of involvement in each of the three behaviours, the PCA suggested that the overall association of the different features such as suffering negative consequences, impairment of control, and motives for involvement were similar across gender, in particular for alcohol and gambling. The finding that the eating-related measures formed two factors instead of one, unlike drinking and gambling, highlights the complexity of eating behaviours and motives. Attitudes symptomatic of eating disorders tended to cluster together, whereas eating expectancies combined with overeating to form a second factor. It is worth noting that the overeating scale was derived for use with obese individuals whereas the eating disorder attitudes scales were developed for use with individuals with eating disorders, and, as such, the conceptualizations of eating behaviors differed somewhat (Karlsson et al. 2000; Fairburn and Beglin 1994). The former scale emphasizes extreme hunger and overeating, whereas the latter measures a wider range of attitudes and behaviours symptomatic of eating disorders, including food restriction, over concern with weight and shape, fear of weight gain, and body dissatisfaction. We speculate that the factor that included eating motives with overeating reflects behaviours and attitudes that are more normally distributed in a nonclinical population than eating disorder symptoms, which are more aberrant and tend to affect fewer individuals. Further research is necessary to understand the relationship of overeating to diagnostic symptoms of eating disorders.

Although this study is useful as expanding our understand of the specific nature of links between additive behaviors in undergraduate students, important limitations exist including the use of a cross-sectional, convenience sample and self-report methodology. The sample size was adequate for the analyses although the use of volunteer recruitment provides a sample that is not necessarily representative of university students in general. Future research should attempt to achieve representative samples. Studies that assess motivations through the use of implicit methods would also overcome the potential of shared method variance with assessing motives only with self-report scales.

Conclusions

Problematic alcohol and gambling are significantly but not strongly linked among university students as are problematic alcohol and disordered eating. Problematic alcohol involvement was linked to eating attitudes and behaviors reflecting mostly specific eating expectancies, suggesting that alcohol and eating serve different purposes among undergraduates and that there is not a straightforward relationship between eating attitudes and behaviors and alcohol involvement and motives. Results did not find a link between the eating and gambling variables, which suggests these behaviours are distinct.

Future research needs to clarify how different types of motives for different behaviours are linked. If specific motives (e.g., coping) represent a common link for students who engage in multiple addictive behaviours, it is possible that altering these motives may lead to simultaneous improvements across behaviours. Targeting motives in prevention and treatment interventions may be an efficacious approach given that motives are considered very proximal predictors of behaviour as they are likely directly linked to the decision to engage in a specific behaviour.

Defining concepts that are parallel across these three types of addictive problems is difficult, perhaps because these behaviours play different roles in students' lives. Eating is not optional, whereas other addictive behaviours are optional, or at least start that way. As a result of constant interactions with and choices about food, we speculate that relationships with eating are more complex and imbued with subtle meanings and motives than with non-essential behaviours such as drinking and gambling.