A Longitudinal Study of the Temporal Relation Between Problem Gambling and Mental and Substance Use Disorders Among Young Adults

Abstract

Objective:

Relatively little is known about the temporal relation between at-risk gambling or problem gambling (PG) and mental and substance use disorders (SUDs) in young adulthood. Our study aimed to examine whether past-year, at-risk, or PG is associated with incident mental disorders and SUDs (that is, depression, generalized anxiety disorder, obsessive–compulsive disorder [OCD], or alcohol dependence) and illegal drug use, and whether past-year mental disorders and SUDs and illegal drug use is associated with incident at-risk or PG.

Method:

Data for this longitudinal study were drawn from the Manitoba Longitudinal Study of Young Adults (MLSYA). Respondents aged 18 to 20 years in 2007 were followed prospectively for 5 years.

Results:

In cross-sectional analyses, at-risk or PG was associated with increased odds of depression, OCD, alcohol dependence, and illegal drug use. In longitudinal analysis at-risk or PG at cycle 1 was associated with incident major depressive disorder, alcohol dependence, and illegal drug use in the follow-up period. Only illegal drug use at cycle 1 was associated with incident at-risk or PG during follow-up.

Conclusions:

At-risk or PG was associated with more new onset mental disorders and SUDs (depression, alcohol dependence, and illegal drug use), compared with the reverse (illegal drug use was the only association with new onset at-risk or PG). Preventing at-risk or PG from developing early in adulthood may correspond with decreases in new onset mental disorders and SUDs later in adulthood.

Many young adults participate in gambling activities and are at-risk for having gambling problems, which are known to be associated with mental disorders.^1–9 Problem gambling (PG) refers to gambling behaviour that has a negative impact on the gambler, friends and family, and the community.¹⁰ At-risk or PG is based on having 1 or more signs of a gambling disorder, whether or not the criteria for a diagnosis is met. In 2002, 76% of Canadians aged 15 years and older endorsed gambling at least 1 time in the previous year,¹¹ with 4.9% of men and 2.7% of women in the adult population meeting criteria for moderate or severe gambling problems.¹² Similarly, past-year estimates of the most severe gambling problems (that is, pathological gambling or those meeting criteria for a diagnosis) in the general United States population are estimated to be at about 1% to 3% for adults,^13,14 with higher estimates for young adults aged 18 to 24 years (7% to 14%) as compared with older adults (2.1% to 5.6%).^15–18

Research on adolescents and young adults suggests that gambling behaviours increase through the teenage years, reaching a peak in the early twenties, before decreasing from the late twenties onward.¹⁹ More specifically, in the United States, Welte et al¹³ found that about 60% of people aged 14 to 15 years reported gambling in the past-year, whereas 72% of 20 to 21 year olds gambled in the past year. A recent study¹³ found that gambling in the past year peaked for people aged 22 to 30 years, while PG peaked later at age 31 to 40 years. Longitudinal prospective data from the United States has also indicated that gambling and PG remained stable and at-risk gambling increased over time in an adolescent sample.²⁰ In longitudinal data from South Australia, gambling participation substantially increased from adolescence (16 to 18 years) into young adulthood (20 to 21 years).²¹ Another longitudinal study of the transitions in gambling participation during late adolescence and young adulthood among inner-city youth in the United States indicated that gambling follows a variable developmental course from adolescence to young adulthood: over a 6-year period, participants who reported past-year gambling had a 36% to 51% chance of reporting gambling at a subsequent assessment.²² Due to inconsistencies in how gambling is measured across studies, it is difficult to determine if the frequency of gambling and at-risk or PG are more or less prevalent in younger age groups (adolescents and young adults), compared with older age groups.²³

A relation between PG and mental disorders has been well established in the literature.^1–9,24 More specifically, cross-sectional data of adolescent and adult samples has found that PG is related to increased odds of mood disorders, anxiety disorders, substance use disorders (SUDs), and personality disorders.^{1,4,9,25–33} Some researchers have extended this work to look at the temporal relation between at-risk or PG and mental disorders. Using cross-sectional data with age of onset assessments, Kessler et al⁶ determined that pathological gambling was linked with subsequent onset of generalized anxiety disorder (GAD), posttraumatic stress disorder (PTSD), and substance dependence. Other studies using longitudinal US general population adult data has found that pathological gambling at baseline was associated with increased odds of incident mental disorders.^2,24 Specific gambling-related symptoms have also been found to be associated with comorbid disorders.²⁴ Similarly, 2 other studies using the same US data indicated that sex moderated the relations between at-risk or PG and incident SUDs with differences noted among men and women³⁴ and that at-risk or PG was associated with increased incidence of GAD and SUD among older adults aged 55 to 90 years.⁸

Longitudinal studies with adolescents and young adult samples have examined gambling, depressive symptoms, and impulsivity over time and found that gambling does increase the likelihood of depressive symptoms^35,36 with some findings suggesting that impulsivity may potentially moderate the relation.³⁵ Another longitudinal study from childhood into adolescence found that gambling was associated with teacher rating of externalizing behaviour and with parent ratings of impulsivity and hyperactivity.³⁷ In addition, a study³⁸ using the same data has shown that childhood and early adolescent aggressive or disruptive behaviours are associated with gambling and at-risk or PG as measured by the South Oaks Gambling Screen—Revised for Adolescents in late adolescence. However, studies using adolescent and young adult samples are limited due to assessments of disorder symptoms and traits, rather than assessments of mental disorders.

With a shortage of longitudinal gambling and mental disorder data collected during young adulthood, it is difficult to examine the temporal relations between at-risk or PG and mental disorder and SUD use exclusively among young adults. At-risk or PG may be associated with new onset mental disorders and SUDs, while for others the mental disorders and SUDs may develop first and lead to at-risk or PG. Young adulthood may present a critical period of intervention before gambling problems have been established over a longer period of time into middle and later adulthood.

The main objectives of our study are to examine: the sociodemographic variables by gambling type; the prevalence of the temporal order of onset between at-risk or PG and mental disorders and SUDs; the cross-sectional associations between at-risk or PG and mental disorders and illegal drug use; whether past year at-risk or PG is associated with incident mental disorder or SUD; and whether past-year mental disorders and SUDs is associated with incident at-risk or PG in a longitudinal sample of young adults.

Methods

Respondents

Data for our study were drawn from the Manitoba Longitudinal Study of Young Adults (MLSYA). The primary objective of the MLSYA was to understand gambling behaviours and attitudes overtime among a young adult sample. Respondents for the study were recruited using random sampling, snowball recruitment, and convenience sampling procedures. These procedures included random-digit dialing, participant providing up to 5 new participant referrals at the end of the interview, Internet advertising of the study, onsite casino recruiting, posters in post-secondary institutions, advertisements in post-secondary institution newspapers, and posters at video lottery terminal sites. The respondents were told that responses to the survey will provide information on various leisure activities, including gambling, that they may or may not participate in and other information that might be related to leisure activities. The sample was similar to sociodemographic characteristics of Manitobans aged 18 to 20 years, with the exception of slightly higher education levels and mostly from urban areas. The mean age of respondents at cycle 1 was 18.9 years and 51.8% were female.

Procedure

Young adults (aged 18 to 20 years at cycle 1) were surveyed at 4 time points across a 5-year span from 2007 to 2011, with an initial sample of 679 young adults. Among these respondents, 517 (76.1%) completed the survey at all 4 time points. Analyses determined that respondents who were missing at cycles 2 through 4 were not different, compared with respondents without missing data, regarding sociodemographic variables. A 2-part survey was administered to respondents during cycle 1, with the first part consisting of a telephone interview and the second part involving the respondents’ choice of an online or mail-in questionnaire. Telephone interviews were used in cycles 2 through 4 (follow-up period). The Manitoba Gaming Control Commission recruited expert academics to ensure that all research and ethical protocol were achieved.³⁹

Measures

At-Risk or Problem Gambling

The valid and reliable Canadian Problem Gambling Index (CPGI) was used to assess past 12-month prevalence of PG. The CPGI is a well-developed tool that was created specifically for assessing PG in general population samples in Canada and has been through extensive psychometric testing during its developmental phases.^10,40 The CPGI uses the following 9 items to assess level of gambling problems: 1) wagered larger amounts to get the same feeling of excitement, 2) tried to win back losses, 3) borrowed money or sold something to get money for gambling, 4) felt you might have a problem with gambling, 5) gambling caused health problems including stress and anxiety, 6) been criticized for your betting or told that you have a problem, 7) gambling has caused financial problems, 8) felt guilty about gambling, and 9) bet more than you could afford to lose. The respondent indicates how frequent each of the above behaviours or problems occurred during the past 12 months: never, sometimes, most of the time, or almost always.

Based on extensive psychometric testing of the CPGI, break points have been identified and used to divide people into 4 gambling categories: nonproblem gambler (score of zero), low risk gambler (score of 1 to 4), moderate risk gambler (score of 5 to 7), and severe risk gambler (score of 8 or more).⁴¹ Psychometric testing of the CPGI have shown the measure to have good sensitivity (78% based on clinical interviews and 83% based on the Diagnostic and Statistical Manual of Mental Disorders (DSM), Fourth Edition, criteria) and specificity (100% based on clinical interviews and DSM-IV criteria).⁴⁰ While the developers of the CPGI indicate different cut-off points for the low- and moderate-risk groupings as reflecting significant risk, more recent psychometric testing indicates that due to lack of meaningful differences, low- and moderate-risk groups could be merged into a single group.⁴¹ In this study, severe gamblers could not be individually examined because of low power (n = 10). Therefore, 2 categories were created that examined any without symptoms of PG (that is, nongamblers and nonproblem gamblers) and other gamblers who indicated any problem symptoms (1 or more). In considering the questions surrounding the establishment of CPGI gambling groupings, it is important to note that our study is interested in studying gambling from a public health and prevention perspective. Therefore, we are interested in understanding at-risk and PG rather than only the most severe gamblers who would meet criteria for gambling disorder. Therefore, our study uses the larger grouping criteria scores of 1 or more, which categorizes low-risk, moderate-risk, and problem gambling groups into an at-risk or PG group.

Axis I Mental Disorders

Axis I mental disorders were assessed using the Composite International Diagnostic Interview—Short Form (CIDI-SF) based on the DSM-IV criteria.⁴² Specifically, in the MLSYA survey the CIDI-SF was used to assess major depressive disorder GAD, and obsessive–compulsive disorder (OCD).

Substance Use and Dependence

Alcohol use was measured using the Alcohol Dependence Scale, which is based on the CIDI-SF and DSM-III-R classification.⁴² This measure was used in the Canadian Community Health Survey and the National Comorbidity Survey and was developed to operationalize both criterion A and criterion B of the DSM-III-R diagnosis for psychoactive SUD. Scores of 3 or more were noted as qualifying for a diagnosis of alcohol dependence.⁴³ Substance use (including marijuana, cocaine, amphetamines, ecstasy, hallucinogens, solvents, heroin, and steroids) was assessed with the Drug Dependence Scale. Drug use was measured by examining whether or not respondents had used illegal drugs, over the past 12 months prior to survey.

Any Mental Disorder or Substance Use

The presence of 1 or more mental disorder or SUD (that is, major depressive disorder, GAD, OCD, alcohol dependence, or any drug use in the past 12 months) was collapsed into a single variable entitled any mental disorder or SUD.

Sociodemographic Variables

The sociodemographic variables assessed at cycle 1 included as covariates in the models were sex, marital status, main activity in the past 12 months, past-year household income, ethnicity, and religion.

Statistical Analyses

First, to provide general descriptive information regarding the sample at cycle 1 (baseline), the prevalence of the 5 gambling categories were computed along with the distribution of each CPGI items. Second, descriptive statistics at cycle 1 were run to understand the distribution of sociodemographic variables by the collapsed gambling groups examining nongambler and nonproblem gambler, compared with at-risk or PGs. Third, the prevalence of at-risk or PG and mental disorders and SUDs onset was computed to determine the prevalence of having either at-risk or PG or mental disorders and SUDs as the antecedent condition. Fourth, cross-sectional logistic regression analyses were conducted to determine the strength of the relation between at-risk or PG and mental disorder at cycle 1. Models were first adjusted for sociodemographic variables (adjusted odds ratios [AOR]-1) and then further adjusted for the presence of any other assessed mental disorder (AOR-2). Longitudinal logistic regression analyses were conducted to determine if the presence of at-risk or gambling problems at cycle 1 were associated with incidence of new onset mental disorders or SUDs at any point during cycles 2, 3, or 4. Models were first adjusted for sociodemographic variables (AOR-1) and then further adjusted for the presence of any other assessed mental disorder (AOR-2). To examine incident mental disorders, respondents with an axis I mental disorder and illegal drug use present at cycle 1 were excluded in the longitudinal analysis, as we were interested in new onset mental disorders in the follow-up period. Finally, longitudinal logistic regression analyses were conducted to assess whether mental disorders or SUDs at cycle 1 were associated with the incidence of new onset at-risk or PG in the follow-up period. Models were first adjusted for sociodemographic variables. Similarly, those with at-risk or PG at cycle 1 were excluded in the analysis of incident gambling. In all longitudinal logistic regressions, the follow-up period included collapsing cycles 2 through 4 into one second time point due to lack of statistical power to assess all cycles individually. This means that our analysis includes 2 time points 5 years apart.

Results

At cycle 1, 11.5% of the sample were nongamblers, 57.0% were nonproblem gamblers, 26.8% were low-risk gamblers, 3.2% were moderate-risk gamblers, and 1.5% were problem gamblers. When collapsing gambling groups, 68.5% of the respondents were nongamblers or nonproblem gamblers and 31.5% were at-risk or problem gamblers (that is, low-risk, moderate-risk, and problem gamblers). Table 1 presents the distribution of each CPGI item by the categorical gambling type. The sociodemographic variables of these groups are presented in Table 2. Differences between the nongamblers or nonproblem gamblers and the at-risk or PG groups were observed in sex and main activity in the past year. Males were more likely to be an at-risk or problem gambler (62%, compared with 38% for females). At-risk or problem gamblers compared with nongamblers or nonproblem gamblers were less likely to report main activity as school (60%, compared with 74%, respectively) and more likely to report work (35%, compared with 23%, respectively).

Table 1.

The distribution of past year problem gambling types at cycle 1.

		Non-gambler,a n = 78, 11.5%		Nonproblem gambler, n = 387, 57.0%		Low-risk gambler, n = 182, 26.8%		Moderate risk gambler, n = 22, 3.2%		Problem gambler, n = 10, 1.5%
Canadian Problem Gambling Inventory items		n	%	n	%	n	%	n	%	n	%
1) Wagered larger amounts to get the same feeling of excitement	Never	0	0.0	386b	99.7	147	80.8	8	36.4	0	0.0
	Sometimes	0	0.0	0	0.0	32	17.6	11	50.0	4	40.0
	Most of the time	0	0.0	0	0.0	2	1.1	2	9.1	5	50.0
	Almost always	0	0.0	0	0.0	1	0.5	1	4.5	1	10.0
2) Tried to win back losses	Never	0	0.0	387	100	126	69.2	6	27.3	1	10.0
	Sometimes	0	0.0	0	0.0	56	30.8	12	54.5	3	30.0
	Most of the time	0	0.0	0	0.0	0	0.0	3	13.5	5	50.0
	Almost always	0	0.0	0	0.0	0	0.0.	1	4.5	1	10.0
3) Borrowed money or sold something to get money for gambling	Never	0	0.0	387	100.0	167	91.8	16	72.7	4	40.0
	Sometimes	0	0.0	0	0	14	7.7	6	27.3	4	40.0
	Most of the time	0	0.0	0	0.0	0	0.0	0	0.0	1	10.0
	Almost always	0	0.0	0	0.0	1	0.5	0	0.0	1	10.0
4) Felt you might have a problem with gambling	Never	0	0.0	387	100	166	91.2	13	59.1	1	10.0
	Sometimes	0	0.0	0	0.0	16	8.8	8	36.4	6	60.0
	Most of the time	0	0.0	0	0.0	0	0.0	1	4.5	1	10.0
	Almost always	0	0.0	0	0.0	0	0.0	0	0.0	2	20.0
5) Gambling caused health problems including stress and anxiety	Never	0	0.0	387	100	171	94.0	17	77.3	7	70.0
	Sometimes	0	0.0	0	0.0	11	6.0	5	22.7	2	20.0
	Most of the time	0	0.0	0	0.0	0	0.0	0	0.0	1	10.0
	Almost always	0	0.0	0	0.0	0	0.0	0	0.0	0	0.0
6) Been criticized for your betting or told that you have a problem	Never	0	0.0	387	100.0	150	82.4	12	54.5	2	20.0
	Sometimes	0	0.0	0	0.0	32	17.6	9	40.9	5	50.0
	Most of the time	0	0.0	0	0.0	0	0.0	0	0.0	0	0.0
	Almost always	0	0.0	0	0.0	0	0.0	1	4.5	3	30.0
7) Gambling has caused financial problems	Never	0	0.0	387	100.0	175	96.2	15	68.2	3	30.0
	Sometimes	0	0.0	0	0.0	7	3.8	7	31.8	6	60.0
	Most of the time	0	0.0	0	0.0	0	0.0	0	0.0	0	0.0
	Almost always	0	0.0	0	0.0	0	0.0	0	0.0	1	10.0
8) Felt guilty about gambling	Never	0	0.0	386b	99.7	91	50.0	5	22.7	2	20.0
	Sometimes	0	0.0	0	0.0	86	47.3	9	40.9	4	40.0
	Most of the time	0	0.0	0	0.0	2	1.1	5	22.7	2	20.0
	Almost always	0	0.0	0	0.0	3	1.6	3	13.6	2	20.0
9) Bet more than you could afford to lose	Never	0	0.0	387	100	123	67.6	4	18.2	0	0.0
	Sometimes	0	0.0	0	0.0	57	31.3	15	68.2	6	60.0
	Most of the time	0	0.0	0	0.0	2	1.1	2	9.1	1	10.0
	Almost always	0	0.0	0	0.0	0	0.0	1	4.5	3	30.0

^aNongamblers were not asked CPGI items; ^bTwo respondents were indicated nonresponse or do not know for 1 CPGI item each.

Table 2.

Sociodemographic variables reported in cycle 1 by gambling type.

	Nongambler or Nonproblem gambler, n = 465, 68.5%	At-risk or problem gamblers, n = 214, 31.5%
Marital status, n (%)	χ2 = 2.917, df = 2, P = 0.23
Single (never married)	301 (64.7)	150 (70.1)
In a relationship	153 (32.9)	62 (29.0)
Married or common-law	11 (2.4)	2 (0.9)
Divorced, separated, or widowed	—	—
Sex, n (%)	χ2 = 24.499, df = 1, P < 0.001
Female	271 (58.3)	81 (37.9)
Male	194 (41.7)	133 (62.1)
Main activity past 12 months, n (%)	χ2 = 13.669, df = 3, P = 0.003
School	345 (74.2)	129 (60.3)
Working	106 (22.8)	74 (34.6)
Looking for work	6 (1.3)	4 (1.9)
Other	8 (1.7)	7 (3.3)
Total household income before taxes in the past 12 months, n (%)	χ2 = 5.190, df = 9, P = 0.82
<$10 000	7 (1.5)	5 (2.3)
$10 001–$19 999	3 (0.6)	3 (1.4)
$20 000–$29 999	13 (2.8)	6 (2.8)
$30 000–$39 999	11 (2.4)	4 (1.9)
$40 000–$49 999	12 (2.6)	4 (1.9)
$50 000–$59 999	13 (2.8)	9 (4.2)
$60 000–$79 999	33 (7.1)	14 (6.5)
$80 000–$99 999	30 (6.5)	15 (7.0)
≥$100 000	130 (28.0)	69 (32.2)
DK or NR	213 (45.8)	85 (39.7)
First identified ethnic group (other than Canadian), n (%)	χ2 = 14.275, df = 3, P = 0.16
European	321 (69.0)	143 (66.8)
Asian	35 (7.5)	24 (11.2)
Other	68 (14.6)	32 (15.0)
DK or NR	41 (8.8)	15 (7.0)
Religion	χ2 = 5.265, df = 5, P = 0.26
No religion, agnostic, or atheist	171 (36.8)	93 (43.5)
Christian	86 (18.5)	30 (14.0)
Roman Catholic	65 (14.0)	34 (15.9)
All Others	133 (28.6)	51 (23.8)
DK or NR	10 (2.2)	6 (2.8)

DK = don’t know; NR = no response

Table 3 presents the prevalence of the various combinations of at-risk or PG and mental disorder onset. Twenty-one per cent (21%) reported neither at-risk or PG nor mental disorders at any point during the study. Eighteen (18%) reported both at-risk or PG and mental disorder at cycle 1 and at follow-up in cycles 2, 3, or 4. About 6% of the sample experience at-risk or PG before mental disorders, while similar proportions (7%) had mental disorders before at-risk or PG.

Table 3.

Prevalence of at-risk or problem gambling and mental disorder onset.

	Total N = 572
Never any at-risk or problem gambling or mental disorders	n = 117 (20.5%)
At-risk or problem gambling at cycle 1 and mental disorders at cycle 2 to 4	n = 36 (6.3%)
At-risk or problem gambling at cycle 1 and no mental disorders	n = 25 (4.4%)
Mental disorders at cycle 1 and at-risk or problem gambling at cycle 2 to 4	n = 41 (7.2%)
Mental disorder at cycle 1 and no at-risk or problem gamble ever	n = 133 (23.3%)
Both at-risk or problem gambling and mental disorders at cycle 1 and at cycle 2 to 4	n = 100 (17.5%)
Neither at-risk or problem gambling or mental disorders at cycle 1 and both at-risk or problem gambling and mental disorders at cycle 2 to 4	n = 8 (1.4%)
Both at-risk or problem gambling and mental disorders at cycle 1, neither at-risk or problem gambling and mental disorders at cycle 2 to 4	n = 9 (1.6%)
Neither at-risk or problem gambling and mental disorders at cycle 1 and only mental disorders cycle 2 to 4	n = 56 (9.8%)
Neither at-risk or problem gambling nor mental disorder at cycle 1 and only at-risk or problem gambling at cycle 2 to 4	n = 16 (2.8%)
Both at-risk or problem gambling and mental disorder at cycle 1 and only mental disorder at cycle 2 to 4	n = 24 (4.2%)
Both at-risk or problem gambling and mental disorder at cycle 1 and only at-risk or problem gambling cycle 2 to 4	n = 7 (1.2%)

The cross-sectional analyses are presented in Table 4. The results indicate that at-risk or PG compared with nongambling or nonproblem gambling was associated with increased odds of major depressive disorder (AOR 2.33; 95% CI 1.47 to 3.68), OCD (AOR 2.57; 95% CI 1.25 to 5.29), alcohol dependence (AOR 2.48; 95% CI 1.57 to 3.88), any drug use (AOR 1.47; 95% CI 1.02 to 2.11), and any mental disorder (AOR 2.09; 95% CI 1.49 to 2.93) in models adjusting for sociodemographic variables. When further adjusting for other assessed mental disorders, all models remained significant with the exception of any drug use.

Table 4.

Cross-sectional analysis of gambling and mental disorders at cycle 1.

	Nongamblers and nonproblem gamblers	At-risk or problem gamblers
Major depressive disorder, n = 679	n = 59 (12.7%)	n = 50 (23.4%)
AOR-1 (95% CI)	1.00	2.33 (1.47 to 3.68), P < 0.001a
AOR-2 (95% CI)	1.00	1.99 (1.17 to 3.38), P = 0.01a
GAD, n = 679	n = 12 (2.6%)	n = 9 (4.2%)
AOR-1 (95% CI)	1.00	1.96 (0.76 to 5.04), P = 0.161
AOR-2 (95% CI)	1.00	1.71 (0.63 to 4.67), P = 0.30
OCD, n = 679	n = 20 (4.3%)	n = 18 (8.4%)
AOR-1 (95% CI)	1.00	2.57 (1.25 to 5.29), P = 0.01a
AOR-2 (95% CI)	1.00	2.35 (1.04 to 5.32), P = 0.04a
Alcohol dependence, n = 679	n = 55 (11.8%)	n = 52 (24.3%)
AOR-1 (95% CI)	1.00	2.48 (1.57 to 3.88), P < 0.001a
AOR-2 (95% CI)	1.00	2.22 (1.40 to 3.50), P = 0.001a
Any drug use in the past 12 months, n = 670	n = 175 (38.1%)	n = 102 (48.3%)
AOR-1 (95% CI)	1.00	1.47 (1.02 to 2.11), P = 0.04a
AOR-2 (95% CI)	1.00	1.22 (0.76 to 1.95), P = 0.41
Any mental health disorder, n = 581	n = 223 (48.6%)	n = 140 (66.4%)
AOR-1 (95% CI)	1.00	2.25 (1.54 to 3.27), P < 0.001
AOR-2 (95% CI)	n/a	n/a

^aSignificant variables are shown as P < 0.05.

AOR-1 = adjusted for sociodemographic variables; AOR-2 = adjusted for sociodemographic variables and presence of any other assessed mental disorder; GAD = generalized anxiety disorder; n/a = not applicable; OCD = obsessive–compulsive disorder

The results for the longitudinal analysis examining the relations between gambling and incident mental disorders are presented in Table 5. At-risk or PG at cycle 1 was significantly associated with increased odds of incident major depressive disorder (AOR 1.98; 95% CI 1.14 to 3.44), alcohol dependence (AOR 2.20; 95% CI 1.17 to 4.13), drug use (AOR 2.72; 95% CI 1.49 to 4.96), and any mental disorder (AOR 3.84; 95% CI 1.89 to 7.79) at cycle 2 through 4. These associations remained significant after further adjusting for comorbid mental disorders.

Table 5.

Longitudinal analysis of incident mental disorders at cycles 2, 3, or 4.

	Nongamblers and nonproblem gamblers	At-risk or problem gamblers
Major depressive disorder, n = 545	n = 52 (15.8%)	n = 31 (25.4%)
AOR-1 (95% CI)	1.00	1.98 (1.14 to 3.44), P = 0.02a
AOR-2 (95% CI)	1.00	1.99 (1.13 to 3.50), P = 0.02a
GAD, n = 532	n = 23 (6.4%)	n = 15 (9.7%)
AOR-1 (95% CI)	1.00	1.62 (0.77 to 3.41), P = 0.21
AOR-2 (95% CI)	1.00	1.40 (0.65 to 2.99), P = 0.39
OCD, n = 577	n = 15 (3.9%)	n = 6 (3.8%)
AOR-1 (95% CI)	1.00	0.95 (0.33 to 2.72), P = 0.92
AOR-2 (95% CI)	1.00	0.82 (0.29 to 2.36), P = 0.72
Alcohol dependence, n = 625	n = 18 (13.3%)	n = 13 (18.8%)
AOR-1 (95% CI)	1.00	2.20 (1.17 to 4.13), P = 0.01a
AOR-2 (95% CI)	1.00	1.98 (1.03 to 3.81), P = 0.04a
Any drug use in the past 12 months, n = 591	n = 52 (20.7%)	n = 37 (38.9%)
AOR-1 (95% CI)	1.00	2.72 (1.49 to 4.96), P = 0.001a
AOR-2 (95% CI)	1.00	2.85 (1.49 to 5.43), P = 0.001a
Any mental health disorder, n = 581	n = 252 (62.7%)	n = 138 (77.1%)
AOR-1 (95% CI)	1.00	3.84 (1.89 to 7.79), P < 0.001a
AOR-2 (95% CI)	n/a	n/a

^aSignificant variables are shown as P < 0.05.

AOR-1 = Adjusted for sociodemographic variables; AOR-2 = Adjusted for sociodemographic variables and presence of any other assessed mental disorder; GAD = generalized anxiety disorder; n/a = not applicable; OCD = obsessive–compulsive disorder.

The longitudinal results for the relations between mental disorders and SUDs and incident at-risk or PG are presented in Table 6. Illegal drug use within the past year (AOR 2.90; 95% CI 1.60 to 5.23) and any mental disorder (AOR 2.68; 95% CI 1.46 to 4.93) were each associated with increased odds of incident at-risk or PG in cycles 2, 3, or 4 after adjusting for sociodemographic variables.

Table 6.

Longitudinal analysis of incident gambling at cycles 2, 3, or 4.

	Nongambler and Nonproblem gamblers	At-risk or problem gamblers
		AOR	P
Major depressive disorder, n = 376	n = 39 (12.6%) 1.00	0.76 (0.30 to 1.94)	0.56
GAD, n = 376	n = 9 (2.9%) 1.00	1.80 (0.44 to 7.39)	0.41
OCD, n = 376	n = 16 (5.2%) 1.00	1.33 (0.41 to 4.31)	0.64
Alcohol dependence, n = 376	n = 35 (29.2%) 1.00	1.82 (0.83 to 3.99)	0.14
Any drug use in the past 12 months, n = 371	n = 98 (32.0%) 1.00	2.90 (1.60 to 5.23)a	<0.001a
Any mental health disorder, n = 371	n = 133 (43.5%) 1.00	2.68 (1.46 to 4.93)a	0.001a

^aSignificant variables are shown as P < 0.05.

AOR = adjusted for sociodemographic variables; GAD = generalized anxiety disorder; OCD = obsessive–compulsive disorder.

Discussion

With gambling activities potentially increasing from adolescence into adulthood, identifying mental disorders and SUDs associated with at-risk and PG in young adults is becoming more important. The cross-sectional findings from our study are consistent with previous work indicating that at-risk and PG is associated with mental disorders and SUDs.^{1,4,9,25–31} The novel findings from our study are as follows. Although at-risk or PG and mental disorders or illegal drug use were both equally likely to be the antecedent condition (6%, compared with 7%, respectively), at-risk or PG was associated with more new onset mental disorders (depression, alcohol dependence, and illegal drug use), compared with the reverse (illegal drug use was the only association with new onset at-risk or PG).

From a clinical and public health perspective, it is important to understand the temporal relation between at-risk and PG and mental disorders. Some studies have indicated that gambling problems are associated with new onset mental disorders using adult and older adult samples.^2,8,34 Previous research has indicated that gambling increases in adolescence and peaks in the early twenties.^19,22 Our findings are the first to identify that at-risk or PG at ages 18 to 20 years is associated with incident mental disorders and illegal drug use at 23 to 25 years. This extends previous studies with the use of prospective data and CIDI-SF diagnoses of 3 axis I disorders (that is, depression, GAD, and OCD) rather than a screening for only depressive symptoms.³³ Notably, at-risk and PG at cycle 1 was associated with increased odds of incident major depressive disorder, alcohol dependence, and illegal drug use and not associated with GAD and OCD. With somewhat stronger magnitude for at-risk gambling and incident externalizing disorders (with the exception of major depressive disorder), it may be that at-risk and PG increases the vulnerability for future externalizing disorders more so than for internalizing disorders. More research in this area is warranted. This research identifies a key period in the lifespan were the prevention of gambling problems as well as of mental disorders and illegal drug use is very important. As well, our study indicated that illegal drug use is associated with incident at-risk or gambling problems among young adults. Therefore, assessing and monitoring gambling behaviours among people using illegal drugs may be an important preventative strategy. With data indicating that cross-sectional associations and incident relations between at-risk gambling or PG and mental disorders, the most effective prevention strategies will likely be those that addressed both PG and mental disorders and illegal drug use.

Our study limitations should be considered when interpreting the findings. First, although the study sample was similar to sociodemographic characteristics of Manitobans aged 18 to 20 years, it was not a representative sample and these findings may not be generalizable to older adults or to adolescents. Second, although several incident mental disorders were assessed, there were many mental disorders that were not included (for example, panic disorder, PTSD, bipolar disorder) that would add to our understanding of gambling and mental health. Finally, due to lack of statistical power for addressing our specific research questions, cycles 2 to 4 of the data had to be collapsed resulting in only 2 time points, which does not allow for modelling trends overtime or patterns of remission. This identifies an important area for future research.

Conclusion

After adjusting for sociodemographic variables and other assessed mental health disorders (including drug use), at-risk or PG predicted later incident major depressive disorder, alcohol dependence, and drug use, while drug use also predicted incident at-risk or PG. The most consistent relations observed in the models were for new onset, bidirectional relations between at-risk or PG and illegal drug use. It is important to note that the criteria for certain mental conditions do change with different DSM editions. However, these findings provide support for the new placement of gambling disorder among substance-related and addictive disorders in the DSM-5. Health care professionals should be aware of the link between at-risk or PG and the potential for increased likelihood of incident mental disorders and illegal drug use among young adults. Specifically, health care professionals should know that at-risk or PG can lead to incident mental disorders and illegal drug use, as well as illegal drug use can lead to new onset at-risk or PG.