Problem gambling in psychotic disorders: A systematic review and meta‐analysis of prevalence

Problem gambling (PBG) is more common in people with mental health disorders, including substance use, bipolar, and personality disorders, than in the general population. Although individuals with psychotic disorders might be expected to be more vulnerable to PBG, fewer studies have focused on this comorbidity. The aim of this review was to estimate the prevalence of PBG in people with psychotic disorders.

appraisal for systematic reviews of prevalence data.The pooled prevalence of PBG was calculated using a fixed effects generalized linear mixed model and presented through forest plots.
Results: Of 1271 records screened, 12 studies (n = 3443) were included.The overall prevalence of PBG was 8.7% (95% CI = 7.8%-9.7%,I 2 = 69%).A lower prevalence was found in studies with a low risk of bias (5.6%; 95% CI = 4.4%-7.0%)compared with studies with a moderate risk of bias (10.4%; 95% CI = 9.2%-11.7%).Different methods used to assess PBG also contributed to the heterogeneity found.Conclusion: This meta-analysis found substantial heterogeneity, partly due to the risk of bias of the included studies and a lack of uniformity in PBG assessment.Although more research is needed to identify those at increased risk for PBG, its relatively high prevalence warrants routine screening for gambling in clinical practice.

| INTRODUCTION
][7][8][9][10] While antipsychotic medications effectively ameliorate positive symptoms (e.g., delusions, hallucinations), only a minority of individuals achieve full recovery. 11,12][18][19][20] Pathological gambling, now termed gambling disorder, was previously categorized as an impulse-control disorder in earlier iterations of the Diagnostic and Statistical Manual of Mental Disorders (DSM). 21,224][25] Although not formally included in the DSM-5 criteria, problem gambling (PBG) encompasses both gambling disorder and gambling behaviors that have adverse consequences for the individual, their social network, and the broader community without meeting the criteria for gambling disorder. 26,27These consequences include financial problems (e.g., job loss, bankruptcy), social isolation, and

Summations
• The results suggest an increased prevalence of problem gambling in people with psychotic disorders compared with the prevalence reported in the general population.• Routine screening for problem gambling and appropriate treatment should be integrated into usual care for people with psychotic disorders.

Limitations
• Relatively few studies have been conducted so far on this comorbidity, and those that have been published have used inconsistent methods to assess problem gambling.• The study samples included were mostly Caucasian men between the ages of 35 and 50 years, so studies in more diverse and vulnerable populations are needed for more generalizable findings.
8][29][30][31] The estimates of gambling disorder prevalence in the general population vary from 0.4% to 2.2%, [32][33][34] contingent upon variations in study methodologies and geographical location.In parallel, estimates of PBG prevalence vary from 0.12% to 6.4%, 32,34,35 being influenced by diverse contextual and methodological factors including the diagnostic instruments and thresholds employed for its definition.While a higher prevalence of PBG has been reported in people with mental health disorders, including substance use, bipolar, and personality disorders, compared with the general population, fewer studies have focused on psychotic disorders. 36,379][40][41][42] Furthermore, the presence of comorbid PBG may add to the deleterious consequences of psychotic disorders, thereby further impeding the recovery process for affected individuals. 43,44Within this framework, the main objective of this systematic review and meta-analysis was to determine the overall prevalence of PBG in individuals with psychotic disorders.

| METHODS
The results presented herein align with the reporting standards outlined in the Preferred Reporting Items for Systematic reviews and Meta-Analyses (PRISMA) 2020 statement. 45The protocol of this study was preregistered in the International Prospective Register of Systematic Reviews (PROSPERO) on June 30, 2023 (CRD42023427793).

| Search strategy and selection criteria
A comprehensive search strategy was developed in collaboration with a health-science librarian (Supplementary Table 1), covering the period from the inception of the databases up to November 1st, 2023.The databases searched were Medline (using Ovid), EMBASE, PsycINFO (using Ovid), CINAHL, Cochrane Central Register of Controlled Trials (CENTRAL), Web of Science, and ProQuest Dissertation and Thesis.The search strategy employed a combination of controlled terms specific to each database, supplemented with uncontrolled vocabulary relevant to gambling (e.g., gambling or betting) and psychotic disorders (e.g., psychosis or schizophrenia).Relevant articles were also identified by manually reviewing the reference lists of the included studies and Google Scholar.There were no language restrictions.Peer-reviewed studies meeting the following inclusion criteria were included: (1) involving a sample of individuals spanning all age groups, diagnosed with either schizophrenia spectrum or psychotic mood disorders as defined based on DSM/ICD criteria; (2) providing the prevalence data of PBG and explicitly stating the methods employed for assessment; (3) studies including observational (i.e., cohort, nested case-control, and cross-sectional) and experimental designs (i.e., randomized and non-randomized trials).There were no other exclusion criteria.
Following removal of duplicate entries, the titles and abstracts of all articles identified through the literature search were independently screened by the principal investigator (O.C.) and an undergraduate psychology student (E.A.).To ensure accuracy and consistency, the same two researchers individually evaluated the suitability of all potentially pertinent full-text articles.The screening process was facilitated using the Covidence systematic review software. 46Any disparities in judgment were subjected to comprehensive discussion, ultimately leading to a consensus, or resolved by a third reviewer (L.Béchard).In situations where multiple articles derived data from the same study sample, only the article featuring the largest sample size or presenting the most detailed results in regard to the data extracted for this review was included.

| Data extraction
The extraction of data was carried out by the principal investigator (O.C.) and two undergraduate psychology students (E.A., and L. Bachand).The information extracted included: (1) study design; (2) country, further categorized according to the regional distribution established by the World Health Organization (WHO), including Africa, Eastern Mediterranean, Europe, Americas, South-East Asia, and Western Pacific, as well as according to the World Bank income classifications for 2022 (i.e., high-income or low-and middle-income); (3) age and sex or gender criteria for inclusion; (4) mean age of the population; (5) proportion of men and women; (6) proportion of Caucasian and African/Afro-American; (7) study setting (i.e., inpatient, outpatient, and mixed); (8) sample size; (9) type of population studied (e.g., psychotic disorders exclusively, any mental health disorders, general population, or other); (10) diagnostic criteria used (e.g., DSM-5, DSM-IV, and ICD-10) to diagnose psychotic disorders; (11) breadth of psychotic disorders included (e.g., only schizophrenia, psychotic disorders excluding or including psychotic mood disorders).Quantitative outcomes extracted encompassed both the count of patients diagnosed with a psychotic disorder (in studies in which the population did not only include individuals with psychotic disorders) and the outcomes of PBG assessment specifically among people with psychotic disorders.Regarding PBG assessment, the specific method/definition used, such as DSM criteria or alternative assessment instruments (e.g., Problem Gambling Severity Index [PGSI], 26 South Oaks Gambling Screen [SOGS], 47 National Opinion Research Center DSM Screen for Gambling Problems [NODS] 48 ) and the corresponding threshold utilized were documented.Where possible, and when DSM criteria were not employed, the prevalence of PBG measured using the threshold recognized as problematic was documented (e.g., PGSI = 8+, SOGS = 5+, and NODS = 5+). 26,47,48The time frame of the assessment was also collected (e.g., lifetime, pastyear).Study authors were contacted in cases of missing data.Any disagreements that arose were thoroughly discussed and resolved through consensus or by a third reviewer (L.Béchard).

| Study quality
The evaluation of the risk of bias was carried out by the principal investigator (O.C.) and two undergraduate psychology students (E.A., L. Bachand) using the Joanna Briggs Institute (JBI) critical appraisal checklist for systematic reviews of prevalence data. 49This 9-item checklist is recognized as the most suitable tool for assessing the methodological quality of prevalence studies. 50The adequacy of the study sample was evaluated based on a hypothesized PBG prevalence of 5% among individuals with psychotic disorders and a 95% confidence level. 51ach study was assigned a quality score ranging from 0 to 9, with 1 point allocated for meeting each criterion.Subsequently, an overall risk of bias was determined, categorized as high (0-3), moderate (4-6), or low (7-9) risk. 49These steps were facilitated by the Covidence systematic review software. 46In instances of disagreements, consensus was reached through comprehensive discussions or resolved by a third reviewer (L.Béchard).

| Statistical analysis
The main characteristics of each study were summarized using descriptive statistics.The overall pooled prevalence estimate of PBG and its corresponding 95% confidence interval (95% CI) were calculated by pooling the raw prevalence from each included study using a fixed effects generalized linear mixed model. 52The presence of heterogeneity was evaluated using a χ 2 test, the betweenstudy variance with τ 2 , and the portion of total variation across studies due to heterogeneity with the I 2 statistic.I 2 values ranging from 0%-25%, 26%-50%, 51%-75%, and 76%-100% indicated low, moderate, substantial, and considerable heterogeneity, respectively. 53To explore potential sources of heterogeneity, subgroup meta-analyses were conducted for predetermined categorical moderators, including study publication year (before/after classification of gambling disorder as a behavioral addiction in the DSM-5, i.e., 2013), geographic region, country income, study setting and type of population, psychotic disorder diagnoses, PBG assessment method, and risk of bias.The results are visually presented through forest plots.All reported CIs represent the 95% range.All statistical analyses were conducted with R software, version 4.3.0(R Project for Statistical Computing), using the meta and metafor packages. 54,55Analyses were performed by the principal investigator (O.C.) and a biostatistician (P.H.C.).

| Modifications to the review protocol
In the review protocol, a meta-analysis using a random effects model was planned, as high heterogeneity was expected due to potential between-study variance.However, of all the moderators examined, the risk of bias of the included studies was the main factor contributing to the substantial heterogeneity found.Studies with a moderate risk of bias generally had smaller sample sizes than studies with a low risk of bias.Since smaller studies are assigned more weight in a random effects model than in a fixed effects model, the latter was considered more appropriate for this meta-analysis to mitigate bias. 56,57Furthermore, given the relatively small total sample sizes of the included studies and the low crude event rates obtained, generalized linear mixed models were preferred over inverse-variance methods because of their greater efficiency in such a scenario. 52,58Sensitivity analyses were performed using the a priori specified methods (i.e., random effects model using the inverse-variance method).Regarding publication bias, although the protocol called for a quantitative analysis, it was assessed qualitatively because methods such as funnel plots and Egger's test are not readily applicable to metaanalysis of prevalence. 59

| RESULTS
The electronic search yielded a total of 1271 unique records after removal of duplicates.1][62][63][64][65][66][67][68][69][70][71][72][73] Two of these studies were excluded from the quantitative synthesis; a study by André et al., 60 which reported a PBG prevalence of 100% in a sample consisting of only one individual with a psychotic disorder, while a study by Bland et al., 63 which reported a PBG prevalence of 0% in 50 individuals with psychotic disorders, was the only study to use a survey conducted in a general population sample and was therefore excluded due to a distinct methodology as well as a high risk of bias assessment.A sensitivity analysis including these two studies was nevertheless performed.

| Study characteristics
A total of 12 studies were included in the meta-analysis, collectively encompassing data from 3443 individuals diagnosed with psychotic disorders.As detailed in Table 1, most studies were cross-sectional (7/12; 58.3%) and only five studies (41.7%) were conducted specifically in people with psychotic disorders.The majority of studies (9/12; 75.0%) did not include people with psychotic mood disorders, including one study with schizophrenia only.Results of PBG assessment are shown in Table 2.One study assessed PBG using the DSM-IV-R diagnostic criteria for pathological gambling, while one study used the DSM-5 criteria for gambling disorder.Five studies (41.7%) used the PGSI, four of which used the 8+ threshold for PBG and one used the 3+ threshold for at-risk gambling, two studies (16.7%) used the NODS, both with the 5+ threshold for PBG, and one study (8.3%) employed the SOGS with the 5+ threshold for PBG.Two studies did not use any validated assessment tool; patients who reported being homeless because of gambling were classified as having PBG in a cross-sectional study of homeless clinic attendees, 71 while PBG was defined as repeated gambling with disproportionate spending in a retrospective cohort chart review. 73One study reported a lifetime prevalence of PBG, two studies did not specify any assessment time frame, and all other studies reported current/ past-year prevalences.As for study quality, the mean score was 6.1 (range, 4-9; Supplementary Table 2); five studies (41.7%) were classified as presenting a low risk of bias, while the remaining seven studies (58.3%) had a moderate risk of bias.
Except for one study conducted in Africa, the five studies that reported ethnicity were predominantly Caucasian, ranging from 57% to 96% of the total population F I G U R E 1 Flow chart of reviewed articles.Adapted from Page et al. 45 T A B L E 1 Characteristics of the included studies.(Supplementary Table 3).The mean age was less than 25 years in only two studies and varied from 34.0 to 53.7 years in the others.Of the 12 studies included in the meta-analysis, one had a majority of women (52%), while the proportion of men ranged from 51% to 94% in the remaining studies.

| Sensitivity analyses
The results obtained using the statistical analyses a priori specified in the review protocol were consistent with those reported.The overall pooled prevalence of PBG using a random effects model and the Freeman-Tukey double arcsine transformation was 8.0% (95% CI, 6.1%-10.1%,I 2 = 70%), with a lower prevalence estimated in the studies with a low risk of bias compared with those with a moderate risk of bias (5.5% [95% CI, 4.2%-6.9%],I 2 = 0%, and 10.1% [95% CI, 7.9%-12.6%],I 2 = 49%, Forest plot of the pooled estimated prevalence of problem gambling in people with psychotic disorders according to risk of bias.CI, confidence interval. respectively, p < 0.01; Supplementary Figure 7).A further sensitivity analysis was performed while including the two excluded studies in a generalized linear mixed model (Supplementary Figure 8).Their inclusion did not significantly change the overall pooled prevalence of PBG (8.6% [95% CI, 7.8%-9.6%],I 2 = 63%).Finally, an additional analysis was performed excluding the only study that did not report the prevalence of PBG using the PGSI threshold recognized as PBG (Supplementary Figure 9), which again did not significantly alter the overall estimated prevalence obtained (9.2% [95% CI, 8.2%-10.3%],I 2 = 63%).

| Summary of main findings
In this meta-analysis, the overall prevalence of PBG in people with psychotic disorders was estimated to be 8.7%, albeit with a significant effect of the risk of bias of the included studies; the pooled prevalence was 5.6% in the five studies with a low risk of bias, and 10.4% in the seven studies with a moderate risk of bias.The main factors contributing to the risk of bias in the included studies were the lack of a reliable measure of either PBG or psychotic disorders, and an inappropriate sample frame and/or sampling method.Indeed, the lack of a systematic procedure for PBG assessment, in addition to different assessment methods and definitions between studies, was evident in a number of studies.Of note, differences in the prevalence of PBG as assessed using different methods were found in subgroup analyses.Intriguingly, the prevalence of PBG estimated with validated instruments, mainly the PGSI, was lower than that found in the studies using the DSM criteria, although only two studies employed the latter, one of which reported a very high prevalence of PBG in Africa in a sample of high school, college, and university students at risk for schizophrenia.
As such, the samples across the included studies were also highly heterogeneous, ranging from inpatients with psychotic disorders only to outpatients with a range of mental health disorders, including but not limited to psychotic disorders.While this could have contributed to the substantial heterogeneity found in this meta-analysis, subgroup meta-analyses could not detect any contributing factors other than the risk of bias and the different methods of PBG assessment, perhaps due to the relatively small number of included studies.
Although direct comparisons cannot be made based on the available evidence, the results of this metaanalysis suggest that PBG is more common in people with psychotic disorders than in the general population.Indeed, a recent meta-analysis of 23 prevalence studies published between 2016 and 2022, including 124,264 individuals, reported a pooled prevalence of PBG in adults from the general population of 1.29% (95% CI, 0.63%-1.95%). 35While this may be expected given the high prevalence of psychiatric comorbidity in people with psychotic disorders, and is particularly well documented for substance use disorders, this is the first quantitative synthesis of the evidence for behavioral addictions, in this case PBG.In addition to possible genetic and neurobiological susceptibilities, 74 one could hypothesize an overrepresentation of risk factors for PBG in this population, such as substance use, mental health comorbidities (e.g., anxiety, depression), and socioeconomic precariousness. 42While these risk factors have been documented in the general population, it remains unclear whether they are the same in people with a psychotic disorder, and the literature on this topic is currently scarce and inconclusive, as highlighted in a recent review. 75Furthermore, this meta-analysis did not find an association between factors such as geographic region, clinical setting, type of psychotic disorder diagnosis (i.e., excluding or including psychotic mood disorders) and the prevalence of PBG.[78][79]

| Strengths and limitations
This systematic review used a robust design in accordance with the best recommendations for conducting meta-analyses of prevalence studies. 49,59As such, statistical methods were planned a priori, and deviations were described and justified, complemented by sensitivity analyses, ensuring transparency at every stage of the review.Although such a meta-analysis must inevitably include studies with heterogeneous methodological designs and greater or lesser risks of bias, subgroup analyses provided a more reliable estimate of the prevalence of PBG in the studies with a low risk of bias.To this end, although the methods used to assess PBG were not consistent across studies, contacts with the authors of some studies made it possible to report prevalences based on accepted PBG thresholds when instruments other than the DSM criteria were used.Despite this lack of uniformity in the methods used to define PBG, and the change in diagnostic classification made in 2013 with the DSM-5, the fact remains that the concept measured in all included studies was that of a problematic gambling behavior that had a negative impact on the individuals and those around them.Finally, several sensitivity analyses were performed and were consistent with the results obtained (e.g., including studies that were excluded from the main meta-analysis due to degeneracy or using a random effects model with the inverse-variance method).
In addition to the limitations associated with the heterogeneity of the included studies, publication bias, which cannot be readily quantified in meta-analyses of prevalence studies, 59 cannot be excluded.To this end, subgroup analyses revealed a higher prevalence of PBG in studies with a higher risk of bias compared with studies with a lower risk of bias, which could indicate publication bias.However, there were both small and large studies reporting low and high estimates of PBG prevalence, reducing the likelihood that publication bias had a significant effect on the results obtained.The lack of information in many of the included studies, particularly with regards to the sociodemographic characteristics of the samples, coupled with the limited number of studies published to date, makes it impossible to verify the existence of associations between some of these variables and the prevalence of PBG.In addition, the cross-sectional nature of most of the studies precludes the determination of a causal relationship and thus the identification of potential risk factors for PBG specific to people with a psychotic disorder.Finally, the samples studied are not representative of the global population as a whole, with an overrepresentation of Caucasian men from developed high-income countries, aged around 35-50 years.Efforts must be made to study this comorbidity in more diverse and vulnerable populations, where PBG may be even more prejudicial.
To conclude, in light of the detrimental repercussions of PBG, particularly amplified among individuals with psychotic disorders, there is a pressing need for further investigation into this coexisting condition.Conducting prospective studies involving diverse populations could be pivotal in identifying those most vulnerable to its effects. 80Additionally, considering the associations between PBG and problematic video gaming, 81 notably prevalent among young adults with psychotic disorders, 82 it is imperative to prioritize research that integrates these two behaviors.4][85] Until more tailored tools and interventions are developed, routine screening for PBG among individuals with psychotic disorders should be consistently implemented within clinical settings.This proactive approach can pave the way for better identification and support for those affected.
bOne study conducted in Africa.