Psychological therapies for pathological and problem gambling

Abstract

Background

Various psychological therapies for pathological and problem gambling have been evaluated in randomised trials. A synthesis of best‐quality evidence is required.

Objectives

The objective was to synthesise evidence from randomised trials of psychological therapies for pathological and problem gambling (cognitive‐behaviour therapy (CBT), motivational interviewing therapy, integrative therapy, other psychological therapy), in order to indicate the efficacy of therapies and durability of therapy effects, relative to control conditions.

Search methods

We conducted a search of the Cochrane Depression, Anxiety and Neurosis Review Group's Specialised Register (CCDANCTR), which includes relevant randomised controlled trials from the following bibliographic databases: CENTRAL (The Cochrane Central Register of Controlled Trials) (all years), EMBASE (1974 ‐), MEDLINE (1950 ‐) and PsycINFO (1967 ‐). We also carried out complementary searches of MEDLINE, EMBASE, PsycINFO, LILACS and CENTRAL for studies published between January 1980 and October 2011. We examined the WHO International Clinical Trials Registry Platform and ClinicalTrials.gov and also conducted manual searches of selected journals and reference lists of included studies.

Selection criteria

Included studies were clinical trials using random allocation to groups, considering pathological or problem gamblers, and evaluating a psychological therapy for pathological or problem gambling. Control conditions included 'no treatment' controls, referral to Gamblers Anonymous and non‐specific treatment component controls.

Data collection and analysis

We systematically extracted data on the characteristics and results of studies. Primary outcomes were measures of gambling symptom severity, financial loss from gambling and frequency of gambling. Secondary outcomes were occurrence of pathological gambling diagnoses and depression and anxiety symptoms. Treatment effects were defined by comparisons between therapy and control conditions at post‐treatment assessments (conducted from 0 to 3 months following completion of treatment) and follow‐up assessments (conducted from 9 to 12 months following completion of treatment), respectively, using the standardised mean difference (SMD) or risk ratio (RR). We synthesised results through random‐effects meta‐analysis.

Main results

Fourteen studies (n = 1245) met the inclusion criteria. Eleven studies compared CBT with control and comparisons at 0 to 3 months post‐treatment showed beneficial effects of therapy that ranged from medium (when defined by financial loss from gambling: SMD ‐0.52; 95% confidence interval (CI) ‐0.71 to ‐0.33, n = 505) to very large (for gambling symptom severity: SMD ‐1.82; 95% CI ‐2.61 to ‐1.02, n = 402). Only one study (n = 147) compared groups at 9 to 12 months follow‐up and produced smaller effects that were not significant. Four studies of motivational interviewing therapy were identified and mainly considered samples demonstrating less severe gambling (relative to studies of pathological gamblers). Data suggested reduced financial loss from gambling following motivational interviewing therapy at 0 to 3 months post‐treatment (SMD ‐0.41; 95% CI ‐0.75 to ‐0.07, n = 244), although comparisons on other outcomes were not significant. The effect approached zero when defined by gambling symptom severity (SMD ‐0.03; 95% CI ‐0.55 to 0.50, n = 163). Studies compared groups at 9 to 12 months follow‐up and found a significant effect of motivational interviewing therapy in terms of frequency of gambling (SMD ‐0.53; 95% CI ‐1.04 to ‐0.02, n = 62), with comparisons on other outcomes that were not significant. Two studies of integrative therapies also considered samples demonstrating overall low gambling severity, and found no significant effects of therapy at 0 to 3 months post‐treatment. Comparisons at 9 to 12 months follow‐up suggested a medium effect from therapy in terms of gambling symptom severity, with no significant differences for other outcomes. One study (n = 18) considered another psychological therapy (i.e.Twelve‐Step Facilitated Group Therapy) and suggested beneficial effects in terms of most outcomes at 0 to 3 months post‐treatment. The evidence supporting these various classes of therapy ranged from very low to low quality.

Authors' conclusions

This review supports the efficacy of CBT in reducing gambling behaviour and other symptoms of pathological and problem gambling immediately following therapy. However, the durability of therapeutic gain is unknown. There is preliminary evidence for some benefits from motivational interviewing therapy in terms of reduced gambling behaviour, although not necessarily other symptoms of pathological and problem gambling. However, the findings are based on few studies and additional research is needed to inform conclusions. There is also evidence suggestive of some possible benefit from integrative therapies, and other psychological therapies for pathological and problem gambling. However, there are too few studies and evidence is insufficient to evaluate these therapies. The majority of studies in this review varied in risk of bias, and much of the evidence comes from studies with multiple limitations. The current data may thus reflect overestimates of treatment efficacy.

Plain language summary

Psychological therapies for pathological and problem gambling

Psychological therapies have been proposed for the treatment of pathological and problem gambling, and this review summarised current evidence for these therapies. It included best‐quality randomised trials, where therapy was compared with conditions including 'no treatment’ controls or referral to Gamblers Anonymous. It considered categories of therapy including: (1) cognitive‐behaviour therapy (CBT); (2) motivational interviewing therapy; (3) integrative therapy; and (4) other psychological therapy. The search identified 14 studies and we combined data from these. Data from nine studies indicated benefits of CBT in the period immediately following treatment. However, there were few studies across longer periods of time (e.g. 12 months) after treatment, and little is known about whether effects of CBT are lasting. Data from three studies of motivational interviewing therapy suggested some benefits in terms of reduced gambling behaviour, but not necessarily other symptoms of pathological and problem gambling. However, the data come from few studies and conclusions regarding motivational interviewing therapy require further research. There were also few studies that provided evidence on integrative therapies (two studies) and other psychological therapies (one study), and there is currently insufficient data to evaluate the efficacy of these therapies

Summary of findings

Summary of findings for the main comparison. CBT versus control for pathological and problem gambling.

CBT versus control for pathological and problem gambling
Patient or population: patients with pathological and problem gambling Settings: outpatient, addiction or mental health, university and general community settings Intervention: CBT
Outcomes	Illustrative comparative risks* (95% CI)		Relative effect (95% CI)	No of participants (studies)	Quality of the evidence (GRADE)	Comments
	Assumed risk	Corresponding risk
	Control	Cognitive‐behaviour therapy (CBT) versus control
Gambling symptom severity ‐ comparisons at 0 to 3 months post‐treatment Lower scores indicate reduced gambling symptom severity, and thus positive outcomes1		Across studies, the typical scores in CBT conditions were, on average, 1.82 standard deviations lower (with a 95% CI ranging from 2.61 to 1.02 standard deviations lower) than scores in the control conditions		402 (7 studies)	⊕⊝⊝⊝ very low2,3,4	This corresponds to a very large effect according to conventions proposed by Cohen 1992
Financial loss from gambling ‐ comparisons at 0 to 3 months post‐treatment Lower scores indicate reduced financial loss from gambling, and thus positive outcomes1		Across studies, the typical scores in CBT conditions were, on average, 0.52 standard deviations lower (with a 95% CI ranging from 0.71 to 0.33 standard deviations lower) than scores in the control conditions		505 (7 studies)	⊕⊕⊝⊝ low2,5	This corresponds to a medium effect according to conventions proposed by Cohen 1992
Frequency of gambling ‐ comparisons at 0 to 3 months post‐treatment Lower scores indicate reduced severity of gambling, and thus positive outcomes1		Across studies, the typical scores in CBT conditions were, on average, 0.78 standard deviations lower (with a 95% CI ranging from 1.11 to 0.45 standard deviations lower) than scores in the control conditions		505 (7 studies)	⊕⊕⊝⊝ low2,5,6,7	This corresponds to a large effect according to conventions proposed by Cohen 1992
Diagnoses of pathological gambling ‐ comparisons at 0 to 3 months post‐treatment Lower scores indicate reduced rates of pathological gambling, and thus positive outcomes1	Across studies, the typical rate of pathological gambling in the control conditions was 81.2 per 100	Across studies, the typical rate of pathological gambling in the CBT conditions was 10.6 per 100 (with a 95% CI ranging from 4.1 to 25.2)	RR 0.13 (95% CI 0.05 to 0.31)	81 (2 studies)	⊕⊕⊝⊝ low2,5,8
Depression symptoms ‐ comparisons at 0 to 3 months post‐treatment Lower scores indicate reduced depression symptoms, and thus positive outcomes1		Across studies, the typical scores in CBT conditions were, on average, 0.66 standard deviations lower (with a 95% CI ranging from 0.93 to 0.39 lower) than scores in the control conditions		276 (4 studies)	⊕⊕⊝⊝ low2,5	This corresponds to a medium to large effect according to conventions proposed by Cohen 1992
Anxiety symptoms ‐ comparisons at 0 to 3 months post‐treatment Lower scores indicate reduced anxiety symptoms, and thus positive outcomes1		Across studies, the typical scores in CBT conditions were, on average, 0.64 standard deviations lower (with a 95% CI ranging from 0.90 to 0.37 lower) than scores in the control conditions		276 (4 studies)	⊕⊕⊝⊝ low2,5	This corresponds to a medium to large effect according to conventions proposed by Cohen 1992
*The basis for the assumed risk is the median control group risk, across studies. The corresponding risk (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CBT: cognitive‐behaviour therapy; CI: confidence interval; RR: risk ratio.
GRADE Working Group grades of evidence High quality: Further research is very unlikely to change our confidence in the estimate of effect. Moderate quality: Further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate. Low quality: Further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate. Very low quality: We are very uncertain about the estimate.

¹ There is currently no consensus on a minimally important difference in this field.

² All studies used random allocation. However, all suffered a number of other limitations. Several studies (4/9) did not clearly describe their method of random allocation. Only 3/9 studies clearly indicated blinding of outcome assessors. 4/9 studies used completers only analysis while 1/9 was identified as having an appropriate method of allocation concealment. 
 ³ There were very large amounts of statistical heterogeneity. Some studies provided unrealistically large estimates of the treatment effect which are likely to be overestimates of the true effects of therapy. 
 ⁴ There were very wide confidence intervals around the summary effect. 
 ⁵ There were moderately wide confidence intervals around the summary effect. 
 ⁶ There were moderate amounts of statistical heterogeneity. 
 ⁷ All studies showed beneficial effects of therapy. 
 ⁸ There were few participants in the studies.

Summary of findings 2. MI therapy versus control for pathological and problem gambling.

MI therapy versus control for pathological and problem gambling
Patient or population: patients with pathological and problem gambling Settings: outpatient and college settings Intervention: MI therapy
Outcomes	Illustrative comparative risks* (95% CI)		Relative effect (95% CI)	No of participants (studies)	Quality of the evidence (GRADE)	Comments
	Assumed risk	Corresponding risk
	Control	Motivational interviewing versus control
Gambling symptom severity ‐ comparisons at 0 to 3 months post‐treatment Lower scores indicate reduced gambling symptom severity, and thus positive outcomes1		Across studies, the typical scores in MI conditions were, on average, 0.03 standard deviations lower (with a 95% CI ranging from 0.55 standard deviations lower to 0.50 standard deviations higher) than scores in the control conditions		163 (2 studies)	⊕⊝⊝⊝ very low2,3,4	This effect approaches zero
Financial loss from gambling ‐ comparisons at 0 to 3 months post‐treatment Lower scores indicate reduced financial loss from gambling, and thus positive outcomes1		Across studies, the typical scores in MI conditions were, on average, 0.41 standard deviations lower (with a 95% CI ranging from 0.75 to 0.07 standard deviations lower) than scores in the control conditions		244 (3 studies)	⊕⊝⊝⊝ very low2,3,4	This corresponds to a medium effect according to conventions proposed by Cohen 1992
Frequency of gambling ‐ comparisons at 0 to 3 months post‐treatment Lower scores indicate reduced financial loss from gambling, and thus positive outcomes1		Across studies, the typical scores in MI conditions were, on average, 0.18 standard deviations lower (with a 95% CI ranging from 0.50 standard deviations lower to 0.15 standard deviations higher) than scores in the control conditions		145 (2 studies)	⊕⊝⊝⊝ very low3,4	This corresponds to a small effect according to conventions proposed by Cohen 1992. The effect is not significantly different from zero.
* The corresponding risk (and its 95% confidence interval) is based on the effect of the intervention (and its 95% CI) relative to the control condition. CI: confidence interval; MI: motivational interviewing
GRADE Working Group grades of evidence High quality: Further research is very unlikely to change our confidence in the estimate of effect. Moderate quality: Further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate. Low quality: Further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate. Very low quality: We are very uncertain about the estimate.

¹ There is currently no consensus on a minimally important difference in this field.

² There were moderate amounts of statistical heterogeneity. 
 ³ There a limited number of studies that have considered a restricted range of sample types and treatment settings. 
 ⁴ There were few participants in the studies.

Summary of findings 3. Integrative therapy versus control for pathological and problem gambling.

Integrative therapy versus control for pathological and problem gambling
Patient or population: patients with pathological and problem gambling Settings: unclear Intervention: integrative therapy
Outcomes	Illustrative comparative risks* (95% CI)		Relative effect (95% CI)	No of participants (studies)	Quality of the evidence (GRADE)	Comments
	Assumed risk	Corresponding risk
	Control	Integrative therapy versus control
Gambling symptom severity ‐ comparisons at 0 to 3 months post‐treatment Lower scores indicate reduced gambling symptom severity, and thus positive outcomes1		Across studies, the typical scores in integrative therapy conditions were, on average, 0.15 standard deviations lower (with a 95% CI ranging from 0.50 standard deviations lower to 0.19 standard deviations higher) than scores in the control conditions		137 (2 studies)	⊕⊝⊝⊝ very low2,3	This effect is not significantly different from zero
Financial loss from gambling ‐ comparisons at 0 to 3 months post‐treatment Lower scores indicate reduced financial loss from gambling, and thus positive outcomes1		Across studies, the typical scores in integrative therapy conditions were, on average, 0.41 standard deviations lower (with a 95% CI ranging from 1.04 standard deviations lower to 0.23 standard deviations higher) than scores in the control conditions		137 (2 studies)	⊕⊝⊝⊝ very low2,3,4	This corresponds to a medium effect according to conventions proposed by Cohen 1992. However, given the small number of studies and participants, this effect is not significantly different from zero.
Frequency of gambling ‐ comparisons at 0 to 3 months post‐treatment Lower scores indicate reduced financial loss from gambling, and thus positive outcomes1		Across studies, the typical scores in integrative therapy conditions were, on average, 0.12 standard deviations lower (with a 95% CI ranging from 0.69 standard deviations lower to 0.46 standard deviations higher) than scores in the control conditions		52 (1 study)	⊕⊕⊝⊝ low5	This effect is not significantly different from zero
*The corresponding risk (and its 95% confidence interval) is based on the effect of the intervention (and its 95% CI) relative to the control condition. CI: confidence interval.
GRADE Working Group grades of evidence High quality: Further research is very unlikely to change our confidence in the estimate of effect. Moderate quality: Further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate. Low quality: Further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate. Very low quality: We are very uncertain about the estimate.

¹ There is currently no consensus on a minimally important difference in this field.

² There a limited number of studies that have considered a restricted range of sample types and treatment settings. 
 ³ There are few participants in the studies. 
 ⁴ There were moderate amounts of statistical heterogeneity. 
 ⁵ There is only 1 trial that considered a college sample.

Background

Description of the condition

Gambling is a common recreational activity, with between 70 and 85 per cent of adults in some Western countries reporting episodes of gambling activity in the past year (Volberg 2006; Wardle 2010). Although not typically a reflection of disordered behaviour, gambling can sometimes escalate to problematic levels characterised by impaired control and adverse personal and social consequences. Shaffer 2002 has described pathological and problem gambling as a global public health concern, associated with impaired health, unemployment, financial difficulties and family violence.

Various terms have been used to describe harmful gambling behaviour, including pathological, problem, at‐risk, disordered and compulsive gambling. The term ‘pathological gambling’ is used frequently and is derived from psychiatric diagnostic systems such as the Diagnostic and Statistical Manual of Mental Disorders, 4^th edition (DSM‐IV; APA 2000). Although currently classified in the DSM‐IV as a disorder of impulse control, the criteria are modelled on those of substance related disorders and include preoccupation with gambling, failed attempts to reduce gambling, and restlessness or irritability when prevented from gambling (see Hodgins 2011). Pathological gambling is likely to be re‐classified under a broad category of addictive disorders in the upcoming DSM‐V (see Petry 2010). The term ‘problem gambling’ is also used frequently and is occasionally interchangable with pathological gambling. However, problem gambling is also sometimes used to describe a subclinical level of the psychiatric disorder or, alternatively, a broader category of severe gambling based on a continuum model of gambling‐related harm. This continuum extends from ‘non‐problem gambling’ to ‘at‐risk’ and then ‘problem gambling’, the latter of which subsumes cases of severe harm that might otherwise receive a diagnosis of pathological gambling (Ferris 2001). Other terms are used less frequently, but all endeavour to distinguish controlled gambling from behaviour that may cause significant problems to the gambler, their family and the community.

The prevalence of pathological and problem gambling has been found to vary internationally, with studies suggesting anywhere between 0.2% (in Norway) and 5.3% (in Hong Kong) of individuals that are pathological or problem gamblers (Hodgins 2011). Various risk factors are implicated in the development and maintenance of the disorder, including demographic variables (such as young age and male sex; Johansson 2009), personality dispositions (e.g., impulsivity traits) and cognitive biases (e.g., inaccurate beliefs about randomness; Sharpe 2002). Aspects of mental health comorbidity (e.g., major depression), neurocognition (e.g., deficits in working memory, inhibition) and alterations in neurobiology (e.g., dopaminergic and opioidergic neurotransmitter systems) have also been connected to gambling disorders (Hodgins 2011). However, the aetiologies of gambling problems are presumed to be heterogeneous and have been distinguished by pathways to problem gambling (that reflect, for example, the varying roles of pre‐morbid psychopathology and deficient coping skills; Blaszczynski 2002). Heterogeneity of pathological and problem gambling has also been postulated and defined by the functions of gambling (e.g., enhancing positive affect versus decreasing negative affect; Stewart 2008) and preferred types of gambling activity (e.g., electronic gaming machines versus sports betting; Petry 2003). Although such models of heterogeneity have received limited empirical consideration (Milosevic 2010), they highlight factors underlying onset and maintenance of gambling that are relevant to the development and delivery of treatment.

Description of the intervention

Various treatment options for pathological and problem gambling have been considered in the literature. Pharmacological therapies target hypothesised dysregulation in neurotransmitter systems (see Achab 2011), while psychological therapies address presumed psychological (e.g., cognitions) and contextual (e.g., environmental cues) determinants of gambling. Psychological therapies vary in intensity from minimal interventions (that may be defined by one therapist contact of limited duration; e.g., 5 to 30 minutes), to brief (one to four sessions), moderate (five to seven sessions) and intensive (eight or more sessions; Babor 1994) therapies. Minimal and brief interventions are typically intended for individuals with low‐level problems, or those who are otherwise disinclined to attend intensive therapy (Hodgins 2011). Psychological therapies have been delivered mainly through face‐to‐face sessions, with some also adapted for remote delivery (e.g., through telephone or internet; Carlbring 2008). Self help programmes, using self completed components (e.g., workbooks; Hodgins 2001) or peer support networks (i.e., Gamblers Anonymous), are also available.

The main psychological therapies for pathological and problem gambling are based on cognitive and behavioural models. Cognitive therapies focus on correcting erroneous beliefs about gambling, as well as biased information processing. Inaccurate beliefs about statistical independence and randomness are core cognitive distortions in pathological and problem gambling (Ladouceur 1996). Other cognitive biases include overconfidence in ability to identify systems of winning, beliefs that attitudes or behaviours can influence gambling outcomes, and tendencies to recollect wins and not losses (Toneatto 1999; Toneatto 2007a). Associated therapies may attempt to correct these misconceptions, such as through use of evidence to demonstrate independence of play. Behavioural therapies, in contrast, are based on assumptions that gambling is a maladaptive learned behaviour. Early exposure to gambling may lead to initiation of the behaviour (Sharpe 1993), which is maintained through conditioning processes. Positive reinforcement may involve a variable‐ratio schedule of financial gain, along with negative reinforcement from escape of negative emotional states. Arousal from gambling may become associated with gambling‐related external stimuli, and this may also maintain the behaviour (Sharpe 1993). Behavioural therapies are typically based on principles of classical and operant conditioning, and can include aversive therapy, systematic desensitisation and behavioural counselling. Behavioural interventions can also include activity planning, problem‐solving training, and social skills and communication training (see Dowling 2008). Cognitive‐behavioural therapy (CBT) combines various aspects of these therapeutic techniques.

Alternative psychological therapies for pathological and problem gambling are informed by more general accounts of behaviour change. The main class of such interventions are motivational interviewing therapies, as developed initially in the context of treatment for problem drinking (Miller 1983). Motivational interviewing is a client‐centred counselling style and clinical method that works from the assumption that a primary obstacle to change is ambivalence (Hettema 2005). This ambivalence takes the form of conflict between courses of action (e.g., changing versus sustaining behaviour), each of which is associated with benefits and costs. Perceived benefits of sustaining behaviour, and costs associated with changing, are hypothesised to create resistance to change. Although conceptual accounts of motivational interviewing are not fully developed, current explanations focus mainly on specific techniques (e.g., targeted reflective listening) that are used to elicit client change talk (i.e., verbalisations that favour change). Having clients describe and listen to their own statements about change are assumed to reduce ambivalence and strengthen commitment to change (e.g., by producing dissonance between present behaviour and the desired goal or self image; Miller 2009). Motivational interviewing is also proposed to function through qualities of the therapist‐client relationship, which "(a) is collaborative rather than authoritarian, (b) evokes the client's own motivation rather than trying to instil it, and (c) honors the client's autonomy" (Miller 2009; p. 533). This relational context is expected to promote client involvement in therapy and the process of change, and also support self efficacy. There are various therapeutic techniques and adaptations of motivational interviewing that are based on these principles and clinical methods (see Hettema 2005).

Other psychological accounts of pathological and problem gambling (e.g., psychodynamic models; Bergler 1957) have been described, including multi‐factorial perspectives (e.g., Sharpe 2002). However, these have not yet been translated into systematic and widely accepted treatment approaches.

Why it is important to do this review

Previous reviews of relevant literature have argued mainly for cognitive‐behavioural approaches. However, many of these reviews (e.g., Pallesen 2005) have been based on uncontrolled and non‐randomised studies. Other reviews (e.g., Toneatto 2003) have focused on randomised trials, but were restricted by the small number of studies available at the time. A recent summary by Westphal 2008 noted that while some forms of CBT might be possibly efficacious, no psychological therapies could be classified as ‘effective’ according to standards described by Chambless 1998. That is, no therapies had been found superior to control conditions in at least two randomised trials across independent research settings. However, this literature continues to evolve and there is need for an updated review of best evidence. The current review supersedes a previous Cochrane review of interventions for pathological gambling (see Other published versions of this review) which is out of date and was withdrawn. This prior review had considered both psychological and pharmacological therapies for pathological gambling. The current review will focus on current evidence for psychological therapies, while pharmacological therapies are the focus of a separate review (Anderson 2011).

Objectives

The objectives of this review were to:

1. examine the efficacy of psychological therapies for pathological and problem gambling (CBT, motivational interviewing therapy, integrative therapy, other psychological therapy) and the durability of treatment effects, in comparison with control conditions;

2. examine the clinical characteristics of studies that may influence the relative efficacy of interventions; and

3. critically evaluate methodological features of studies that may bias research findings.

The current review focused on brief, moderate and intensive psychological therapies for pathological and problem gambling, that were delivered in face‐to‐face treatment sessions. Minimal interventions, involving less than 30 minutes of therapist contact (Babor 1994), as well as self help programmes (e.g., Gamblers Anonymous), were not considered. Remotely delivered therapies (e.g., via telephone or internet) were also out of scope. Literature on these therapies for gambling is relatively new (Gainsbury 2011) and there are unresolved concerns about these types of interventions (e.g., pertaining to patient safety; see Proudfoot 2004). Given that it also remains unclear whether or when such remote interventions are appropriate, given the typically complex nature of pathological and problem gambling (which is, for example, characterised by high levels of mental health comorbidity; Lorains 2011), these therapies require separate consideration.

Methods

Criteria for considering studies for this review

Types of studies

Eligible studies were randomised controlled trials. Cross‐over trials were eligible if participants were randomly allocated to treatment sequence (using the between‐group comparison from the first treatment stage only). Quasi‐randomised trials, using an alternative form of allocation to groups (e.g., sequential allocation), were not eligible. Sample size and language of the report were not used to determine inclusion, and no restrictions were placed on the study settings that were eligible for inclusion.

Types of participants

Participants were males and females of any age and ethnicity who were identified as pathological or problem gamblers using accepted assessment strategies. These included general clinical interviews (based on DSM criteria; APA 2000) and structured clinical interviews (like the Structured Interview for Pathological Gambling (SCI‐PG); Grant 2004) that provide formal diagnoses of pathological gambling. Also included were studies that used self report assessment tools, like the South Oaks Gambling Screen (SOGS; Lesieur 1987) and the Problem Gambling Severity Index (PGSI) of the Canadian Problem Gabmling Index (CPGI; Ferris 2001). These instruments provide cut‐off criteria for classifying the severity of gambling problems, frequently described in terms of problem gambling or ‘probable’ pathological gambling. For discussion of the distinctions between classification of pathological and problem gambling see Description of the condition. Also see Types of outcome measures for a longer listing of measurement tools typically used in the literature, and reviews by Abbott 2006 and PGRTC 2011 for discussion of diagnostic and measurement issues.

Types of interventions

Psychological therapies

This review considered any type of psychological therapy intended to reduce pathological and problem gambling. For current purposes, it was required that these interventions were classified as brief, moderate or intensive therapies (see Description of the intervention), that were delivered face‐to‐face by trained clinicians or student therapists under supervision. Several main categories of psychological therapy were identified a priori.

1) Cognitive‐behaviour therapy (CBT)

CBT included any intervention that was based on a cognitive or behavioural account of the condition, or a combination of the two. In the current update of the review, both individual and group therapy formats were considered together. Stand‐alone cognitive therapy and behaviour therapy interventions, as well as interventions adapting both cognitive and behavioural techniques, were also included under the single category. Future updates may consider separately the cognitive, behavioural and cognitive‐behavioural interventions, as more studies become available.

2) Motivational interviewing therapy

Motivational interviewing therapies referred to any treatment that was based predominantly on a motivational interviewing approach to change (Miller 1983). Motivational interviewing therapies are often brief and can be delivered as a freestanding treatment. An adaptation of motivational interviewing is called motivational enhancement therapy (MET), which combines motivational interviewing with a standardised assessment of problematic behaviours and personal feedback on results (Hettema 2005).

3) Integrative therapy

Integrative therapies included any interventions that combined significant aspects of different types of treatments (such as CBT and motivational interviewing). Where combined therapies were apparent, we made initial efforts to classify the therapy as predominantly one type of treatment (where around 80% of sessions were dedicated to one component of treatment). Where it was not possible to classify as one predominant type of treatment, we classified the intervention as an integrative therapy.

4) Other psychological therapy

This category included any other psychological therapies that did not fall under the primary categories noted above. As additional studies become available, future updates of this review will endeavour to consider specific interventions as independent categories. 

Interventions that were not considered

Interventions that were based on psychological principles but lacked systematic and face‐to‐face contact with a trained clinician were not considered. These were therapies delivered remotely (via telephone or internet; e.g., Carlbring 2008), as well as minimal interventions and self help programmes, including those using self completed workbooks and peer support programmes. Studies in which a psychological therapy was administered in combination with a pharmacological treatment were also beyond the scope of the review and were not considered.

Control conditions

The main types of control conditions considered in this review were:

1) No treatment

‘No treatment’ conditions include wait‐list control and assessment only. In typical applications of wait‐list control, participants undergo a baseline assessment but do not begin treatment until the experimental group receive and complete therapy. The wait‐list control participants receive another assessment prior to their own entry to active treatment. Assessment‐only conditions involve collection of outcome data at baseline and follow‐up.

2) Gamblers Anonymous referral

Gamblers Anonymous is the parallel organisation to Alcoholics Anonymous (AA), and is a voluntary self help fellowship that employs abstinent gamblers as peer‐counsellors (see Petry 2005a). In the context of therapy for pathological and problem gambling, studies may treat Gamblers Anonymous referral as a control condition. In such instances, actual attendance of Gamblers Anonymous sessions is not required and is typically unreported or low. Gamblers Anonymous referral can be described as a 'real world' (Petry 2006; p. 555) control condition that accounts for naturalistic recovery over time.

3) Non‐specific treatment component controls

Non‐specific treatment controls refer to any therapy condition that provides some general characteristics that are common to different psychological therapies. These include therapist attention, characteristics of therapist or group interaction (such as warmth, empathy and social support), and a treatment rationale that facilitates patient outcome expectancies (Mohr 2009).

Although these main classes of control may produce different estimates of effects of therapy (for example, 'no treatment' conditions like wait‐list control may reduce participants' expectations of change and are least likely to positively affect outcomes, leading to large estimates of therapy effects; Mohr 2009), we considered them together for purposes of this review. Other control conditions (e.g., pill placebos, in comparative studies evaluating psychological and pharmacological treatment conditions) may be considered in future updates as studies become available.

Control conditions that were not considered

The aim of this review was not to consider the comparative superiority of interventions. Therefore, comparisons between different psychological therapies classified under the types of interventions described above (e.g., CBT versus motivational interviewing therapies) were out of scope. Treatment‐as‐usual controls, which can comprise current best practice interventions or routine clinical care (Mohr 2009) were also out of scope. This was because treatment‐as‐usual also likely involves significant components of the main therapies considered here (e.g., cognitive and behavioural interventions). This represents a change from the published protocol associated with this review (see Differences between protocol and review). Comparisons between a psychological therapy and a pharmacological intervention were also not considered.

Types of outcome measures

We considered three primary and three secondary outcome measures and these correspond mainly to variables emphasised by experts in gambling research as reflected in the Banff consensus (Walker 2006). These variables include self reported gambling behaviours, measures of multiple symptoms of pathological and problem gambling, and criteria for diagnosis of pathological gambling. In order to reduce the number of potential analyses, we recorded additional outcomes (e.g., gambling cognitions; collateral reports of gambling behaviour) but did not systematically examine them. These outcomes may be considered in future updates as studies become available.

Primary outcomes

1) Reduction in gambling symptom severity, ascertained using standardised measurement instruments completed by participants or clinicians. These included the South Oaks Gambling Screen (SOGS; Lesieur 1987), the Massachusetts Gambling Screen (MAGS; Shaffer 1994), the Yale Brown Obsessive Compulsive Scale adapted for Pathological Gambling (PG‐YBOCS; Pallanti 2005), the Gambling Symptom Assessment Scale (G‐SAS; Kim 2009), the Problem Gambling Severity Index (PGSI) of the Canadian Problem Gambling Index (CPGI; Ferris 2001), the gambling subscale of the Addiction Severity Index (ASI‐G; Lesieur 1992), the National Opinion Research Centre DSM‐IV Screen (NODS; Gerstein 1999) and measures of DSM symptoms (e.g., DSM‐IV Criteria Interview; Cox 2004). When a study provided data on multiple measures, only one was used. We gave preference to scales that: (a) referenced symptoms over a period of time relevant for a clinical trial (e.g., past month); and (b) were most comparable to commonly used scales from other treatment outcomes studies (see Stinchfield 2004).

2) Reduction in financial loss from gambling, ascertained using instruments such as gambling diary records, timeline follow‐back interviews (Sobell 1992) or single‐item scales, measuring self reports of gambling behaviour (e.g., amount spent gambling or amount lost gambling in the past week or month).

3) Reduction in frequency of gambling, ascertained using instruments such as gambling diary records, timeline follow‐back interviews (Sobell 1992) or single‐item scales, measuring self reports of gambling behaviour (e.g., number of sessions gambled or the number of days spent gambling in the past month).

Secondary outcomes

1) Reduced occurrence of clinical diagnoses of pathological gambling, ascertained through clinician‐administered diagnostic interviews, based on DSM criteria (APA 2000).

2) Reduction in depression symptoms, ascertained using standardised and validated measurement instruments such as the Beck Depression Inventory II (Beck 1996) and the Hamilton Rating Scale for Depression (Hamilton 1960).

3) Reduction in anxiety symptoms, ascertained using standardised and validated measurement instruments such as the Spielberger State‐Trait Anxiety Inventory (Spielberger 1970) and the Hamilton Anxiety Scale (Hamilton 1959).

Search methods for identification of studies

We conducted a systematic search to identify studies that informed the current review, as well as a separate review on pharmacological interventions (Anderson 2011). This search comprised the following strategies:

Electronic searches

We searched the following electronic databases:

• The Cochrane Depression, Anxiety and Neurosis Review Group's Specialised Register (CCDANCTR, all years to 9 November 2011). See Appendix 1 for further details.

The author team carried out complementary searches for studies published between January 1980 and October 2011, on:

• The Cochrane Central Register of Controlled Trials (CENTRAL 2011, Issue 4);

• MEDLINE (1980 to 11 October 2011);

• EMBASE (1980 to 11 October 2011);

• PsycINFO (1980 to 11 October 2011);

• Literature in the Health Sciences in Latin America and the Caribbean (LILACS) to 11 October 2011.

For a detailed description of the search strategies see Appendix 2.

We also searched the WHO International Clinical Trials Registry Platform (ICTRP) and ClinicalTrials.gov (5 December 2011)

Searching other resources

We manually searched the following journals to identify potentially relevant studies that may not have been indexed in the included databases:

• Gambling Research (2003 onwards);

• International Gambling Studies (2001 to 2003);

• Journal of Gambling Issues (2000 to 2006).

We manually searched the reference lists of all included studies to identify other relevant references.

Data collection and analysis

Selection of studies

Studies were deemed eligible for inclusion if they:

1. used a systematic process of randomly allocating participants to groups;

2. included males and females of any age and ethnicity, who were identified as pathological or problem gamblers according to an accepted assessment strategy (see Types of participants); and

3. evaluated a psychological intervention to treat pathological or problem gambling, compared to a relevant control condition (see Types of interventions).

To determine eligibility of studies, several members of the review team independently scanned the title and abstract (where available) of each record retrieved. Two authors examined every record. We resolved disagreement about selection of studies through discussion. Agreement was reached in all cases. We retrieved full‐text articles where information in the title and abstract was insufficient to determine eligibility. We also retrieved the full‐text articles of all potentially eligible studies to confirm eligibility. Where information required to evaluate the inclusion criteria could not be obtained from published material, we contacted the authors of the primary study in efforts to obtain the information (see Dealing with missing data).

We identified duplicate publications and listed these along with the primary publication. We examined such publications for relevant information not reported in the primary report. Where multiple publications reported on the same outcome, we used only one set of results from the report with most complete or relevant data for each meta‐analysis.

Data extraction and management

We extracted data from studies using a structured data extraction form (Centre for Clinical Effectiveness 2010). Data included study descriptive characteristics (title, reference/source, year of publication), design features (e.g., setting, intervention, control, inclusion/exclusion criteria), sample characteristics (e.g., sample size, age, sex, predominant gambling activity), missing data strategies (e.g., loss to follow‐up, ‘intention‐to‐treat’ or ‘completers only’ analysis), results (e.g., means, standard deviations, frequency counts) and risk of bias (see below). Two review authors independently extracted information from the included studies.

We planned several main comparisons to evaluate the efficacy of each category of therapy relative to controls, as shown below:

1. CBT versus control;

2. motivational interviewing therapy versus control;

3. integrative therapy versus control; and

4. other psychological therapy versus control.

It was intended that comparisons would be conducted according to all primary and secondary outcomes. These comparisons were based on data relating to: (1) post‐treatment assessments, conducted from 0 to 3 months following completion of treatment (i.e., 0 to 3 months post‐treatment); and (b) follow‐up assessments, conducted from 9 to 12 months following completion of treatment (i.e., 9 to 12 months follow‐up). Future updates of this review may consider longer periods of post‐treatment follow‐up as studies become available. We combined studies comparing therapies with different types of control conditions (e.g., 'no treatment' and non‐specific treatment controls) in the same analyses.

Subgroup analyses (see Subgroup analysis and investigation of heterogeneity) and sensitivity analyses (Sensitivity analysis) were planned.

Assessment of risk of bias in included studies

Two review authors independently assessed the risk of bias associated with each included study. Both authors indicated 'High', 'Low' or 'Unclear' with regard to several design characteristics among the main sources of bias in clinical trials (Higgins 2008a; Westphal 2007):

Random allocation to groups (sequence generation): Only studies that indicated randomly allocating participants to groups were eligible for inclusion. However, the level of detail about the procedures varied, and it was not always possible to fully evaluate the method of allocation. We classified such studies with limited detail as 'Unclear’ and as having a potentially high risk of bias.

Allocation concealment: Proper randomisation depends on adequate concealment of allocation sequence (Schulz 2002), whereby participants and researchers are kept unaware, and are unable to foresee, the groups to which participants are allocated. We classified studies that lacked allocation concealment as having a high risk of bias.

Blinding of outcome assessors: Blinding can refer to hiding the randomisation sequence from several potential groups (e.g., participants, treatment providers, outcome assessors; Montori 2002). While blinding of participants and treatment providers is appropriate for studies of pharmacological treatments, these are less relevant to psychological therapies. For this review, blinding was thus used in reference to blinding of outcome assessors (e.g., researchers administering symptom scales). We classified studies that failed to blind outcome assessors (including studies relying on measures self completed by participants) as having a high risk of bias.

Analysis of ‘intention‐to‐treat’ data: Where data from participants were missing because of attrition, primary studies generally reported analyses conducted on either: (a) data from participants providing complete information (i.e., ‘completers only’); or (b) an intention‐to‐treat (ITT) sample, whereby data from all participants were included through use of various missing data strategies (e.g., last observation carried forward; Wood 2004). Given that attrition often reflects a non‐random process that varies across condition, results from analysis of ‘completers only’ data have a high risk of bias (Shih 2002) and we classified them as such in this review. There is also variability in risk of bias from crude (e.g., last observation carried forward) versus sophisticated (e.g., multilevel modelling) missing data techniques (Wood 2004). This additional variability was not considered in the current review (in efforts to reduce the complexity of proposed sensitivity analyses), but may be considered in future updates.

Selective outcome reporting: Selective outcome reporting refers to the selection and presentation of a limited subset of data or analyses, according to the nature (e.g., statistical significance) of the results (Hutton 2000). There are different types of selection bias (see Higgins 2008a), and these generally require access to study protocols to compare against published reports. As such, we classified studies as having a high risk of bias in the current review if: (1) they had study protocols available which listed outcomes or measures that were not reported in the results; or (2) outcomes were reported with inadequate detail for inclusion in the meta‐analyses.

Other sources of bias: We also documented two other sources of potential bias. These related to management of systematic pre‐treatment differences between groups (which may suggest a failure of randomisation to equate groups), and the measurement of treatment fidelity (which indicates whether therapy was delivered as intended). We classified studies as having a high risk of bias if they: (a) identified pre‐treatment differences and failed to control for these (we classified studies that did not evaluate differences, or did not indicate whether or not differences were accommodated in the data analysis, as 'Unclear'); and/or (b) failed to conduct some measurement of therapist adherence to treatment (we classified studies that did not indicate whether or not measurements of therapist adherence were obtained as 'Unclear').

We resolved disagreements between review authors with regard to the classification of studies through discussion. Agreement was reached in all cases. In line with recommendations (e.g., Juni 1999), we considered each source of bias independently. Where sufficient studies were available, these classifications were used to inform sensitivity analyses (see Assessment of heterogeneity).

Measures of treatment effect

Where skewed data were apparent, trials could use the mean to represent central tendency, even though long tails in a distribution can distort this statistic. Other studies may conduct a logarithmic transformation of data, or use the median to represent central tendency. Where transformations have been applied, the descriptive statistics may be reported in their original metric or geometric form. Results based on untransformed and log‐transformed data should not be included in a single meta‐analysis, while estimates of the median also should not be used (see Deeks 2008). In the current review, we extracted available information on the strategies used to address the skewed data from primary studies. It was envisaged that should studies consistently report the transformed (geometric) means, these would be used in the analysis. However, should un‐transformed statistics comprise the only information consistently available, these were assumed to be the best available estimate of central tendency. Where studies did not report examination of distribution of scores, the current review screened for asymmetrical outcome data using the check proposed by Altman 1996. Where data become available in future updates, sensitivity analysis and data transformations (Higgins 2008b) may be conducted.

Unit of analysis issues

Studies with multiple treatment groups

Where a study reported comparisons involving multiple treatment arms (e.g., individual CBT; group CBT) versus control, we combined data from the different therapies using the formulae reported by Higgins 2008c. Where different groups (e.g., men and women) were involved in the same treatment, but had reported data separately, we also combined these data.

Studies using cluster‐randomised designs

Where a cluster‐randomised trial (see Higgins 2008c) was identified, it was determined that the methods used to analyse data would be extracted, while the inflated standard error approach (Higgins 2008c) would be used to adjust standard errors for non‐independence of observations. To facilitate this adjustment, we extracted the degree of non‐independence, as reflected in the intra‐class correlation (ICC). Where the ICC was not reported, we assumed a value of 0.05.

Dealing with missing data

Missing information about study design and results/statistics

Information about research design that was not reported in a primary publication was initially ascertained through examination of duplicate publications. Where informative duplicate publications were unavailable, and where missing data related to the inclusion criteria or risk of bias (as defined in this review), we made contact with the study authors. This occurred in 10 cases. We also contacted the authors where necessary statistics (e.g., standard deviations) were missing. This occurred in 14 cases. We also made subsequent efforts to obtain missing statistics from other published results (e.g., standard errors, t‐values), based on formulae provided by Lipsey 2001. Alternatively, we obtained approximate estimates of certain statistics (e.g., means) from published figures. As a last resort, the study was excluded from the analysis.

Missing observations from primary studies due to attrition

The decision to consider 'completers only' or ITT data was initially determined by the type of information reported; for example, if the study only reported analyses of the ‘completers only’ sample. However, preference was given to data from the ITT sample where available. It was proposed that the influence of using completers only or ITT samples would inform sensitivity analyses (see Assessment of heterogeneity).

Assessment of heterogeneity

Clinical heterogeneity

The decision to synthesise results across studies was based on the clinical homogeneity of trials. We only considered studies or inclusion in a meta‐analysis if they: (a) evaluated samples of treatment seeking pathological or problem gamblers; (b) involved treatment using a similar type of psychological intervention; (c) compared the intervention with a relevant control condition; (d) measured treatment efficacy in terms of similar outcome measures; and (e) conducted an evaluation at a follow‐up time that was broadly comparable across investigations. For studies that were clinically heterogeneous or presented insufficient information to facilitate quantitative synthesis, we presented a narrative summary of results.

Statistical heterogeneity

We assessed statistical heterogeneity across studies using the I² statistic, which indicates the percentage of total variability across studies due to between‐study differences (for further discussion, see Huedo‐Medina 2006). We also examined the Chi² statistic and associated significance test (P value). However, given that this statistic lacks power to detect true differences (Deeks 2008), greater emphasis was placed on I². Although thresholds for I² are arbitrary, there are overlapping bands that can be interpreted as suggesting minor (0% to 40%), moderate (30% to 60%), substantial (50% to 90%) and considerable (75% to 100%) levels of heterogeneity (Deeks 2008). Interpretation of the I² statistic was also qualified through evaluation of the pattern of variability, and whether all studies indicated beneficial effects of treatment, or whether effects ranged, such that some studies suggested null or even harmful effects. Where there was strong evidence of true heterogeneity, the pooled effect was considered to a limited, although 'best available' estimate of the expected magnitude of the treatment effect. We qualified the interpretation through discussion of diversity.

Assessment of reporting biases

We examined multiple databases to identify published research, while we searched trial registers to identify unpublished studies. We considered funnel plots to evaluate publication bias, but there were an insufficient number of studies (Sterne 2008) to allow for meaningful presentation. Funnel plots will be presented in future updates. We also examined relevant databases and trial registers to identify reports published in a non‐English language. For example, we searched the LILACS database to examine for Portuguese‐ and Spanish‐language reports of trials, while we also examined trial registers from countries including Germany (e.g., German Clinical Trials Register) and China (e.g., Chinese Clinical Trial Register) through the ICTRP (see Search methods for identification of studies). 

Data synthesis

Two review authors entered data into Review Manager 5.1 (RevMan 2011). We conducted random‐effects meta‐analyses to provide a weighted estimate of the efficacy of each intervention compared to control. This random‐effects model assumed true variability in effect sizes across studies, and estimated both the average effect and the degree of variability across studies (from a distribution with a mean and variance; see Normand 1999).

Subgroup analysis and investigation of heterogeneity

Where statistical heterogeneity was observed, and where sufficient studies were available, we conducted subgroup analyses to examine factors explaining between‐study variability. These analyses evaluated potential differences in treatment effects according to study clinical characteristics:

1. intensity of therapy (brief, moderate, intensive therapy);

2. severity of gambling problems (e.g., pathological versus problem gambling); and

3. modality of therapy (individual versus group therapy).

We calculated estimates of treatment effect and precision for each subgroup when at least 10 studies were available. We compared these estimates across subgroups, whereby we assumed significant differences when the confidence intervals did not overlap (Hunter 2004). When additional studies become available, further subgroup analyses will be considered in subsequent editions of this review.

Sensitivity analysis

We conducted sensitivity analyses to examine methodological characteristics associated with risk of bias. The characteristics considered were specified in Assessment of risk of bias in included studies. These were:

1. random allocation to groups;

2. allocation concealment;

3. blinding of outcome assessors;

4. analysis of 'intention‐to‐treat' data.

For the purposes of the sensitivity analyses, we grouped studies according to risk of bias. We calculated estimates of treatment effect and precision for studies with low and high risk of bias and compared across subgroups. We made comparisons when at least 10 studies were available. To minimise the number of proposed analyses, we did not consider other sources of potential bias (relating to selective outcome reporting, evidence and management of pre‐treatment differences, and the measurement of treatment fidelity) in sensitivity analyses in the current update of the review.

'Summary of findings' tables

The main findings of the review are presented in the 'Summary of findings' tables. These tables include interpretations of meta‐analysis findings based on comparison with controls, conducted at 0 to 3 months post‐treatment: for CBT (see Table 1), motivational interviewing therapy (see Table 2) and integrative therapy (see Table 3). Given that a single study (with few participants) considered an other psychology therapy, while there were also limited data on outcomes at 9 to 12 months follow‐up, we have excluded 'Summary of findings' tables for these comparisons. They will be considered in future updates as more studies become available.

The magnitude of SMD point estimates are interpreted using conventions proposed by Cohen 1992. Also included are evaluations of quality of evidence. These evaluations reflect the confidence held that an effect size point‐estimate is correct and unbiased (Guyatt 2008). The Grading of Recommendation, Assessment, Development and Evaluation (GRADE) Working Group (GRADE Working Group 2004) has provided a system for classifying an overall body of evidence according to such levels of confidence. It identifies four potential grades of evidence:

• high: additional research is unlikely to change confidence in the estimate of a treatment effect;

• moderate: additional research will impact on confidence in the estimate and may change this estimate;

• low: additional research is very likely to change the estimate; and

• very low: any estimate of a treatment effect is uncertain.

This classification is based on several characteristics (including the risk of bias) of studies that comprise a body of evidence, including study design, quality, consistency and directness. By way of summary, the studies in the current review were all randomised trials with generally good directness (i.e., the extent that participants, interventions and outcomes are generally relevant to the applied settings of interest) and we thus classified them initially as having a 'high' grade of evidence. However, the small number of studies informing several analyses, the variable risk of bias of many studies, the inconsistency across studies in some analyses (as reflected in unexplained between‐study heterogeneity) and wide confidence intervals meant that we downgraded this classification. The extent of the downgrade varied across therapies and analyses of the different outcomes. For CBT, the ratings of quality of evidence extended from very low (e.g., for gambling symptom severity) to low (e.g., for financial loss from gambling). The other classes of therapy suffered generally from a lack of studies and most analyses were based on evidence that was classified as very low quality.

Results

Description of studies

Results of the search

See Figure 1 for the PRISMA flow diagram of results of the systematic search.

1.

Study flow diagram.

Once duplicate citations were deleted, the search identified 3163 abstracts. This number included articles that were potentially relevant to the current review, and a separate review of pharmacological treatments for pathological and problem gambling (Anderson 2011). After deletion of irrelevant articles based on title and abstract, we retrieved 72 full‐text articles for further investigation. A total of 14 studies (reported in 16 articles) met the inclusion criteria for the current review. All were available in English language and were published in peer‐reviewed journals.

Included studies

See Characteristics of included studies

Design

All 14 studies were randomised controlled trials. All studies were randomised at the participant level and used a parallel‐group design. See Characteristics of excluded studies for studies that were described as randomised but were found to use an alternative method of allocation.

Participants

A total of 1245 participants across studies were allocated to groups. The average sample size was around 89, and ranged from 13 (Melville 2004a) to 231 (Petry 2006). All trials were conducted using adults (mean age approximately 44 years), although one study examined college students (Petry 2009). One study considered males only (Sylvain 1997), while another considered females only (Dowling 2007). The remaining 12 studies examined males and females, although four samples were predominantly male (Carlbring 2010; Ladouceur 2001; Ladouceur 2003; Petry 2009). Dowling 2007 focused on participants with a primary problem concerning electronic gaming machines (EGMs), while over 80% of the sample in Ladouceur 2001 reported a problem with EGMs.

Eleven of the 14 studies evaluated pathological gamblers diagnosed through a structured or semi‐structured interview (Carlbring 2010; Dowling 2007; Grant 2009; Petry 2006), or a general clinical interview (Ladouceur 2001; Ladouceur 2003; Marceaux 2011; Melville 2004a; Melville 2004b; Oei 2010; Sylvain 1997). One study identified ‘moderate risk’ and ‘problem’ gamblers using a low cut‐off score (three or more) on the Problem Gambling Severity Index (PGSI) (Diskin 2009). Two studies (Petry 2008; Petry 2009) administered the South Oaks Gambling Screen (SOGS) and used lower thresholds (scores of three or more) than is usually the case (scores of five or more identify probable pathological gambling). Around 60% of the participants in Petry 2008 were probable pathological gamblers, compared with around 38% in Petry 2009.  

Five of the 14 studies excluded participants with common comorbid disorders. Exclusions were based on the presence of substance use disorders (Carlbring 2010; Grant 2009), current depression (Carlbring 2010) and personality disorders (Oei 2010). Three studies excluded patients with bipolar disorder (Carlbring 2010; Ladouceur 2001; Ladouceur 2003), while one study (Oei 2010) excluded patients if manic symptoms accounted for their gambling behaviour. The remaining studies did not indicate excluding participants based on common comorbidities (with an occasional exception being a history of schizophrenia or psychotic episodes).

Setting

Studies were conducted in the U.S. (Grant 2009; Marceaux 2011; Melville 2004a; Melville 2004b; Petry 2006; Petry 2008; Petry 2009), Canada (Diskin 2009; Ladouceur 2001; Ladouceur 2003; Sylvain 1997), Australia (Dowling 2007; Oei 2010) and Sweden (Carlbring 2010). Nine of the 14 studies were conducted in outpatient settings (Carlbring 2010; Diskin 2009; Dowling 2007; Grant 2009; Ladouceur 2001; Ladouceur 2003; Melville 2004a; Melville 2004b; Petry 2006). These included addiction or mental health treatment centres (Ladouceur 2001; Ladouceur 2003), university clinics (Diskin 2009; Dowling 2007), or more general community or research outpatient settings (Carlbring 2010; Grant 2009; Melville 2004a; Melville 2004b; Petry 2006). The remaining studies did not fully describe the context of the trial (Marceaux 2011; Oei 2010; Petry 2008; Petry 2009; Sylvain 1997).

Interventions

The interventions were administered by psychologists (Dowling 2007; Oei 2010; Sylvain 1997), cognitive therapists (Ladouceur 2001; Ladouceur 2003), masters level counsellors (Marceaux 2011; Melville 2004a; Melville 2004b), or therapists with bachelors, masters or doctoral training (Grant 2009; Petry 2006; Petry 2008; Petry 2009). Some studies relied on mixed groups of clinical professionals (Carlbring 2010), or used doctoral students under supervision (Diskin 2009) to deliver therapy.

Cognitive‐behaviour therapy (CBT)

Eleven of the 14 studies assessed a form of CBT. Six of the 14 studies assessed CBT that was individually administered (Dowling 2007; Grant 2009; Ladouceur 2001; Oei 2010; Petry 2006; Sylvain 1997), while seven studies assessed a form of group CBT (Carlbring 2010; Dowling 2007; Ladouceur 2003; Marceaux 2011; Melville 2004a; Melville 2004b; Oei 2010).

Nine of the 11 studies evaluated therapies that involved cognitive and behavioural components. Carlbring 2010 evaluated group therapy (Ortiz 2006) that focused on cognitive restructuring, coping skills and identification of high‐risk situations. Sessions included imaginary exposure with response prevention. Dowling 2007 evaluated therapy delivered through either individual or group sessions. Therapy included cognitive restructuring (relating to misconceptions about randomness and other biases), financial limit setting and activity scheduling of leisure activities. Problem‐solving training, relapse prevention and imaginal desensitisation were also included. Grant 2009 described the therapy as imaginal desensitisation plus motivational interviewing; however, treatment included cognitive and behavioural strategies, while motivational interviewing comprised only half of one session. The therapy made use of audio recordings of gambling scenarios to teach management of urges. Oei 2010 evaluated therapy (Raylu 2010) that was delivered through either individual or group sessions. Therapy included psycho‐education and motivational interviewing, cognitive correction and imaginal exposure. Problem‐solving training, instruction in cognitive and behavioural self management and relapse prevention were also included. Petry 2006 evaluated individual therapy (Petry 2005b) that involved mainly behavioural techniques with limited cognitive components. Sylvain 1997 delivered individual therapy that focused on cognitive misconceptions about randomness, and delivered instruction about randomness and other biases. Further components included cognitive restructuring and problem‐solving and social skills training. Melville 2004a evaluated a group intervention (Melville 2000a) based on Sylvain 1997, but considered two methods of delivering this intervention. In one form (natural language), therapists used verbal counselling techniques to correct misconceptions. In a second form (node‐link mapping), visual representation techniques (i.e., ‘mapping’) were used additional to verbal communication. Melville 2004b and Marceaux 2011 also evaluated this ‘node‐link mapping enhanced’ group CBT.

Two of the seven studies considered interventions that were primarily cognitive in nature. Ladouceur 2001 delivered individual therapy that focused on the correction of erroneous cognitions and beliefs about randomness, and also included a cognitive approach to relapse prevention. It was explicitly required that no behavioural techniques were used. Ladouceur 2003 evaluated a group therapy intervention that was based on the same therapy content.

Motivational interviewing therapy

Four of the 14 studies assessed a motivational interviewing therapy (Carlbring 2010; Diskin 2009; Petry 2008; Petry 2009). Two trials considered motivational interviewing (or motivational interviewing counselling). Carlbring 2010 evaluated four sessions of motivational interviewing that focused on the exploration of positive and negative consequences of gambling, and mapping of reasons for gambling. Diskin 2009 delivered a single session of motivational interviewing counselling that involved techniques outlined by Miller 2002. Two trials (Petry 2008; Petry 2009) evaluated single sessions of motivational enhancement therapy (MET). This included motivational interviewing counselling and the provision of personalised feedback about gambling. Three of the studies (Diskin 2009; Petry 2008; Petry 2009) considered participants with a lower level of gambling severity than is typical of studies of pathological gamblers.

Integrative therapy

Two of the 14 studies (Petry 2008; Petry 2009) considered an integrative therapy. Both studies considered MET and condensed CBT, which involved a single session of motivational interviewing and personalised feedback about gambling (i.e., MET), and three optional sessions of condensed CBT. These latter sessions focused on the determination of triggers for gambling, and discussion of methods for coping with internal triggers. Sessions also included instruction on coping with triggers, assertiveness training, and were supplemented using handouts adapted from Petry 2005b. Both studies considered participants with a lower level of gambling severity than is typical of studies of pathological gamblers.

Other psychological therapy

One of the 14 studies (Marceaux 2011) evaluated an intervention called Twelve‐Step Facilitated Treatment that involved therapy content modelled after the 12 steps of Gamblers Anonymous (see Ferentzy 2009). The therapy (Melville 2000b) took place across two sessions per week for eight weeks, and was delivered to small groups of pathological gamblers by trained clinicians. All sessions had a specific agenda and format, and were comprised of specific recovery tasks and suggested reading material relating to objectives in cognitive, emotional, behavioural, social and spiritual domains. This manualised therapy was delivered by therapists and differs from typical and naturalistic forms of Gamblers Anonymous, the latter of which is self or peer directed, and frequently lacks formal and uniform structures for sessions. None of the other included studies considered interventions that were classified as ‘Other psychological approaches to treatment’.

Control conditions

Ten of the 14 studies (Carlbring 2010; Dowling 2007; Grant 2009; Ladouceur 2001; Ladouceur 2003; Marceaux 2011; Melville 2004a; Melville 2004b; Oei 2010; Sylvain 1997) compared the intervention with a wait‐list control. One study (Grant 2009) used wait‐list control and also provided Gamblers Anonymous referral during the wait‐list period. Two trials (Petry 2008; Petry 2009) compared against assessment only. One study (Diskin 2009) compared the intervention with a control interview that was also classified as assessment only.

Two of the 14 studies (Grant 2009; Petry 2006) compared the intervention with Gamblers Anonymous referral. As noted above, Grant 2009 used Gamblers Anonymous referral during the wait‐list period, and reported that around 74% of control participants attended at least one Gamblers Anonymous meeting. Participants attended around one session, on average, across the treatment period. Petry 2006 reported that participants in the Gamblers Anonymous condition attended around two sessions, on average, over the course of treatment. Only 60% of the participants attended at least one Gamblers Anonymous session by the 12‐month follow‐up.

Duration of therapy and timing of post‐treatment assessments and follow‐up

Twelve of the 14 studies reported treatment duration in terms of number of sessions, ranging from 1 to 20 sessions (Carlbring 2010; Diskin 2009; Dowling 2007; Grant 2009; Ladouceur 2003; Marceaux 2011; Melville 2004a; Melville 2004b; Oei 2010; Petry 2006; Petry 2008; Petry 2009). One study (Ladouceur 2001) indicated that treatment was of variable duration and was administered for a maximum of 20 sessions. Participants in this study received around 11 hours of treatment, on average. Another study (Sylvain 1997) described administering one or two sessions weekly, for a maximum of 30 hours of treatment. They provided participants with around 17 hours of treatment, on average.

Four of the 14 studies reported the collection of post‐treatment data that coincided with the end of treatment (Marceaux 2011; Melville 2004a; Melville 2004b; Petry 2006). Three studies (Dowling 2007; Oei 2010; Sylvain 1997) did not indicate the timing of the post‐treatment assessment, and it was assumed that this measurement also coincided with end of treatment. In contrast, two studies reported post‐treatment data that were collected around one week following treatment (Carlbring 2010; Grant 2009), while another (Diskin 2009) reported that data were collected after one month. Two trials (Ladouceur 2001; Ladouceur 2003) reported that post‐treatment data were collected around three to four months after baseline; which allowed time for completion of the intervention condition. Another two trials (Petry 2008; Petry 2009) reported that post‐treatment data were collected six weeks after baseline. These trials both evaluated multiple interventions of variable duration, and post‐treatment data were thus collected several weeks after the completion of certain interventions.

Studies conducted follow‐up assessments at three months (Diskin 2009), six months (Carlbring 2010; Diskin 2009; Dowling 2007; Grant 2009; Ladouceur 2001; Ladouceur 2003; Marceaux 2011; Melville 2004a; Melville 2004b; Oei 2010; Petry 2006; Sylvain 1997), nine months (Diskin 2009; Petry 2008; Petry 2009), 12 months (Carlbring 2010; Diskin 2009; Ladouceur 2001; Ladouceur 2003; Petry 2006; Sylvain 1997) and 24 months (Ladouceur 2003) after treatment. Due to the nature of the wait‐list control condition (where control participants eventually received treatment), several studies only provided data on the treatment groups at follow‐up (Carlbring 2010; Dowling 2007; Grant 2009; Ladouceur 2001; Ladouceur 2003; Marceaux 2011; Melville 2004a; Melville 2004b; Oei 2010; Sylvain 1997). Studies that compared treatment and controls at follow‐up used conditions characterised as Gamblers Anonymous referral (Petry 2006) and assessment only (Diskin 2009; Petry 2008; Petry 2009).

Outcome measures

Eleven of the 14 studies reported measures of gambling symptom severity. Some studies used more than one tool. Several trials used the number of DSM‐IV (Ladouceur 2001; Ladouceur 2003; Marceaux 2011; Melville 2004a; Melville 2004b) or DSM‐III‐R criteria (Sylvain 1997). Other studies used the SOGS (Diskin 2009; Ladouceur 2001; Petry 2006; Petry 2009; Sylvain 1997), PGSI (Diskin 2009), Addiction Severity Index (ASI‐G) (Petry 2006; Petry 2008; Petry 2009), National Opinion Research Centre DSM‐IV Screen (NODS) (Carlbring 2010), Yale Brown Obsessive Compulsive Scale adapted for Pathological Gambling (PG‐YBOCS) and the Gambling Symptom Assessment Scale (G‐SAS) (Grant 2009). Many of the measures originally referenced lifetime or past year gambling, and some studies adapted measures across shorter periods of time. Carlbring 2010 adapted the NODS to assess symptoms over the past month, while three studies (Petry 2006; Petry 2008; Petry 2009) considered a version of the SOGS that measured past month symptoms. Marceaux 2011 adapted the DSM‐IV criteria to assess across the past month. Grant 2009 used the PG‐YBOCS and the G‐SAS (which were both developed as treatment outcome measures) to assess change over one week. Diskin 2009 administered the PGSI and SOGS to measure past year gambling, but only administered these measures at the 12‐month follow‐up. The remaining studies (Ladouceur 2001; Ladouceur 2003; Melville 2004a; Melville 2004b; Sylvain 1997) did not indicate whether the time frames of instruments were adapted for use as an outcome measurement.

Thirteen of the 14 studies reported data on a financial loss from gambling. Nine studies measured the amount spent or lost gambling (Carlbring 2010; Diskin 2009; Dowling 2007; Ladouceur 2001; Ladouceur 2003; Marceaux 2011; Oei 2010; Petry 2006; Sylvain 1997). Two studies (Petry 2008; Petry 2009) measured net expenditure (not including amounts won and re‐invested), while two other studies (Melville 2004a; Melville 2004b) measured the amount spent and calculated the percentage of baseline amount spent gambling. Studies asked variously about losses per day, week or month.

Thirteen of the 14 studies reported data on frequency of gambling. Three studies (Dowling 2007; Marceaux 2011; Oei 2010) reported broadly on ‘frequency’ of gambling, while others reported specifically on number of gambling sessions (Ladouceur 2001; Ladouceur 2003; Sylvain 1997) or number of times gambled (Melville 2004a; Melville 2004b). The remaining studies reported number of days gambled (Carlbring 2010; Diskin 2009; Petry 2006; Petry 2008; Petry 2009). These studies measured behaviour using weekly records (Dowling 2007), timeline follow‐back techniques (Carlbring 2010; Diskin 2009; Marceaux 2011; Petry 2006; Petry 2009) or ordinal scales (Oei 2010). One study (Petry 2008) used items from the ASI‐G. Other studies were unclear about the instruments used (Ladouceur 2001; Ladouceur 2003; Melville 2004a; Melville 2004b; Sylvain 1997). Studies asked variously about frequency per day, week or month.

Four of the 14 studies (Dowling 2007; Ladouceur 2001; Ladouceur 2003; Marceaux 2011) measured the rate of occurrence of clinical diagnoses of pathological gambling, ascertained by clinicians according to the DSM‐IV criteria at post‐treatment or follow‐up. Six of the 14 studies reported data on depression symptoms, using the Beck Depression Inventory‐II (Carlbring 2010; Dowling 2007; Marceaux 2011; Melville 2004b), the Hamilton Rating Scale for Depression (Grant 2009) and the Depression Anxiety Stress Scale‐21 (Oei 2010). Six of the 14 studies reported data on anxiety symptoms, using the Beck Anxiety Inventory (Carlbring 2010; Marceaux 2011; Melville 2004b), the State‐Trait Anxiety Inventory (Dowling 2007), the Hamilton Anxiety Scale (Grant 2009) or the Depression Anxiety Stress Scale (Oei 2010). The total score of the Depression Anxiety Stress Scale was used to represent both depression and anxiety for this review.

Some of the included studies also considered a limited number of additional outcomes that were not considered in the current update of the review. These included measures of urges and perceptions of control over gambling, gambling cognitions and other measures of psychological functioning (e.g., general psychological distress and self esteem). See Characteristics of included studies for further detail. Three studies (Petry 2006; Petry 2008; Petry 2009) reported documentation of adverse events relating to study or treatment participation.

Skewed data

Seven of the 14 studies (Carlbring 2010; Diskin 2009; Grant 2009; Marceaux 2011; Melville 2004a; Melville 2004b; Oei 2010) did not report examination of the distribution of scores, but were often found to suffer skewed data according to Altman 1996. Skewed data seemed to particularly characterise financial loss from gambling, although frequency of gambling was also often non‐normally distributed. Studies that did identify asymmetrical data used various analytic strategies. Dowling 2007 noted a small number of extreme outliers on weekly measures of gambling behaviour and removed these data‐points from calculations of average weekly gambling. Three studies (Ladouceur 2001; Ladouceur 2003; Sylvain 1997) used non‐parametric analyses. Three studies (Petry 2006; Petry 2008; Petry 2009) conducted transformations of data and used transformed scores in the substantive analyses, while summary statistics were reported in terms of raw means and standard deviations for frequency of gambling, and the median and interquartile range for financial loss from gambling. Raw means and standard deviations were the most common statistics reported across studies, and were used to inform the analyses in the current update of the review. 

Excluded studies

See Characteristics of excluded studies

Twenty‐three studies were identified as being potentially eligible for this review, but were excluded after closer examination. Four of these studies were excluded because they were quasi‐randomised controlled trials that used alternate methods of allocation to groups (Echeburua 1996; Echeburua 2000; Echeburua 2011; Myrseth 2009). Four studies delivered the intervention by telephone, internet or without any face‐to‐face interaction with a trained clinician (Carlbring 2008; Cunningham 2009; Hodgins 2001; Hodgins 2009). Ten studies compared CBT interventions with other CBT interventions, and did not have an appropriate control group for the purposes of this review (Blanco 2010; Harvey 2010; Hodgins 2007; McConaghy 1983; McConaghy 1988; McConaghy 1991; Milton 2002; Petry 2005c; Toneatto 2007b; Tse 2010). One study (Petry 2008b) compared three different types of psychological interventions, and did not have an appropriate control group for the purpose of this review. One study (Korman 2008) compared CBT with a treatment‐as‐usual condition that also comprised significant cognitive and behavioural components, and thus had no appropriate control group for the purposes of this review. Three studies were excluded because participants were not identified as pathological or problem gamblers using an accepted assessment strategy (Dickerson 1990; Petry 2006b; Sani 2005).

Studies awaiting classification

See Characteristics of studies awaiting classification.

There are four studies that are currently awaiting classification. These include studies that were identified through records in clinical trial registries, but without full text reports that could be identified or accessed and a study that was published and identified after the most recent update to the systematic search (Cunnigham 2012). These studies have not yet been formally evaluated for eligibility, and may be included or excluded in a future update to this review.

Risk of bias in included studies

See Figure 2 and Figure 3.

2.

'Risk of bias' summary: review authors' judgements about each risk of bias item for each included study.

3.

'Risk of bias' graph: review authors' judgements about each risk of bias item presented as percentages across all included studies.

Allocation

Randomisation

Nine of the 14 studies clearly described an appropriate method of randomisation (Carlbring 2010; Diskin 2009; Dowling 2007; Grant 2009; Marceaux 2011; Oei 2010; Petry 2006; Petry 2008; Petry 2009). These included computer‐generated urn‐randomisation procedures (Diskin 2009; Grant 2009; Petry 2006), computer‐generated randomisation schedules (Dowling 2007; Oei 2010), true random number services (Carlbring 2010) and a table of random numbers (Marceaux 2011). Two studies (Petry 2008; Petry 2009) used sealed envelopes. The remaining studies (Ladouceur 2001; Ladouceur 2003; Melville 2004a; Melville 2004b; Sylvain 1997) did not fully describe the randomisation procedure and we classified them as unclear and having a high risk of bias.

Allocation concealment

Four of the 14 studies described an appropriate method of allocation concealment (Carlbring 2010; Dowling 2007; Petry 2008; Petry 2009). In such instances, allocation was conducted by teams that were independent of the primary investigators (Carlbring 2010; Dowling 2007) and/or by using sealed envelopes (Carlbring 2010; Dowling 2007; Petry 2008; Petry 2009). Two studies indicated no attempt to conceal allocation (Ladouceur 2001; Ladouceur 2003), while one study concealed allocation from the participants but not members of the research team (Marceaux 2011). We classified these studies as having a high risk of bias for the purpose of this review. Seven studies did not indicate whether or not any process was used to conceal allocation and we classified them as having an unclear risk of bias (Diskin 2009; Grant 2009; Melville 2004a; Melville 2004b; Oei 2010; Petry 2006; Sylvain 1997).

Blinding

Blinding of outcome assessors

Blinding of the outcome assessors was clearly indicated in seven of the 14 studies (Diskin 2009; Grant 2009; Marceaux 2011; Melville 2004a; Melville 2004b; Petry 2008; Petry 2009). This type of blinding was typically possible because outcome assessors did not deliver the interventions. Three studies did not blind outcome assessors or relied on forms that were self completed by participants and we classified them as having a high risk of bias (Dowling 2007; Ladouceur 2001; Ladouceur 2003). Four studies did not indicate whether outcome assessors were blinded and we classified them as having an unclear risk of bias (Carlbring 2010; Oei 2010; Petry 2006; Sylvain 1997).

Incomplete outcome data

Management of attrition

Eight of the 14 studies used a missing data strategy that facilitated ITT analysis (Carlbring 2010; Diskin 2009; Dowling 2007; Grant 2009; Oei 2010; Petry 2006; Petry 2008; Petry 2009). These included last observation carried forward (Dowling 2007; Grant 2009; Oei 2010) and multilevel modelling procedures (that used all available data). These latter procedures were described as hierarchical linear modelling (Petry 2006; Petry 2008; Petry 2009), mixed‐effects modelling (Carlbring 2010) and linear mixed modelling (Diskin 2009). ‘Completers only’ data were used by four studies (Ladouceur 2001; Ladouceur 2003; Marceaux 2011; Sylvain 1997). Two studies did not clearly indicate the procedures used for the management of attrition and we classified them as having an unclear risk of bias (Melville 2004a; Melville 2004b).

Selective reporting

We classified 11 of the 14 studies as having an unclear risk of bias with regards to selective outcome reporting (Diskin 2009; Dowling 2007; Grant 2009; Ladouceur 2001; Ladouceur 2003; Marceaux 2011; Oei 2010; Petry 2006; Petry 2008; Petry 2009; Sylvain 1997) because no protocol papers were available. As such, it could not be ascertained whether data were reported on all expected outcomes. However, three studies reported inadequate or incomplete results on certain outcomes and could not be included in the meta‐analysis (Carlbring 2010; Melville 2004a; Melville 2004b). We classified these studies as having a high risk of bias.

Other potential sources of bias

Pre‐treatment differences between groups: 10 of the 14 studies reported examination of pre‐treatment differences between groups and reported no evidence of systematic variability according to group (Carlbring 2010; Diskin 2009; Dowling 2007; Grant 2009; Ladouceur 2001; Ladouceur 2003; Marceaux 2011; Petry 2008; Petry 2009; Sylvain 1997). Two of the 14 studies also examined pre‐treatment differences and found systematic variability according to gambling severity (Petry 2006), income (Oei 2010) and employment status (Oei 2010). Petry 2006 took initial scores of gambling severity into account in subsequent analyses (and was classified as having a low risk of bias), while Oei 2010 did not report whether or not this was done (and we thus classified it as having an unclear risk of bias). Two studies did not report examination of pre‐treatment differences and we classified them as having an unclear risk of bias (Melville 2004a; Melville 2004b).

Measurement of treatment fidelity: 11 of the 14 studies reported use of some methodology to evaluate treatment fidelity (Carlbring 2010; Diskin 2009; Dowling 2007; Grant 2009; Ladouceur 2001; Ladouceur 2003; Oei 2010; Petry 2006; Petry 2008; Petry 2009; Sylvain 1997). These studies typically recorded therapy sessions (Carlbring 2010; Diskin 2009; Dowling 2007; Grant 2009; Ladouceur 2001; Ladouceur 2003; Oei 2010; Petry 2006; Petry 2008; Petry 2009; Sylvain 1997). Dowling 2007 used audiotapes to solicit feedback on therapy. The remaining studies scrutinised the materials to evaluate therapist competency and/or adherence to treatment according to a manual or standardised coding system. All of these studies reported moderate to high levels of therapist competency and adherence to treatment. Three of the 14 studies reported no measure of treatment fidelity and were classified as having an unclear risk of bias (Marceaux 2011; Melville 2004a; Melville 2004b).

Effects of interventions

See: Table 1; Table 2; Table 3

Comparison 1: Cognitive‐behaviour therapy (CBT) versus control

Eleven studies out of 14 compared CBT with control at 0 to 3 months post‐treatment. Nine of the 11 studies provided data to compare groups on at least one outcome (Dowling 2007; Grant 2009; Ladouceur 2001; Ladouceur 2003; Marceaux 2011; Melville 2004a; Oei 2010; Petry 2006; Sylvain 1997). Seven studies compared groups on gambling symptom severity (Grant 2009; Ladouceur 2001; Ladouceur 2003; Marceaux 2011; Melville 2004a; Petry 2006; Sylvain 1997), while seven studies provided data to compare groups on financial loss (Dowling 2007; Ladouceur 2001; Ladouceur 2003; Marceaux 2011; Oei 2010; Petry 2006; Sylvain 1997) and frequency of gambling (Dowling 2007; Ladouceur 2001; Ladouceur 2003; Marceaux 2011; Oei 2010; Petry 2006; Sylvain 1997). Two studies provided data to compare groups on diagnoses of pathological gambling (Ladouceur 2003; Marceaux 2011), while four studies provided data to compare groups on depression symptoms (Dowling 2007; Grant 2009; Marceaux 2011; Oei 2010) and anxiety symptoms (Dowling 2007; Grant 2009; Marceaux 2011; Oei 2010). Two studies (Dowling 2007; Oei 2010) provided data on both individual and group CBT, and these data were combined. One study (Melville 2004a) provided data from two types of group CBT (group CBT and ‘mapping enhanced’ group CBT) and their data were also combined. Only one study (Petry 2006) compared groups at 9 to 12 months follow‐up.

Data from most comparisons in three of the 11 studies could not be included in the analyses because of missing data. Carlbring 2010 evaluated both CBT and motivational interviewing therapy in comparison with control. The published analyses combined data from different treatment conditions and compared ‘any treatment’ with control, and no data were provided on the individual therapy conditions. Two studies (Melville 2004a; Melville 2004b) reported the post‐treatment means for treatment and control conditions, but these data could not be used because they did not report standard deviations. In one instance (Melville 2004a) the required statistic was obtained from an independent groups t‐test for the ‘mapping enhanced’ CBT versus control comparison on gambling symptom severity. This standard deviation was used as the best estimate of the variance estimate for the CBT (combined group CBT and ‘mapping enhanced’ group CBT) versus control analysis on this outcome.

Primary outcomes

1) Gambling symptom severity

Seven of the 11 studies with 402 participants compared CBT with control on gambling symptom severity. Results indicated that the overall difference between groups at 0 to 3 months post‐treatment was significant and showed a beneficial effect of treatment (standardised mean difference (SMD) ‐1.82; 95% confidence interval (CI) ‐2.61 to ‐1.02; see Analysis 1.1). All individual studies were consistent in showing beneficial effects of therapy (i.e., no studies suggested adverse effects). However, the precise magnitude of the summary effect should be interpreted with caution given the high level statistical heterogeneity across studies (Chi² = 57.82, P < 0.01, I² = 90%). One study with 147 participants compared groups on gambling symptom severity at 9 to 12 months follow‐up. This trial found no significant difference (SMD ‐0.11; 95% CI ‐0.43 to 0.22; see Analysis 1.1). Because only one study was available it was not possible to estimate statistical heterogeneity.

1.1. Analysis.

Comparison 1 Cognitive‐behaviour therapy (CBT) versus control, Outcome 1 Gambling symptom severity.

2) Financial loss from gambling

Seven of the 11 studies with 505 participants compared CBT with control on financial loss from gambling. Results indicated that the difference in financial loss from gambling between groups at 0 to 3 months post‐treatment was significant and showed a beneficial effect of therapy (SMD ‐0.52; 95% CI ‐0.71 to ‐0.33; see Analysis 1.2). There was no evidence of statistical heterogeneity (Chi² = 3.90, P = 0.69, I² = 0%). One study with 147 participants compared groups at 9 to 12 months follow‐up and found no significant difference (SMD ‐0.15; 95% CI ‐0.47 to 0.18; see Analysis 1.2).

1.2. Analysis.

Comparison 1 Cognitive‐behaviour therapy (CBT) versus control, Outcome 2 Financial loss from gambling.

3) Frequency of gambling

Seven of the 11 studies with 505 participants compared CBT with control on frequency of gambling. Results indicated that the difference in gambling frequency between groups at 0 to 3 months post‐treatment was significant and showed a beneficial effect of therapy (SMD ‐0.78; 95% CI ‐1.11 to ‐0.45; see Analysis 1.3). There was moderate statistical heterogeneity (Chi² = 15.46, P = 0.02, I² = 61%). One study with 147 participants compared groups at 9 to 12 months follow‐up and found no significant difference (SMD ‐0.12; 95% CI ‐0.45 to 0.20; Analysis 1.3).

1.3. Analysis.

Comparison 1 Cognitive‐behaviour therapy (CBT) versus control, Outcome 3 Frequency of gambling.

Secondary outcomes

1) Pathological gambling diagnoses

Two of the seven studies with 81 participants compared CBT with control on the risk ratio of pathological gambling diagnoses at 0 to 3 months post‐treatment. Across studies it was found that around 8% of participants in the CBT condition met criteria for pathological gambling at post‐treatment, compared with around 81% in the control group. This difference was significant and demonstrated a beneficial effect of therapy (risk ratio (RR) 0.13; 95% CI 0.05 to 0.31; see Analysis 1.4). There was no evidence of heterogeneity (Chi² = 0.85, P = 0.36, I² = 0%).

1.4. Analysis.

Comparison 1 Cognitive‐behaviour therapy (CBT) versus control, Outcome 4 Diagnoses of pathological gambling.

2) Depression symptoms

Four of the 11 studies with 276 participants compared CBT with control on depression symptoms. Results indicated that the difference in depression symptoms between groups at 0 to 3 months post‐treatment was significant and showed a beneficial effect of therapy (SMD ‐0.66; 95% CI ‐0.93 to ‐0.39; see Analysis 1.5). There was no evidence of statistical heterogeneity (Chi² = 1.14, P = 0.77, I² = 0%). No studies compared groups at 9 to 12 months follow‐up.

1.5. Analysis.

Comparison 1 Cognitive‐behaviour therapy (CBT) versus control, Outcome 5 Depression symptoms.

3) Anxiety symptoms

Four of the 11 studies with 276 participants compared CBT with control on anxiety symptoms. Results indicated that the difference in anxiety symptoms between groups at 0 to 3 months post‐treatment was significant and showed a beneficial effect of therapy (SMD ‐0.64; 95% CI ‐0.90 to ‐0.37; see Analysis 1.6). There was no evidence of statistical heterogeneity (Chi² = 0.40, P = 0.94, I² = 0%). No studies compared groups at 9 to 12 months follow‐up.

1.6. Analysis.

Comparison 1 Cognitive‐behaviour therapy (CBT) versus control, Outcome 6 Anxiety symptoms.

Subgroup analyses

There were insufficient studies available to conduct subgroup analyses.

Sensitivity analyses

There were insufficient studies available to conduct sensitivity analyses.

Comparison 2: Motivational interviewing therapy versus control

Four of the 14 studies (Carlbring 2010; Diskin 2009; Petry 2008; Petry 2009) compared an motivational interviewing therapy with control at 0 to 3 months post‐treatment. Three of the four studies provided data for inclusion in the analyses. Two studies compared groups on gambling symptom severity (Petry 2008; Petry 2009); while three and two studies, respectively, compared groups on financial loss from gambling (Diskin 2009; Petry 2008; Petry 2009) and frequency of gambling (Diskin 2009; Petry 2009). No studies provided data on the secondary outcomes. Three studies provided data on group comparisons conducted at 9 to 12 months follow‐up.

One of the four studies (Carlbring 2010) compared groups at post‐treatment but could not be included in the analysis because of missing data. This study combined data from different therapy conditions and compared ‘any treatment’ with control. It did not report individual comparisons between motivational interviewing therapy and control.

Primary outcomes

1) Gambling symptom severity

Two of the four studies with 163 participants compared motivational interviewing therapy with control on gambling symptom severity. Results indicated that the difference between groups at 0 to 3 months post‐treatment was not significant and approached zero (SMD ‐0.03; 95% CI ‐0.55 to 0.50; see Analysis 2.1). There was moderate statistical heterogeneity across studies (Chi² = 2.76, P = 0.10, I² = 64%). Three trials with 221 participants compared groups at 9 to 12 months follow‐up. These studies found no significant differences (SMD ‐0.25; 95% CI ‐0.61 to 0.12; see Analysis 2.1). There was moderate statistical heterogeneity (Chi² = 3.66, P = 0.16, I² = 45%).

2.1. Analysis.

Comparison 2 Motivational interviewing versus control, Outcome 1 Gambling symptom severity.

2) Financial loss from gambling

Three of the four studies with 244 participants compared motivational interviewing therapy with control on financial loss from gambling. Results indicated that the difference between groups at 0 to 3 months post‐treatment was significant and showed a beneficial effect of therapy (SMD ‐0.41; 95% CI ‐0.75 to ‐0.07; see Analysis 2.2). There was moderate statistical heterogeneity (Chi² = 3.53, P = 0.17, I² = 43%). Two studies with 152 participants compared groups at 9 to 12 months follow‐up and found no significant differences (SMD ‐0.26; 95% CI ‐0.58 to 0.06; see Analysis 2.2). There was no statistical heterogeneity (Chi² = 0.41, P = 0.52, I² = 0%).

2.2. Analysis.

Comparison 2 Motivational interviewing versus control, Outcome 2 Financial loss from gambling.

3) Frequency of gambling

Two of the four studies with 145 participants compared motivational interviewing therapy with control on frequency of gambling. Results indicated that the difference between groups at 0 to 3 months post‐treatment was not significant (SMD ‐0.18; 95% CI ‐0.50 to 0.15; see Analysis 2.3). There was no statistical heterogeneity (Chi² = 0.55, P = 0.46, I² = 0%). One study with 62 participants compared groups on gambling frequency at 9 to 12 months follow‐up. Results showed a significant difference and beneficial effect of therapy (SMD ‐0.53; 95% CI ‐1.04 to ‐0.02; see Analysis 2.3).

2.3. Analysis.

Comparison 2 Motivational interviewing versus control, Outcome 3 Frequency of gambling.

Secondary outcomes

None of the studies reported data to compare motivational interviewing therapy with control on the secondary outcomes.

Subgroup analyses

There were insufficient studies available to conduct subgroup analyses.

Sensitivity analyses

There were insufficient studies available to conduct sensitivity analyses.

Comparison 3: Integrative therapy versus control

Two of the 14 studies (Petry 2008; Petry 2009) evaluated integrative therapies. These studies both considered motivational enhancement therapy (MET) and condensed CBT. Both studies compared groups on gambling symptom severity and financial loss from gambling, while one study compared groups on frequency of gambling (Petry 2009)

Primary outcomes

1) Gambling symptom severity

Two studies with 137 participants compared integrative therapy with control on gambling symptom severity. Results indicated that the difference between groups at 0 to 3 months post‐treatment was not significant (SMD ‐0.15; 95% CI ‐0.50 to 0.19; see Analysis 3.1). There was no statistical heterogeneity (Chi² = 0.15, P = 0.70, I² = 0%). Both studies with 128 participants also compared groups at 9 to 12 months follow‐up. Results showed a significant difference and beneficial effect of therapy (SMD ‐0.38; 95% CI ‐0.74 to ‐0.02; see Analysis 3.1). There was no statistical heterogeneity (Chi² = 0.00, P = 0.97, I² = 0%).

3.1. Analysis.

Comparison 3 Integrative therapy versus control, Outcome 1 Gambling symptom severity.

2) Financial loss from gambling

Two studies with 137 participants compared integrative therapy with control on financial loss from gambling. Results indicated that the difference between groups at 0 to 3 months post‐treatment was not significant (SMD ‐0.41; 95% CI ‐1.04 to 0.23; see Analysis 3.2). There was moderate statistical heterogeneity across studies (Chi² = 3.09, P = 0.08, I² = 68%). Both trials with 128 participants also compared groups at 9 to 12 months follow‐up. Results showed no significant difference (SMD ‐0.06; 95% CI ‐0.42 to 0.29; see Analysis 3.2). There was no evidence of heterogeneity (Chi² = 0.50, P = 0.48, I² = 0%).

3.2. Analysis.

Comparison 3 Integrative therapy versus control, Outcome 2 Financial loss from gambling.

3) Frequency of gambling

One of the two studies with 52 participants compared integrative therapy with control on frequency of gambling. Results indicated that the difference between groups at 0 to 3 months post‐treatment was not significant (SMD ‐0.12; 95% CI ‐0.69 to 0.46; see Analysis 3.3). This study with 52 participants also compared groups at 9 to 12 months follow‐up. Results showed no significant difference (SMD ‐0.45; 95% CI ‐1.02 to 0.13; see Analysis 3.3).

3.3. Analysis.

Comparison 3 Integrative therapy versus control, Outcome 3 Frequency of gambling.

Secondary outcomes

None of the studies reported data to compare integrative therapy with control on the secondary outcomes.

Subgroup analyses

There were insufficient studies available to conduct subgroup analyses.

Sensitivity analyses

There were insufficient studies available to conduct sensitivity analyses.

Comparison 4: Other psychological therapy versus control

One of the 14 studies (Marceaux 2011) evaluated an ‘Other psychological therapy’. This study evaluated Twelve‐Step Facilitated Group Therapy (Melville 2000b). It compared the therapy with control at 0 to 3 months post‐treatment on all primary and secondary outcomes. This study did not compare the groups at 9 to 12 months follow‐up.

Primary outcomes

1) Gambling symptom severity

One study with 18 participants compared an ‘Other psychological therapy’ with control on gambling symptom severity. Results indicated that the difference between groups at 0 to 3 months post‐treatment was significant and showed a beneficial effect of therapy (SMD ‐1.82; 95% CI ‐2.98 to ‐0.66; see Analysis 4.1). Because only one study was included, it was not possible to estimate statistical heterogeneity.  

4.1. Analysis.

Comparison 4 Other psychological therapy versus control, Outcome 1 Gambling symptom severity.

2) Financial loss from gambling

One study with 18 participants compared an ‘Other psychological therapy’ with control on financial loss from gambling. Results indicated that the difference between groups at 0 to 3 months post‐treatment was significant and showed a beneficial effect of therapy (SMD ‐1.38; 95% CI ‐2.46 to ‐0.30; see Analysis 4.2).

4.2. Analysis.

Comparison 4 Other psychological therapy versus control, Outcome 2 Financial loss from gambling.

3) Frequency of gambling

One study with 18 participants compared an ‘Other psychological therapy’ with control on frequency of gambling. Results indicated that the difference between groups at 0 to 3 months post‐treatment was significant and showed a beneficial effect of therapy (SMD ‐1.66; 95% CI ‐2.78 to ‐0.53; see Analysis 4.3).

4.3. Analysis.

Comparison 4 Other psychological therapy versus control, Outcome 3 Frequency of gambling.

Secondary outcomes

1) Pathological gambling diagnoses

One study with 18 participants compared an ‘Other psychological therapy’ with control on the occurrence of pathological gambling diagnoses at 0 to 3 months post‐treatment. This study found that 27.3% of participants in the therapy condition met criteria for pathological gambling at post‐treatment, compared with 85.7% in the control group. This difference was significant and demonstrated a beneficial effect of therapy (RR 0.32; 95% CI 0.12 to 0.87; see Analysis 4.4).  

4.4. Analysis.

Comparison 4 Other psychological therapy versus control, Outcome 4 Diagnoses of pathological gambling.

2) Depression symptoms

One study with 18 participants compared an ‘Other psychological therapy’ with control on depression symptoms. Results indicated that the difference between groups at 0 to 3 months post‐treatment was significant and showed a beneficial effect of therapy (SMD ‐1.55; 95% CI ‐2.66 to ‐0.44; see Analysis 4.5).

4.5. Analysis.

Comparison 4 Other psychological therapy versus control, Outcome 5 Depression symptoms.

3) Anxiety symptoms

One study with 18 participants compared an ‘Other psychological therapy’ with control on anxiety symptoms. Results indicated that the difference between groups at 0 to 3 months post‐treatment was not significant (SMD ‐0.72; 95% CI ‐1.70 to 0.26; see Analysis 4.6).

4.6. Analysis.

Comparison 4 Other psychological therapy versus control, Outcome 6 Anxiety symptoms.

Subgroup analyses

There were insufficient studies available to conduct subgroup analyses.

Sensitivity analyses

There were insufficient studies available to conduct sensitivity analyses.

Discussion

Summary of main results

Cognitive‐behaviour therapy (CBT)

The current review supports the short‐term efficacy of CBT. Relative to controls, participants randomised to CBT demonstrated reduced gambling symptom severity, financial loss from gambling and less frequent gambling at 0 to 3 months post‐treatment. CBT conditions also demonstrated reduced rates of diagnoses of pathological gambling and symptoms of depression and anxiety. The magnitude of treatment effects frequently varied across studies and current estimates of the magnitude of therapy effects should be evaluated with caution. This was particularly the case for gambling symptom severity, with some trials providing unrealistically large estimates of effects of therapy. Despite this between‐study heterogeneity, however, the available trials consistently suggested benefits from therapy immediately following treatment. In contrast, there was only one study that evaluated therapy effects across longer periods of time (e.g., 9 to 12 months post‐treatment), and the durability of effects from CBT for pathological and problem gambling remains unknown. 

There are two main caveats for interpreting the findings relating to CBT. First, there is variability in the nature of interventions classified as CBT. Several therapies included cognitive and behavioural techniques, while others evaluated pure forms of cognitive therapy. The effects of individual and group CBT were also combined. As such, there may be unexplored variability in effects associated with type of CBT and the modality of delivery. Future updates of this review will endeavour to explore this variability as more studies become available. Second, there are few trials available to support most specific types of manualised therapy. Chambless 1998 describes criteria for interventions to be classified as ‘empirically supported’, and emphasises the need for at least two randomised trials that are rigorously designed and conducted in independent research settings. Although Ladouceur 2001 and Ladouceur 2003 both evaluated the same form of cognitive therapy (that was based on Sylvain 1997, but minus any behavioural techniques), the same research team conducted the studies. The ‘mapping enhanced’ group CBT (Melville 2004a) has also received support in multiple trials, but these were also conducted by the same research team (while the studies were characterised by small samples and other methodological limitations). All other interventions have been supported by single trials, and no form of CBT can be classified as empirically supported at the current time.

Motivational interviewing therapy

Synthesis of three trials of motivational interviewing therapy provided evidence of some beneficial effects from therapy in terms of less financial loss from gambling at 0 to 3 months post‐treatment, and less frequent gambling at 9 to 12 months follow‐up. However, the effect was close to zero when defined by gambling symptom severity, while comparisons on other outcomes were associated with wide confidence intervals and were not significant. The findings may suggest that there are beneficial effects from motivational interviewing that are limited to gambling behaviour, but that do not extend to other symptoms of pathological and problem gambling. However, the small number of studies and wide confidence intervals associated with many comparisons highlight the imprecision of current evidence, and indicate the need for additional research.    

Two features of the studies evaluating motivational interviewing therapies deserve comment. First, in several cases the trials used low cut‐off criteria for participant eligibility, and yielded samples that were characterised by less severe difficulties overall, relative to studies of pathological gamblers. Accordingly, the results of such studies do not generalise to severe forms of the disorder. Second, the motivational interviewing therapies considered here all consisted of single sessions of around an hour, which contrasts with some usages of motivational interviewing that can comprise multiple sessions of up to 12 hours of therapist contact (Hettema 2005). Prior research has shown that ‘dosage’ of motivational interviewing can relate to the magnitude of treatment effects (Burke 2003), and it seems plausible that longer therapies may bring about greater change in pathological and problem gambling. Future updates of this review will endeavour to explore the role of severity of gambling and intensity of treatment as more studies become available.

Integrative therapies

Two studies provided evidence concerning MET and condensed CBT, and considered samples demonstrating, on average, less severe gambling problems than studies of pathological gamblers.  Although comparisons yielded generally consistent trends suggesting modest benefits from therapy, none were significant at 0 to 3 months post‐treatment and only one was significant at 9 to 12 months follow‐up. Given the small number of studies, it is not possible to ascertain whether non‐significant findings reflect limited effects of treatment or Type II errors (reflected in the wide confidence intervals and imprecise estimates of the treatment effect).

Other types of psychological therapies

One trial considered a therapy that was classified as an ‘Other psychological therapy’. The intervention was Twelve‐Step Facilitated Group Therapy (Melville 2000b). The study showed beneficial effects of this treatment on gambling symptom severity, financial loss from gambling, frequency of gambling and the rate of occurrence of diagnoses of pathological gambling and depression symptoms, when compared to wait‐list control at 0 to 3 months post‐treatment. As such, there is preliminary evidence to suggest that this therapy is a promising form of treatment and should be further considered in randomised trials.

Alternative types of psychological intervention, such as mindfulness‐based therapies (Korman 2008), psychodynamic therapies (Bergler 1957) and eye movement desensitisation/reprocessing (Henry 1996), were not considered in eligible trials, and rigorous data on their efficacy relative to control is currently lacking. 

Clinical heterogeneity of studies

There were high levels of clinical heterogeneity across the included studies. Therapies varied widely in intensity (extending in duration from 1 to 20 sessions) and included brief, moderate and intensive therapies. Studies also varied in the approaches used to assess gambling problems, and considered participants formally diagnosed as pathological gamblers (e.g., using clinical interviews based on the DSM‐IV), as well as samples of problem or ‘at risk’ gamblers, demonstrating varying (or sub‐clinical) levels of gambling‐related harm. There were too few studies to conduct subgroup analyses and provide systematic consideration of whether such factors are related to the expected magnitude of treatment effects, and such analyses will be included in future updates as more studies become available.  

Overall completeness and applicability of evidence

Completeness of evidence

There were 14 randomised trials included in this review that considered CBT (11 studies), motivational interviewing therapies (four studies), integrative therapies (two studies) and other types of therapy (one study). There was heterogeneity within categories and few specific forms of treatment had been addressed in multiple studies. The evidence mainly addressed the short‐term efficacy of therapy and few studies followed participants across longer periods of time. The treatment effects were typically defined in terms of outcomes that are important indicators of improvement in gambling problems, including gambling symptom severity and gambling behaviour (e.g., financial loss from gambling) and most studies provided data on at least one of these outcomes. In contrast, fewer studies provided data on secondary symptoms of depression and anxiety, or other outcomes (e.g., gambling cognitions). The small number of studies meant that subgroup and sensitivity analyses could not be conducted, such that understanding of the sources of heterogeneity across studies remains incomplete.

Applicability of evidence

The current evidence has moderate to high levels of external validity overall. The available studies typically used few exclusion criteria, and often considered males and females, as well as participants with various types of preferred gambling activity. Many of the studies included patients with significant mental health comorbidity, and current evidence may generalise well to individuals with co‐occurring conditions. The studies also delivered therapy across a range of outpatient settings (e.g., addiction or mental health treatment centres, general community settings), and the findings may also be applicable to a range of clinical contexts. Therapists had various types and levels of training, which suggests that findings can be applied to settings where therapists vary in clinical background and expertise.

There were no studies that considered therapy in the context of inpatient treatment. The evidence also has low to moderate international relevance, with most studies conducted in the U.S. (seven studies) and Canada (four studies), and a smaller number of trials in Australia (two studies) and Europe (one study). The nature of pathological and problem gambling, as well as the context of delivery of treatment services, may differ across jurisdictions and it remains unclear whether findings will generalise to other parts of the world. Finally, the available trials of both motivational interviewing and the integrative therapies were generally focused on participants demonstrating overall less severe levels of problem gambling, relative to studies of pathological gambling. As such, the evidence for these therapies may have limited applicability to patients with severe levels of a gambling disorder.

Quality of the evidence

The randomised trials identified in the current review varied in quality. Some provided limited or no description of the method of random allocation, and may have deviated from a rigorous protocol of random allocation. Several studies managed attrition through ‘completers only’ analysis, which may also overestimate the effects of gambling treatment. There was further variability in the use of allocation concealment and blinding of outcome assessors. Although there were too few studies to conduct sensitivity analyses, and thus provide a systematic evaluation of impacts of risk of bias, some primary studies provided very large and probably unrealistic treatment effects (particularly relating to CBT). This also suggests that the current evidence may provide overestimates of the true effects of therapies.

Potential biases in the review process

The current review was limited mainly to published research, and while efforts were made to identify unpublished material (e.g., through searches of trial registers), the efficacy of these strategies (as well as the quality of such unpublished studies) is uncertain. Conclusions may be biased if there are unpublished studies with discrepant findings. The small number of studies meant that many analyses yielded findings that were imprecise and suffered from low statistical power. Furthermore, some analytical techniques (e.g., funnel plots; subgroup analyses) could not be used. The findings from some studies could not be included in the analyses because of missing data. Although attempts were made to recover this information, this was not possible in all instances and such findings, if available, may have changed the results. Finally, several of the primary outcomes were found to suffer from skewed data, which may have biased the only estimates of central tendency (e.g., mean) that were commonly available for analysis.

Agreements and disagreements with other studies or reviews

The current findings are generally consistent with prior reviews of this treatment literature, that have argued for the efficacy of any psychological therapy for the treatment of pathological and problem gambling (Pallesen 2005), as well as specific cognitive‐behavioural interventions (Gooding 2009; Leibetseder 2011; Westphal 2008). However, the current review represents a more comprehensive and rigorous extension of existing work. Several prior reviews have not used systematic search procedures (e.g., Brewer 2008; Westphal 2008) and were not necessarily based on all available evidence. Other reviews (Gooding 2009; Leibetseder 2011; Pallesen 2005) used systematic search procedures, but also adopted less stringent inclusion criteria than in the current instance. Probably the most rigorous prior review was conducted by Toneatto 2003, which was also limited by the small number of relevant trials available at the time. Accordingly, the current review provides the strongest support presently available for the efficacy of CBT for pathological and problem gambling. The review also provides the best available synthesis of evidence relating to other types of psychological therapies, including motivational interviewing and integrative therapies.

Authors' conclusions

Implications for practice.

The current evidence supports cognitive‐behavioural approaches for the treatment of pathological and problem gambling, and is consistent with prior recommendations (e.g., Lopez Viets 1997; VGDHS 2000) that cognitive‐behaviour therapy (CBT) should be classified as best practice at the present time. However, the current review highlights important caveats to this advice. A substantial amount of the evidence comes from studies that suffered from multiple limitations, and these may have led to overestimates of treatment efficacy. Furthermore, the evidence only shows short‐term benefits from therapy, and there is insufficient evidence to indicate whether or not treatment effects observed soon after therapy are maintained across longer periods of time. The available evidence is suggestive of some potential benefits from alternative psychological therapies, including motivational interviewing therapies, integrative therapies and Twelve‐Step Facilitated Group Therapy, but there is a need for additional research before conclusions can be drawn. 

Implications for research.

Several of the psychological therapies considered in this review showed promise for the treatment of pathological or problem gambling, and there is a strong need for additional studies to further elucidate these potential benefits. Independent replication studies are required to evaluate manualised CBT interventions. Once a specific form of intervention has been studied sufficiently to meet standards for classification as an empirically supported treatment (see Chambless 1998), this intervention can serve as a benchmark against which other therapies can be compared. Future studies should also consider therapies for gambling problems that range in severity, as well as samples of mainly pathological gamblers. Intensive therapies for the treatment of severe forms of the disorder should be further considered. However, given the low rates of treatment seeking that typically characterise pathological and problem gambling (suggesting barriers to accessing existing therapies; Slutske 2006), further consideration of brief and moderate intensity interventions is also important.

The lack of credible long‐term studies means that the evidence base is presently inadequate concerning relapse rates and long‐term efficacy. A limited body of evidence (e.g., Echeburua 2001) suggests that even the best techniques may be associated with high rates of relapse, and there is a need for trials across longer periods of follow‐up. The use of a wait‐list control requires that group comparisons are limited to immediate post‐treatment, and alternative controls (such as non‐specific treatment controls) should be used to facilitate long‐term comparisons. Relative to 'no treatment' conditions, such non‐specific treatment controls should also provide more conservative tests of effects of therapy (e.g., by matching treatment and control conditions on positive outcome expectancies; Mohr 2009). Future trials should aim to recruit a sufficient sample to allow for stable estimates of treatment effects (e.g., a minimum of around 25 to 30 per treatment condition; Chambless 1998), and use an appropriate method of random allocation (which is clearly described in the published report). Although it may be necessary to deviate from a strictly random allocation procedure (e.g., if participants present with extreme levels of distress), data from these patients should be excluded from the analyses. Studies should include procedures that allow for allocation concealment and blinding of outcome assessors, and adopt analytic approaches that are consistent with intention‐to‐treat principles (e.g., multilevel modelling). Trials should report means and standard deviations for a range of outcome variables that are recommended to define effects of gambling treatment (see Walker 2006).

Notes

This review supersedes a previous review: Oakley‐Browne M, Adams P, Mobberley P. Interventions for pathological gambling. Cochrane Database of Systematic Reviews 2000, Issue 1. Art. No.: CD001521. DOI: 10.1002/14651858.CD001521.pub2.

Appendices

Appendix 1. CCDAN Specialised Register (CCDANCTR)

The Cochrane Depression, Anxiety and Neurosis Group (CCDAN) maintain two clinical trials registers at their editorial base in Bristol, UK, a references register and a studies‐based register. The CCDANCTR‐References Register contains over 30,000 reports of randomised controlled trials in depression, anxiety and neurosis. Approximately 65% of these reports (references) have been tagged to individual, coded trials. The coded trials are held in the CCDANCTR‐Studies Register and records are linked between the two registers through the use of unique Study ID tags. Coding of trials is based on the EU‐Psi coding manual. Please contact the CCDAN Trials Search Co‐ordinator for further details. Reports of trials for inclusion in the Group's registers are collated from routine (weekly), generic searches of MEDLINE (1950‐), EMBASE (1974‐) and PsycINFO (1967‐); quarterly searches of the Cochrane Central Register of Controlled Trials (CENTRAL) and review‐specific searches of additional databases. Reports of trials are also sourced from international trials registers c/o the World Health Organisation’s trials portal (ICTRP), drug companies, the handsearching of key journals, conference proceedings and other (non‐Cochrane) systematic reviews and meta‐analyses. Details of CCDAN's generic search strategies can be found on the Group's website.

CCDANCTR‐Studies Register

The studies register was searched using the following controlled (coded) phrase: 
 Condition = "Pathological Gambling"

CCDANCTR‐References Register

The references register was searched using the following free‐text terms to identify additional untagged/uncoded references: 
 (gambl* or betting or wagering or ludomania* or ludopath* or pokie* or ((gaming or fruit or slot) and machine*) or (video* and lotter*))

The CCDANCTR (studies and references) was up‐to‐date as of 9 November 2011.

Appendix 2. Additional search strategies

OVID MEDLINE (1980 to 11 October 2011)

1. Gambling/ 
 2. Video Games/px [psychology] 
 3. (gambling or gamble$ or betting or wagering).tw. 
 4. ((patholog$ or problem$ or addict$ or compulsiv$ or impulsive$ or crav$) adj7 gambl$).tw. 
 5. ludomania$.tw. 
 6. pokie$.tw. 
 7. ((gaming or fruit or slot) adj machine$).tw. 
 8. (video$ adj1 lotter$).tw. 
 9. or/1‐8 
 10. Randomized Controlled Trial.pt. 
 11. Controlled Clinical Trial.pt. 
 12. randomi#ed.ti,ab. 
 13. randomly.ab. 
 14. placebo$.ab. 
 15. drug therapy.fs. 
 16. trial.ti. 
 17. groups.ab. 
 18. (control$ adj3 (trial or study)).ab,ti. 
 19. ((singl$ or doubl$ or tripl$ or trebl$) adj3 (blind$ or mask$ or dummy)).mp. 
 20. (animals not (humans and animals)).sh. 
 21. or/10‐19 
 22. (21 not 20) 
 23. (9 and 22)

OVID EMBASE (1980 to 11 October 2011)

1. Pathological Gambling/ 
 2. (gambling or gamble$ or betting or wagering).tw. 
 3. ((patholog$ or problem$ or addict$ or compulsiv$ or impulsive$ or crav$) adj7 gambl$).tw. 
 4. ludomania$.tw. 
 5. pokie$.tw. 
 6. ((gaming or fruit or slot) adj machine$).tw. 
 7. (video$ adj1 lotter$).tw. 
 8. or/1‐7 
 9. Randomized Controlled Trial.de. 
 10. Randomization.de. 
 11. Placebo.de. 
 12. placebo$.ti,ab. 
 13. randomi#ed.ti,ab. 
 14. randomly.ab. 
 15. ((singl$ or doubl$ or trebl$ or tripl$) adj3 (blind$ or mask$ or dummy)).mp. 
 16. factorial$.ti,ab. 
 17. allocat$.ti,ab. 
 18. assign$.ti,ab. 
 19. volunteer$.ti,ab. 
 20. crossover procedure.de. 
 21. (crossover$ or cross over$).ti,ab. 
 22. (quasi adj (experimental or random$)).mp. 
 23. (control$ adj3 (trial$ or study or studies or group$)).ti,ab. 
 24. ((animal or nonhuman) not (human and (animal or nonhuman))).de. 
 25. or/9‐23 
 26. (25 not 24) 
 27. (8 and 26)

OVID PsycINFO (1980 to 11 October 2011)

1. exp Gambling/ 
 2. (gambling or gambler$ or betting or wagering).tw. 
 3. ((patholog$ or problem$ or addict$ or compulsiv$ or impulsive$ or crav$) adj7 gambl$).tw. 
 4. ludomania$.tw. 
 5. pokie$.tw. 
 6. ((gaming or fruit or slot) adj machine$).tw. 
 7. (video$ adj1 lotter$).tw. 
 8. or/1‐7 
 9. Treatment Effectiveness Evaluation.sh. 
 10. Clinical Trials.sh. 
 11. Mental Health Program Evaluation.sh. 
 12. Placebo.sh. 
 13. placebo$.ti,ab. 
 14. randomly.ab. 
 15. randomi#ed.ti,ab. 
 16. trial.ti,ab. 
 17. ((singl$ or doubl$ or trebl$ or tripl$) adj3 (blind$ or mask$ or dummy)).mp. 
 18. (control$ adj3 (trial$ or study or studies or group$)).ti,ab. 
 19. factorial$.ti,ab. 
 20. allocat$.ti,ab. 
 21. assign$.ti,ab. 
 22. volunteer$.ti,ab. 
 23. (crossover$ or cross over$).ti,ab. 
 24. (quasi adj (experimental or random$)).mp. 
 25. "2000".md. [methodology=treatment outcome/clinical trial] 
 26. (animal not ((human or inpatient or outpatient) and animal)).po. 
 27. or/9‐25 
 28. (27 not 26) 
 29. (8 and 28)

Cochrane Register of Controlled Trials (CENTRAL 2011, Issue 4)

#1. Gambling/ 
 #2. (gambling or gamble* or betting or wagering) 
 #3. ((pathology* or problem* or addict* or compulsive* or impulsive* or crav*) NEAR gambl*) 
 #4. ludomania* 
 #5. pokie* 
 #6. ((gaming or fruit or slot) NEAR/2 (machine*)) 
 #7. (video* NEAR/2 lotter*) 
 #8. (#1 or #2 or #3 or #4 or #5 or #6 or #7)

LILACS (to 11 October 2011)

1. (ludopatia or ((juego or jugadores or enjugadores) adj (patológico$ or problema))) 
 2. ((random$ or aleatori$ or casual or acaso or azar) or ((duplo or doble or simple or triplo or triple) and (cego or ciego)) or ((doubl$ or singl$ or tripl$ or trebl$) and (blind$ or mask$) or single‐masked study/ or double‐masked study/ or prophylatic controlled trials/ or (placebo$ and control$) or (clinical$ and trial$)) 
 3. (1 and 2)

Data and analyses

Comparison 1. Cognitive‐behaviour therapy (CBT) versus control.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Gambling symptom severity	7		Std. Mean Difference (IV, Random, 95% CI)	Subtotals only
1.1 Comparisons at 0‐3 months post‐treatment	7	402	Std. Mean Difference (IV, Random, 95% CI)	‐1.82 [‐2.61, ‐1.02]
1.2 Comparisons at 9‐12 months follow‐up	1	147	Std. Mean Difference (IV, Random, 95% CI)	‐0.11 [‐0.43, 0.22]
2 Financial loss from gambling	7		Std. Mean Difference (IV, Random, 95% CI)	Subtotals only
2.1 Comparisons at 0‐3 months post‐treatment	7	505	Std. Mean Difference (IV, Random, 95% CI)	‐0.52 [‐0.71, ‐0.33]
2.2 Comparisons at 9‐12 months follow‐up	1	147	Std. Mean Difference (IV, Random, 95% CI)	‐0.15 [‐0.47, 0.18]
3 Frequency of gambling	7		Std. Mean Difference (IV, Random, 95% CI)	Subtotals only
3.1 Comparisons at 0‐3 months post‐treatment	7	505	Std. Mean Difference (IV, Random, 95% CI)	‐0.78 [‐1.11, ‐0.45]
3.2 Comparisons at 9‐12 months follow‐up	1	147	Std. Mean Difference (IV, Random, 95% CI)	‐0.12 [‐0.45, 0.20]
4 Diagnoses of pathological gambling	2		Risk Ratio (M‐H, Random, 95% CI)	Subtotals only
4.1 Comparisons at 0‐3 months post‐treatment	2	81	Risk Ratio (M‐H, Random, 95% CI)	0.13 [0.05, 0.31]
5 Depression symptoms	4		Std. Mean Difference (IV, Random, 95% CI)	Subtotals only
5.1 Comparisons at 0‐3 months post‐treatment	4	276	Std. Mean Difference (IV, Random, 95% CI)	‐0.66 [‐0.93, ‐0.39]
6 Anxiety symptoms	4		Std. Mean Difference (IV, Random, 95% CI)	Subtotals only
6.1 Comparisons at 0‐3 months post‐treatment	4	276	Std. Mean Difference (IV, Random, 95% CI)	‐0.64 [‐0.90, ‐0.37]

Comparison 2. Motivational interviewing versus control.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Gambling symptom severity	3		Std. Mean Difference (IV, Random, 95% CI)	Subtotals only
1.1 Comparisons at 0‐3 months post‐treatment	2	163	Std. Mean Difference (IV, Random, 95% CI)	‐0.03 [‐0.55, 0.50]
1.2 Comparisons at 9‐12 months follow‐up	3	221	Std. Mean Difference (IV, Random, 95% CI)	‐0.25 [‐0.61, 0.12]
2 Financial loss from gambling	3		Std. Mean Difference (IV, Random, 95% CI)	Subtotals only
2.1 Comparisons at 0‐3 months post‐treatment	3	244	Std. Mean Difference (IV, Random, 95% CI)	‐0.41 [‐0.75, ‐0.07]
2.2 Comparisons at 9‐12 months follow‐up	2	152	Std. Mean Difference (IV, Random, 95% CI)	‐0.26 [‐0.58, 0.06]
3 Frequency of gambling	2		Std. Mean Difference (IV, Random, 95% CI)	Subtotals only
3.1 Comparisons at 0‐3 months post‐treatment	2	145	Std. Mean Difference (IV, Random, 95% CI)	‐0.18 [‐0.50, 0.15]
3.2 Comparisons at 9‐12 months follow‐up	1	62	Std. Mean Difference (IV, Random, 95% CI)	‐0.53 [‐1.04, ‐0.02]

Comparison 3. Integrative therapy versus control.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Gambling symptom severity	2		Std. Mean Difference (IV, Random, 95% CI)	Subtotals only
1.1 Comparisons at 0‐3 months post‐treatment	2	137	Std. Mean Difference (IV, Random, 95% CI)	‐0.15 [‐0.50, 0.19]
1.2 Comparisons at 9‐12 months follow‐up	2	128	Std. Mean Difference (IV, Random, 95% CI)	‐0.38 [‐0.74, ‐0.02]
2 Financial loss from gambling	2		Std. Mean Difference (IV, Random, 95% CI)	Subtotals only
2.1 Comparisons at 0‐3 months post‐treatment	2	137	Std. Mean Difference (IV, Random, 95% CI)	‐0.41 [‐1.04, 0.23]
2.2 Comparisons at 9‐12 months follow‐up	2	128	Std. Mean Difference (IV, Random, 95% CI)	‐0.06 [‐0.42, 0.29]
3 Frequency of gambling	1		Std. Mean Difference (IV, Random, 95% CI)	Subtotals only
3.1 Comparisons at 0‐3 months post‐treatment	1	52	Std. Mean Difference (IV, Random, 95% CI)	‐0.12 [‐0.69, 0.46]
3.2 Comparisons at 9‐12 months follow‐up	1	52	Std. Mean Difference (IV, Random, 95% CI)	‐0.45 [‐1.02, 0.13]

Comparison 4. Other psychological therapy versus control.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Gambling symptom severity	1		Std. Mean Difference (IV, Random, 95% CI)	Subtotals only
1.1 Comparisons at 0‐3 months post‐treatment	1	18	Std. Mean Difference (IV, Random, 95% CI)	‐1.82 [‐2.98, ‐0.66]
2 Financial loss from gambling	1		Std. Mean Difference (IV, Random, 95% CI)	Subtotals only
2.1 Comparisons at 0‐3 months post‐treatment	1	18	Std. Mean Difference (IV, Random, 95% CI)	‐1.38 [‐2.46, ‐0.30]
3 Frequency of gambling	1		Std. Mean Difference (IV, Random, 95% CI)	Subtotals only
3.1 Comparisons at 0‐3 months post‐treatment	1	18	Std. Mean Difference (IV, Random, 95% CI)	‐1.66 [‐2.78, ‐0.53]
4 Diagnoses of pathological gambling	1		Risk Ratio (M‐H, Random, 95% CI)	Subtotals only
4.1 Comparisons at 0‐3 months post‐treatment	1	18	Risk Ratio (M‐H, Random, 95% CI)	0.32 [0.12, 0.87]
5 Depression symptoms	1		Std. Mean Difference (IV, Random, 95% CI)	Subtotals only
5.1 Comparisons at 0‐3 months post‐treatment	1	18	Std. Mean Difference (IV, Random, 95% CI)	‐1.55 [‐2.66, ‐0.44]
6 Anxiety symptoms	1		Std. Mean Difference (IV, Random, 95% CI)	Subtotals only
6.1 Comparisons at 0‐3 months post‐treatment	1	18	Std. Mean Difference (IV, Random, 95% CI)	‐0.72 [‐1.70, 0.26]

Characteristics of studies

Characteristics of included studies [ordered by study ID]

Carlbring 2010.

Methods	Design: RCT Study duration: 8 weeks Follow‐up: 6 and 12 months
Participants	Sample size: 198 individuals were assessed for eligibility; 150 were randomised and 127 (that attended at least 1 treatment session) were included in the analyses Mean age: 40.5 (SD = 12.3) years Gender: 21 (16.5%) women and 106 (83.5%) men Ethnicity: 84 (66%) native Swedes; 65 (51.2%) a parent born in another country Gambling modality: the most frequent primary problematic game for the 127 participants was state sanctioned video lottery terminals in restaurants (37.8%), at legal casinos (7.1%) or at unregulated clubs (2.4%). Poker on the Internet (15.7%) and horse betting (13.4%) were also common. Casino‐type games at casinos (6.3%) or restaurants (3.1%) were less frequent. Inclusion criteria: not clearly specified Exclusion criteria: suicidal ideation (n = 13), unwillingness to be randomised (n = 6), recently commenced medication for anxiety and/or depression or being in a parallel treatment for gambling problems (n = 6), not having an ongoing gambling problem (n = 5), primary drug and/or alcohol dependence (n = 4), ongoing severe depression (n = 3), unwillingness to participate (n = 3), ongoing bipolar disorder (n = 2), imprisonment (n = 2), inability to speak Swedish (n = 2) or complete self report questionnaires (n = 1), and ongoing psychosis (n = 1)
Interventions	Group 1: Group CBT ‐ n = 59. Group CBT was administered in groups for one 3‐hour session per week for 8 weeks. The treatment was manualised, and each session focused on a set theme. Psycho‐education, exercises and homework were included in all sessions. The treatment was partly focused on cognitive restructuring and encouraging alternative behavioural strategies. Therapy also focused on identifying personal high‐risk situations for gambling and increasing skills to cope with these situations. A recurrent feature throughout treatment was the reduction of gambling urges through imaginary exposure and response prevention. Treatment goals were individually set by each client. Therapists (1 licensed clinical psychologist with psychotherapist training, 2 licensed clinical psychologists, 1 licensed social worker and 1 licensed psychiatric nurse) received continuous supervision and exclusively provided the treatment in the CBGT condition. Group 2: Motivational Interviewing (MI) ‐ n = 68. The manualised motivational interviewing treatment was administered across 50‐minute sessions (on average) that were spaced to extend across the same number of weeks as CBGT. The first 2 sessions were separated by around 7 days. The following 2 sessions were separated by around 3 to 4 weeks. The sessions used standard MI principles and explored the positive and negative consequences of gambling, and mapped the reasons for gambling. The patient was encouraged to make a decision about gambling and a change plan. MI sessions were delivered one‐on‐one. As such, total therapist time per patient was 2.5 hours on average. The typical patient attended 2.94 (SD = 1.1) sessions. Therapists (1 licensed clinical psychologist with psychotherapist training and 20 years MI experience, 1 licensed clinical psychologist with 2 years clinical MI experience, and 2 licensed social workers, 1 of whom had 10 years experience and the other was newly trained in MI) supervised themselves as a group once a month through assessment of their own audiotaped sessions, using the Motivational Interviewing Treatment Code 2.0. Group 3: No treatment control group ‐ n = 46. After 9 weeks the wait‐list control group received the allotted treatment and participants were included in the 2 active arms.
Outcomes	NORC DSM‐IV Screen for gambling problems (NODS), modified to assess gambling at 1 month (instead of 1 year); Timeline Follow‐Back (TLFB) instruments; Beck Depression Inventory‐2 (BDI‐2); Beck Anxiety Inventory (BAI); a 5‐item, 10‐point treatment credibility scale
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Randomisation was conducted by a true random‐number service independent of the investigators and therapists
Allocation concealment (selection bias)	Low risk	Randomisation was conducted by a true random‐number service independent of the investigators and therapists. Participants randomly selected an envelope that indicated their assigned condition.
Incomplete outcome data (attrition bias) All outcomes	Low risk	All participants that attended at least 1 session were included in the analysis
Selective reporting (reporting bias)	High risk	Post‐treatment means and standard deviations for the 2 treatment conditions and control group were not reported. Rather, parameter estimates from mixed‐effects analyses were presented. Statistical comparisons with the control condition combined data from the 2 different treatment conditions into a single estimate (i.e., 'any treatment' versus wait‐list control), and separate comparisons between each type of treatment and control were not reported.
Other bias	Low risk	Pre‐treatment differences between groups: the study compared groups at pre‐treatment on demographic variables and other measures (at pre‐treatment) including a measure of perceived credibility. Groups did not differ significantly on any variable. Measurement of treatment fidelity:Group CBT ‐ all therapy sessions were audiotaped and 20% of sessions were randomly selected and coded by an independent clinical psychologist with experience in the treatment method. Results showed a 93% adherence to the manual. MI ‐ all therapy sessions were audiotaped and 20% of sessions were randomly selected for coding according to the Motivational Interviewing Treatment Code 2.0 (MITI) by 1 of 4 independent and blinded coders. Therapist competency was considered good, with high attainment of reference values for MI proficiency as determined by the coding manual.
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	It is noted that self report data were collected during the clinical interview at baseline. However, it is unclear whether the outcome data collected at the post‐treatment, 6 and 12‐month follow‐up time points were completed by an outcome assessor (blinded or unblinded) or were self completed by participants.

Diskin 2009.

Methods	Design: RCT Study duration: 1 session Follow‐up: 3, 6, 9 and 12 months
Participants	Sample size: 97 individuals were randomised to groups; 81 were included in the analysis (most of the remaining 16 individuals that were randomised to groups did not attend the initial interview) Mean age: 45 years (SD = 10.6) Gender: 35 (43%) women and 46 (57%) men Ethnicity: not clearly stated; the study was conducted in Canada Gambling modality: 60.5% reported concerns with video lottery terminals. 33.3% reported concerns with slot machines. Card and dice games were a concern for 18.5%. 6.2% reported concerns with bingo. 6.2% reported problems with lottery type games and sports select lotteries. Inclusion criteria: over 17 years of age. Scores of 3 or more on the Problem Gambling Severity Index of the Canadian Problem Gambling Index. Gambled in the 2 months preceding the intake interview. Were willing to provide the name of a collateral informant and provide follow‐up data on gambling. Exclusion criteria: currently receiving treatment for problem gambling; unable to speak English well enough to participate
Interventions	Group 1: Motivational Intervention (MI) ‐ n = 42. MI therapists employed the counselling skills from Miller and Rollnick (1991, 2002). These included reflective listening, summarising and supporting self efficacy. The protocol included techniques such as a discussion of the good and not so good things about gambling, a decisional balance exercise and a readiness ruler of motivation to change. The intervention was manualised to ensure that all participants were offered a similar experience. However, the therapists were free to follow the discussion as it occurred rather than force participants to follow the pre‐determined format. 2 doctoral students each delivered half the CI and half the MI interventions. Group 2: Control Intervention (CI) ‐ n = 39.  The CI interview included the SCID II modules for histrionic, narcissistic, obsessive compulsive and avoidant personality disorders. These semi‐structured modules were intended to provide a consistent structure for the interviews. The modules were not presented as an assessment. Rather they were used as a basis for conversation and to provide structure for the interview. Participants were encouraged to talk about their gambling. The interviewers were required to be interested and pleasant, but to avoid any specific MI techniques or advice about behaviour change. When emotions were aroused during the interview the interviewers were empathic and supportive.
Outcomes	TLFB interview for number of days and dollars gambled; NODS; SOGS; PGSI‐CPGI; Global Severity Index of the Brief Symptom Inventory
Notes	Multiple measures of gambling symptom severity were available from this study. Data from the PGSI were included in the analyses.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Computer generated urn‐randomisation procedure
Allocation concealment (selection bias)	Unclear risk	The study did not indicate whether a method of allocation concealment was used
Incomplete outcome data (attrition bias) All outcomes	Low risk	An intention to treat approach was used, whereby baseline values or the 4‐week interview values were carried forward (LOCF). Separate analyses were conducted using the sample of participants who completed at least the 3‐month interview.
Selective reporting (reporting bias)	Unclear risk	It could not be determined if the report provided data on all expected outcomes
Other bias	Low risk	Pre‐treatment differences between groups: the study compared groups pre‐treatment and found no significant between‐group differences on any variables measured at baseline (gender, age, education, income, depression, alcohol and drug abuse, problem gambling measures, psychological distress, days and dollars gambled per month) Measurement of treatment fidelity: all but one interview was audiotaped and 2 blinded raters listened to 25% of the interviews. The raters were asked to complete a checklist of intended components for each interview. Ratings of the percentage of appropriate interview components documented in each interview were high.
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Telephone interviews were conducted by a research assistant who was blind to group assignment

Dowling 2007.

Methods	Design: RCT Study duration: 12 sessions Follow‐up: 6 months
Participants	Sample size: 56 individuals were randomly allocated to groups Mean age: 43.5 years (SD = 8.0) for the individual CBT group; 42.6 years (SD = 11.7) for the group CBT condition; 44.3 years (SD = 11.0) for the wait‐list control group Gender: 100% women Ethnicity: 77% of participants were born in Australia Gambling modality: 100% had a preferred gambling modality of electronic gaming machines Inclusion criteria: preferred gambling modality of electronic gaming machines; diagnosis of pathological gambling Exclusion criteria: none specified
Interventions	Group 1: Individually administered CBT ‐ n = 14. Treatment comprised 12 sessions, each around 1.5 hours in length. Therapy extended across 22.2 (SD = 13.2) weeks on average. The treatment involved components focused on setting limits on gambling behaviour, identifying and participating in alternative leisure activities, examination of thoughts and beliefs underlying gambling, examination of general problematic thoughts and beliefs, problem solving training, communication training (with a focus on assertiveness training) and relapse prevention. Treatment goals of abstinence or controlled gambling were allowed, but only participants who chose abstinence were considered in the analyses. Group 2: Group administered CBT ‐ n = 17. Treatment comprised 12 group treatment sessions, each around 2 hours in length. Group sizes ranged from 4 to 6. All participants in group treatment completed sessions across a 12‐week period. Treatment content was the same as for individually administered CBT (see above). Only participants who chose abstinence were considered in the analyses. Group 3: Wait list control ‐ n = 25. It was required that none of the participants were receiving psychosocial or supportive therapy for gambling problems.
Outcomes	Frequency, duration, amount of money inserted and expenditure gambling, measured using diary records; BDI; STAI and Coopersmith Self‐esteem Inventory; DSM‐IV‐TR criteria
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	A computer‐generated randomisation schedule was used
Allocation concealment (selection bias)	Low risk	Envelopes containing details of treatment allocation were prepared by an individual external to the trial
Incomplete outcome data (attrition bias) All outcomes	Low risk	The primary analyses were initially conducted on a treatment completer (non‐intention‐to‐treat) basis. These analyses were then repeated using an intention‐to‐treat approach. Baseline values were carried forward (LOCF) for participants who failed to complete the treatment programme or waiting list. Last available values were carried forward as follow‐up scores for participants who were not available for the duration of the follow‐up period.
Selective reporting (reporting bias)	Unclear risk	It could not be determined if the report provided data on all expected outcomes
Other bias	High risk	Pre‐treatment differences between groups: the study compared groups pre‐treatment on gambling behaviour measures and psychological functioning measures. No significant between‐group differences were found on any variable. Measurement of treatment fidelity: a treatment manual, training video, training sessions and feedback from treatment videos were employed to ensure treatment integrity. The study reported no significant differences in Goal Achievement Scale outcomes (for gambling frequency, duration and expenditure) across patients treated by the primary author and those treated by other clinicians. However, there was no quantified measurement of treatment fidelity.
Blinding of outcome assessment (detection bias) All outcomes	High risk	Outcome assessors were not blinded

Grant 2009.

Methods	Design: RCT Study duration: 6 sessions Follow‐up: 6 months
Participants	Sample size: 68 individuals were randomised to groups Mean age: 48.7 (SD = 12.8) years Gender: 43 (63.2%) women and 25 (36.8%) men Ethnicity: not clearly stated; the study was conducted in the U.S. Gambling modality: not reported Inclusion criteria: men and women aged 18 to 75; had gambled at least once per week for the past 2 months; DSM‐IV criteria for pathological gambling as measured by the SCI‐PG Exclusion criteria: past 3 months substance use disorder; positive urine drug screen at baseline; current psychotherapy or medication for pathological gambling; previous GA attendance; suicidal ideation or current use of psychotropic medications
Interventions	Group 1: CBT ‐ n = 33. The manualised CBT programme consisted of 6 sessions, each lasting 1 hour and extending across an 8‐week period. Patients commenced therapy 1 week after the baseline assessment, and were seen weekly for 6 sessions. 1 week after the final session they returned for evaluation. Session 1 consisted of psycho‐education and motivational enhancement. Subsequent sessions focused on functional analysis and behavioural strategies (Session 2), coping with gambling urges and changing irrational thinking (Session 3), imaginal desensitisation using audiotaped gambling scenarios (Session 4), relapse prevention and assertiveness training (Session 5), and significant other involvement, education and therapy (Session 6). Group 2: GA referral (wait‐list control) ‐ n = 35. The control condition involved GA referral during the treatment period, followed by the provision of therapy after the treatment group had completed the intervention. Participants in the GA group received a list of meeting times and locations for GA meetings. Individuals assigned to GA returned after 8 weeks for follow‐up. People in this group were then provided with 6 sessions of CBT, commencing shortly after the final assessment (8 weeks after baseline).
Outcomes	PG‐YBOCS; G‐SAS; CGI‐severity scale; HRSD; HAS; SDS; QOLI
Notes	Multiple measures of gambling symptom severity were available from this study. Data from the PG‐YBOCS were included in the analyses.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Computer‐generated randomisation was used
Allocation concealment (selection bias)	Unclear risk	The study did not indicate whether an appropriate method of allocation concealment was used
Incomplete outcome data (attrition bias) All outcomes	Low risk	Intention‐to‐treat analysis was conducted using last observation carried forward (LOCF)
Selective reporting (reporting bias)	Unclear risk	It could not be determined if the report provided data on all expected outcomes
Other bias	Low risk	Pre‐treatment differences between groups: the study compared groups pre‐treatment on all baseline variables. There were no significant between‐group differences on any variable. Measurement of treatment fidelity: therapist adherence and competence were assessed by audiotaping 12 subjects for independent examination by external reviewers. The adherence measure was based on the Collaborative Study Psychotherapy Rating Scale. A competency measure was based on the Cognitive Therapy Scale. Mean scores across reviewers indicated that a sufficient range of methods were applied with skill and flexibility.
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Evaluations were performed by raters blind to treatment assignment

Ladouceur 2001.

Methods	Design: RCT Study duration: 1 weekly session for a maximum of 20 sessions Follow‐up: 6 and 12 months
Participants	Sample size: 88 individuals were randomised to groups. 59 participants undertook therapy and 29 were allocated to wait‐list control. 7 of the 29 in the wait‐list condition eventually undertook treatment. 35 participants (from 66 who began treatment) completed the intervention. Mean age: 40.8 years (SD = 10.2) for the 35 participants who completed treatment; 43.4 years (SD = 10.2) for the 29 participants in the control group Gender: the 35 participants who received treatment group included 4 women and 31 men. The 29 participants in the control group included 7 women and 22 men. Ethnicity: most participants were Caucasians born into Catholic families Gambling modality: participants reported problems with gambling on gaming machines (85%), cards (28.3%), horse races (15%), sports betting (13.3%), blackjack (13.3%), bingo (11.7%), skill games (10%) and keno (10%) Inclusion criteria: primary diagnosis of pathological gambling (DSM‐IV); willing to undergo randomisation Exclusion criteria: evidence of immediate suicidal intent; current or past schizophrenia; bipolar disorder or organic mental disorder
Interventions	Group 1: CBT ‐ n = 35. Treatment was administered on an individual basis with 1 weekly session lasting 60 minutes for a maximum of 20 sessions. Treatment included cognitive correction of misconceptions about randomness and a cognitive approach to relapse prevention. Patients received an average of 11.03 (SD = 5.25) hours of treatment. 3 experienced cognitive therapists administered treatment (all were licensed psychologists). Therapists were supervised by the first author (a psychologist with 20 years experience in CBT). Group 2: Wait‐list control ‐ n = 29. After evaluation, participants in the control group were contacted and told that their name would be placed on a waiting list. Participants were offered treatment after the 3‐month waiting period, when a post‐test evaluation was conducted. No other instructions concerning treatment were delivered.
Outcomes	DSM‐IV criteria for pathological gambling assessed by clinical interview and the SOGS. Participants rated (on 0 to 10 scale) the extent of belief that they could refrain from gambling in high‐risk situations. Participants rated (on a 0 to 100 scale) their perception of control over gambling problems. Participants rated (on a 0 to 10 scale) their desire to gamble. Number of gambling sessions, number of hours spent gambling and the total amount of money spent gambling during the previous week
Notes	7 participants in the wait‐list control condition undertook the treatment following the wait‐list period. Data from these patients may have been included in both the treatment and wait‐list control group outcome data that were presented in the published report. As such, a small amount of the data may have been from non‐independent observations.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	The study did not indicate whether an appropriate randomisation procedure was used
Allocation concealment (selection bias)	High risk	Allocation to intervention was not concealed
Incomplete outcome data (attrition bias) All outcomes	High risk	Participants who dropped out from the study were excluded from the analyses
Selective reporting (reporting bias)	Unclear risk	It could not be determined if the report provided data on all expected outcomes
Other bias	Low risk	Pre‐treatment differences between groups: the study compared groups pre‐treatment on the gambling outcome measures. No significant differences on any outcome measure were observed. Measurement of treatment fidelity: all sessions were recorded and a checklist of permitted interventions (cognitive) and prohibited interventions (behavioural) guided the administration of treatment. An independent evaluator then listened to recordings of 14% of sessions for each participant to ensure that the treatment was administered as indicated on the checklist. The therapist and the independent evaluator reached a 96.4% agreement. Only 2 of the sessions evaluated were found to address slightly behavioural issues relating to money management.
Blinding of outcome assessment (detection bias) All outcomes	High risk	Outcome assessors were not blinded. Several measures were completed by participants.

Ladouceur 2003.

Methods	Design: RCT Study duration: 10 sessions Follow‐up: 6, 12 and 24 months
Participants	Sample size: 46 participants engaged in group treatment; 34 participants completed treatment; 25 participants were allocated to wait‐list control Mean age: 42.56 (SD = 10.48) years for the 34 participants who completed treatment; 44.56 (SD = 10.7) years for the 25 participants in the control group Gender: the 34 participants in the treatment group included 9 women (26.5%) and 25 men (73.5%); the 25 participants in the control group included 4 women (16%) and 21 men (84%) Ethnicity: most participants were Caucasian Gambling modality: not reported Inclusion criteria: primary diagnosis of pathological gambling; willing to undergo randomisation Exclusion criteria: evidence of current or past schizophrenia, bipolar disorder or organic mental disorder
Interventions	Group 1: Group CBT ‐ n = 34. Treatment was administered on a group basis with 10 weekly sessions, each lasting 2 hours. Treatment included cognitive correction of misconceptions about randomness and a cognitive approach to relapse prevention. 5 experienced cognitive therapists administered treatment. Group 2: Wait‐list control ‐ n = 25. Participants allocated to the control group were contacted and told that their names would be placed on a waiting list. They were offered treatment after a 4‐month waiting period, when a post‐intervention evaluation was conducted. No other instruction concerning treatment was delivered.
Outcomes	Number of DSM‐IV criteria for pathological gambling assessed through clinical interview. Participants rated (on 0 to 10 scale) the extent of their belief that they could refrain from gambling in high‐risk situations. Participants rated (on a 0 to 100 scale) their perception of control over gambling problems. Participants rated (on a 0 to 10 scale) their desire to gamble. Number of gambling sessions, number of hours spent gambling and total amount of money spent gambling during the previous week
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	The study did not indicate whether an appropriate randomisation procedure was used
Allocation concealment (selection bias)	High risk	Allocation to intervention was not concealed
Incomplete outcome data (attrition bias) All outcomes	High risk	Participants who dropped out from the study were excluded from the analyses
Selective reporting (reporting bias)	Unclear risk	It could not be determined if the report provided data on all expected outcomes
Other bias	Low risk	Pre‐treatment differences between groups: the study compared groups at pre‐treatment on demographic and outcomes variables. There were no significant differences between groups on any variable. Measurement of treatment fidelity: all sessions were recorded and a checklist of permitted interventions (cognitive) and prohibited interventions (behavioural) guided the administration of treatment. An independent evaluator (a trained graduate student) then listened to recordings of 20% of sessions for each participant to ensure that the treatment was administered as indicated on the checklist. The therapist and the independent evaluator reached a 100% agreement. Only cognitive interventions were administered in the sessions that were evaluated.
Blinding of outcome assessment (detection bias) All outcomes	High risk	Outcome assessors were not blinded. Several measures were completed by participants.

Marceaux 2011.

Methods	Design: RCT Study duration: 8 weeks Follow‐up: 6 months
Participants	Sample size: 49 participants were allocated to groups; 11 dropped out before receiving information about group assignment. As such, 38 individuals were allocated to groups. 33 of these participants completed treatment (or wait‐list control). Mean age: 48.56 years (SD = 10.38) for the wait‐list control group; 47.64 years (SD = 12.11) for the Twelve‐Step Facilitated Group Therapy; 47.44 years (SD = 10.50) for the Cognitive‐Behavioural Group Therapy with Mapping Gender: of the 38 participants allocated to groups, 23 (60.5%) were women and 15 (39.5%) were men Ethnicity: of the 38 participants allocated to groups, 33 (86.8%) were Caucasian, 4 (10.5%) were African‐American and 1 (2.6%) were 'other' Gambling modality: not reported Inclusion criteria: at least 21 years of age; met 5 or more DSM‐IV criteria in the past year and endorsed 5 or more items on the SOGS (lifetime) Exclusion criteria: none specified
Interventions	Group 1: Twelve‐Step Facilitated (TSF) Group Therapy ‐ TSF comprised an 8‐week programme that consisted of 2 90‐minute group sessions per week. The treatment was manualised and reflected a structured programme with a specific agenda and format. The treatment objectives were congruent with the GA view of pathological gambling and addressed cognitive, emotional, behavioural, social and spiritual issues relating to gambling. Principles of motivational enhancement therapy were also included. The primary focus of the intervention was on the first 5 'steps' of GA, with each session focusing on specific recovery tasks and reading material. Participants were asked keep a journal throughout the intervention. Group 2: Cognitive‐Behavioural Group Therapy with Mapping (CBGT‐mapping) ‐ CBGT‐mapping was an 8‐week program consisting of 2 90‐minute sessions per week. The first 2 sessions include a formal assessment of participants, rapport building, introduction of node‐link mapping and the detailed imagination of a gambling experience. Remaining sessions used node‐link mapping techniques and focused on 3 'thought‐sets'. These corresponded to: (1) understanding randomness; (2) problem‐solving; and (3) preventing relapse. Each session utilised node‐link maps corresponding to the session topic. Group 3: Wait‐List Control ‐ participants in the wait‐list condition were informed that there would be an 8‐week wait until the next group would begin therapy
Outcomes	DSM‐IV criteria interview (past month time‐frame at both post‐ and 6‐month assessments), assessment of control (measuring ability to control gambling, ability to refrain from gambling, and desire to gamble, each scored using 10‐point Likert scale response format), BDI‐II, BAI and TLFB (frequency of gambling, money spent gambling and time spent gambling)
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	A table of random numbers was used to assign participants to groups
Allocation concealment (selection bias)	High risk	Allocation was concealed from the participants but not from all the investigators
Incomplete outcome data (attrition bias) All outcomes	High risk	Participants who dropped out from the study were excluded from the analyses
Selective reporting (reporting bias)	Unclear risk	It could not be determined if the report provided data on all expected outcomes
Other bias	Unclear risk	Pre‐treatment differences between groups: the study compared groups at pre‐treatment on demographic variables, gambling outcomes (at pre‐treatment), and depression and anxiety scores. No significant differences were found between the groups on any variables. Measurement of treatment fidelity: the study did not indicate whether any procedures were used to evaluate treatment fidelity
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Outcome assessors were blinded

Melville 2004a.

Methods	Design: RCT Study duration: 8 weeks Follow‐up: 6 months
Participants	Sample size: 20 individuals responded to an advertisement and 13 were randomised to groups Mean age: 68 years (SD not reported) for the node‐link mapping group; 52 years (SD not reported) for the non‐mapping therapy group; 53 years (SD not reported) for the control group Gender: from the 13 participants that were randomised and undertook treatment, 8 (61.5%) were women and 5 (38.5%) were men Ethnicity: from the 20 individuals that responded to the advertisement, 16 (85%) were white Gambling modality: individuals responding to the advertisement identified slot machines (62%) or video poker (37%) as their primary gambling activity Incusion criteria: DSM‐IV criteria for pathological gambling; scores of 5 or more on the SOGS Exclusion criteria: none specified
Interventions	Group 1: Node‐Link Mapping Group CBT ‐ n = 4. Prior to group treatment, each participant met individually with a therapist for 30 minutes. They were provided with a description of the format of the sessions along with session dates and times. Participants also made audio recordings of thoughts that occurred during an episode of imagined gambling. Subsequent group sessions then used node‐link mapping to address 3 topic areas: (1) understanding randomness; (2) problem solving; and (3) relapse prevention. 8 hours of therapy were devoted to each topic. Node‐link mapping is a visual representation technique designed to highlight interrelations between thoughts, emotions, actions and environmental influences. Group sessions were 90 minutes in length, and were conducted twice per week for 8 weeks. Therapists were Masters level counsellors. Group 2: Non‐Mapping Group CBT ‐ n = 4. The non‐mapping group received the same therapy content as the node‐link mapping group CBT (focused on understanding randomness, problem solving and relapse prevention). However, treatment was delivered using natural language and without node‐link maps. Group sessions were 90 minutes in length, and were conducted twice per week for 8 weeks. Therapists were Masters level counsellors. Group 3: Wait‐List Control ‐ n = 5. Participants were informed that because of random assignment they were placed on a waiting list. They were also informed that the treatment would begin in 8 weeks.
Outcomes	DSM‐IV criteria. Self ratings of control over gambling, desire to gamble and ability to refrain from gambling, each scored using a 10‐point Likert scale response format. Number of times gambled, number of hours spent gambling and amount spent gambling in the past 30 days. Number of sessions of Gamblers Anonymous (or other treatment for problem gambling) attended in the past 30 days.
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	The study did not indicate whether an appropriate randomisation procedure was used
Allocation concealment (selection bias)	Unclear risk	The study did not indicate whether an appropriate method of allocation concealment was used
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	The study did not indicate how incomplete outcome data was managed
Selective reporting (reporting bias)	High risk	No standard deviations for any outcome measures were reported
Other bias	Unclear risk	Pre‐treatment differences between groups: the study did not indicate whether pre‐treatment differences between groups were examined Measurement of treatment fidelity: the study did not indicate whether any procedures were used to evaluate treatment fidelity
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Assessment personnel were trained interns who were blind to the treatment conditions

Melville 2004b.

Methods	Design: RCT Study duration: 8 weeks Follow‐up: 6 months
Participants	Sample size: 28 individuals responded to an advertisement and 19 were allocated to groups Mean age: 51 years (SD not reported) for node‐link mapping group therapy; 54 years (SD not reported) for the control group Gender: from the 19 participants that undertook treatment, 16 (84.2%) were women and 3 (15.8%) were men Ethnicity: from the 28 individuals that responded to the advertisement, 25 (89%) were white Gambling modality: the 28 individuals that responded to the advertisement mainly identified slot machines (62%) or video poker (38%) as their primary gambling activity Inclusion criteria: DSM‐IV criteria for pathological gambling; scores of 5 or more on the SOGS Exclusion criteria: none specified
Interventions	Group 1: Node‐Link Mapping Group CBT ‐ n = 9. The treatment procedures were the same as those described under the Node‐Link Mapping Group CBT in Melville 2004a. Group 2: Wait‐list control group ‐ n = 10. The treatment procedures were the same as those described under the Wait‐List Control in Melville 2004a.
Outcomes	DSM‐IV criteria. Self ratings of control over gambling, desire to gamble and ability to refrain from gambling, each scored using a 10‐point Likert scale response format. Number of times gambled, number of hours spent gambling and amount spent gambling in the past 30 days. Number of sessions of Gamblers Anonymous (or other treatment for problem gambling) attended in the past 30 days. Measures of depression (BDI‐II) and anxiety (BAI).
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	The study did not indicate whether an appropriate randomisation procedure was used
Allocation concealment (selection bias)	Unclear risk	The study did not indicate whether an appropriate method of allocation concealment was used
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	The study did not indicate how incomplete outcome data was managed
Selective reporting (reporting bias)	High risk	No standard deviations for outcome measures were reported
Other bias	Unclear risk	Pre‐treatment differences between groups: the study did not indicate whether pre‐treatment differences between groups were examined Measurement of treatment fidelity: the study did not indicate whether any procedures were used to evaluate treatment fidelity
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Assessment personnel were trained interns who were blind to the treatment conditions

Oei 2010.

Methods	Design: RCT Study duration: 6 weeks Follow‐up: 6 months
Participants	Sample size: 102 individuals were allocated to groups Mean age: 43.43 years (SD = 11.87) for individual CBT; 48.67 years (SD = 11.75) for group CBT; 43.38 years (SD = 9.71) years for the control group Gender: 42 (41.18%) women and 60 (58.82%) men Ethnicity: across groups, around 80% of participants were Caucasian Gambling modality: around 60% to 70% of each group were regular (more than once a week) players of electronic gaming machines Inclusion criteria: over 18 years of age; met the DSM‐IV criteria for pathological gambling; provided informed consent to participate Exclusion criteria: receiving additional assistance for gambling from a health professional (e.g., psychologist, counselling agency); poor proficiency with English; met DSM‐IV criteria for personality disorder(s) (assessed via clinical interview); other conditions (e.g., hypomania) accounting for gambling behaviour
Interventions	Group 1: Individual CBT ‐ n = 37 (sample size used in the analysis was n = 51; see Notes below). The manualised individual CBT programme was delivered across one 2‐hour session per week for 6 weeks. The treatment programme included components focused on psycho‐education, motivational interviewing, correction of erroneous cognitions, relapse prevention, goal setting skills training, problem solving skills training and imaginal exposure. Therapy also included cognitive and behavioural self management strategies to stabilise gambling and cope with the urges and negative consequences of gambling. Group 2: Group CBT ‐ n = 37 (sample size used in the analysis was n = 51; see Notes below). The manualised group CBT program was delivered across one 2.5 hour session per week for 6 weeks. The components of therapy were the same as those for individual CBT (see above). Group 3: Wait‐List Control ‐ n = 28. Participants received treatment after a 6‐week waiting period. These participants were then allocated to either the Individual CBT or Group CBT condition after the wait‐list period.
Outcomes	Gambling Related Cognitions Scale; Gambling Urge Scale; Depression Anxiety Stress Scale‐21; Satisfaction with Life Scale; average amount gambled per day; frequency of gambling measured using 5‐point ordinal scales extending from 'never' to 'monthly or less' to '2‐4 times a week' to '2‐3 times a week' and '4 or more times a week' for frequency of gambling on gaming machines, table games, betting on animals and other forms of gambling (in the past 2 weeks)
Notes	The n = 28 participants allocated to wait‐list control were subsequently allocated to Individual CBT or Group CBT after the waiting period. The post‐treatment data from these participants (originally in the wait‐list condition) were then included when calculating summary statistics for the outcomes variables for these treatment conditions. As such, the sample size for both treatment conditions was inflated to n = 51 (with the inclusion of an additional 14 participants in each treatment condition). The treatment groups thus include data from non‐independent observations and cases that also made up the wait‐list control condition.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Participants were randomly assigned to groups using a computer generated randomised number system
Allocation concealment (selection bias)	Unclear risk	The study did not indicate whether an appropriate method of allocation concealment was used
Incomplete outcome data (attrition bias) All outcomes	Low risk	Intention‐to‐treat analysis was conducted using LOCF
Selective reporting (reporting bias)	Unclear risk	It could not be determined if the report provided data on all expected outcomes
Other bias	Unclear risk	Pre‐treatment differences between groups: the study compared groups at pre‐treatment on demographic variables and engagement in various forms of gambling. Significant differences between the groups were found on 2 demographic variables. Participants in the individual and wait‐list conditions were more likely to have a higher distribution of income, and be employed full time, compared to participants in the group treatment condition. The report did not indicate whether subsequent analyses adjusted for these baseline differences. Measurement of treatment fidelity: therapy sessions were video‐taped. An adaptation of the Yales Adherence and Competence Scale showed that therapists in both treatment conditions adhered to the manual closely and were competent in delivering CBT for gambling problems.
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	The study did not indicate whether outcome assessors were blinded

Petry 2006.

Methods	Design: RCT Study duration: 8 weeks Follow‐up: 6 and 12 months
Participants	Sample size: 231 individuals were randomised to groups Mean age: 44.4 years (SD = 11.7) for the GA referral group; 44.3 years (SD = 9.4) for the CB workbook group; 45.8 years (SD = 11.6) years for the individual CBT group Gender: 104 (45%) women and 127 (55%) men Ethnicity: 20 (8.7%) African American, 195 (84.4%) European American, 10 (4.3%) Hispanic American and 6 (2.6%) other Gambling modality: across groups, the preferred forms of gambling were electronic machines (> 40% of each group), cards (11% to 20%), scratch/lottery (10% to 16%), sports (6% to 14%), animal races, craps or dice games (all < 10%) Inclusion criteria: DSM‐IV criteria for pathological gambling; gambled in the past 2 months; 18 years or older; reading at the 5th grade level Exclusion criteria: current suicidal intentions; past month psychotic symptoms; already receiving gambling treatment
Interventions	Group 1: Individual CBT and GA Referral ‐ n = 84. Participants were initially provided with a list of locations and times for GA meetings. They subsequently met with a therapist for 1 hour each week for 8 weeks. Sessions were structured by handouts and addressed: (a) triggers for gambling; (b) functional analysis of gambling; (c) scheduling pleasant activities; (d) self management planning; (e) coping with urges to gamble; (f) assertiveness training (and gambling refusal skills training); (g) correcting irrational thinking; and (h) coping with gambling lapses. A workbook on handling gambling‐related debt was provided, while most sessions had homework exercises. 10 masters‐level and 3 doctoral‐level therapists delivered treatment. They received training and close supervision of at least 1 case. Ongoing supervision consisted of regular review of therapy notes, audiotapes and case discussion. Group 2: CBT Workbook and GA Referral ‐ n = 84. During a 10 to 15‐minute session with a therapist, participants were provided with a list of locations and times for GA meetings. They subsequently received a 70‐page workbook containing CB exercises and a section on handling gambling‐related debt. The workbook contained descriptions and exercises identical to those in the therapy condition. The evaluator instructed participants to complete 1 chapter per week for 8 weeks. Group 3: GA Referral ‐ n = 63. During a 10 to 15‐minute session with a therapist, participants were provided a list of the locations and meeting times of GA meetings held throughout the state of Connecticut. GA was discussed, including prior attendance, expectations and potential concerns. Participants were told that many people who become involved with GA reduce or stop gambling, and they were encouraged to select a GA meeting to attend.
Outcomes	SOGS; Addiction Severity Index (ASI); TLFB (days gambled and amounts lost daily were recorded); Brief Symptom Inventory (BSI); Service Utilisation Form; collaterals identified by participants were interviewed about the participants' gambling behaviours
Notes	Multiple measures of gambling symptom severity were available from this study. Data from the SOGS were included in the analyses.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Computerised urn randomisation balanced 3 groups on SOGS scores, age, gender, ethnicity and prior gambling treatment
Allocation concealment (selection bias)	Unclear risk	The study did not indicate whether an appropriate method of allocation concealment was used
Incomplete outcome data (attrition bias) All outcomes	Low risk	Intention‐to‐treat analysis was conducted using random‐effects regression
Selective reporting (reporting bias)	Unclear risk	It could not be determined if the report provided data on all expected outcomes
Other bias	Low risk	Pre‐treatment differences between groups: the study compared groups at pre‐treatment on a wide range of demographic variables and gambling outcomes (at pre‐treatment). A significant difference was found between the groups on ASI gambling scores. The GA referral condition had lower ASI‐G scores compared to the other conditions. Subsequent analyses of this variable took these differences into account. Measurement of treatment fidelity: therapy sessions were audiotaped and were evaluated by 4 individuals according to a modification of the Yale Adherence Competence Scale. The raters scored CB and non‐CB (e.g., psycho‐educational/case management) items on a 7‐point Likert scale (1 = none/poor, 3 = some/adequate, 7 = extensive/exceptional). For CB items, average scores were 4.3 (SD = 0.8) and around 'good/quite a bit'. For non‐CB items (which were not intended to be covered in sessions), average scores were 1/1 (SD = 0.3) and around 'none/poor'.
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	The study did not indicate if outcome assessors were blinded

Petry 2008.

Methods	Design: RCT Study duration: interventions ranged from 10 minutes to a 4 sessions Follow‐up: 9 months
Participants	Sample size: 180 individuals were randomised to groups Mean age: 41.4 years (SD = 12.5) years for the assessment only group; 43.5 years (SD = 14.4) for the brief advice group; 45.0 years (SD = 13.8) years for the MET group; 44.0 years (SD = 10.2) years for the MET and condensed CBT group Gender: 72 (40%) women and 108 (60%) men Ethnicity: 41 (22.8%) African American, 110 (61.1%) European American, 21 (11.7%) Hispanic American and 3 (1.7%) other (with some missing data) Gambling modality: across groups, the most common preferred forms of gambling were scratch/lottery (> 40% of each group), cards (8% to 28%), slot machines (14% to 22%), sports (9% to 14%) and dice (2% to 14%) Inclusion criteria: at least 3 items on the SOGS; at least USD 100 wagering and gambled on at least 4 occasions in the past 2 months; 18 years or older Exclusion criteria: reading below 5th grade level; past month suicidal intentions or psychotic symptoms; interest in receiving more intensive gambling treatment than provided in the study
Interventions	Group 1: MET ‐ n = 55. During a 50‐minute session, therapists provided personalised feedback about participants' gambling. They subsequently explored positive and negative consequences of gambling, and discussed how gambling fit with their goals and values. Participants also completed a change plan worksheet. 9 therapists (2 with bachelors degrees and 7 with masters degrees) delivered this form of therapy, as well as the other interventions (see below). They received didactic training and close supervision of at least 1 case. Supervision consisted of reviews of therapy notes and audiotapes, and case discussion. Group 2: MET and Condensed CBT ‐ n = 40. Participants met with a therapist and received the same MET session as described above. They were also encouraged to return for 3 sessions of CBT at subsequent times. CBT sessions involved determination of internal and external triggers of gambling, discussion of methods for coping with internal gambling triggers, external gambling triggers, and for coping with gambling cues or cravings. Group 3: Brief Advice ‐ n = 37. Participants met with a therapist for around 10 minutes. Using a 1‐page handout, the therapist described the participants' level of gambling in relation to the general population. They also outlined risk factors for severe gambling problems, and provided advice on steps to curtail development of problems. These steps included limiting the amount of money spent gambling, reducing the time spent gambling, not viewing gambling a means of making money, and spending time engaged in other activities. Group 4: Assessment Only Control ‐ n = 48. After the baseline evaluation, research assistants informed participants that they would be re‐contacted for follow‐up evaluations in 6 weeks and 9 months.
Outcomes	NODS (at baseline only); ASI; SOGS (past‐month version); Brief Symptom Inventory (at baseline only); treatment service review
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	From envelopes, participants selected slips of paper indicating group assignment
Allocation concealment (selection bias)	Low risk	From envelopes, participants selected slips of paper indicating group assignment
Incomplete outcome data (attrition bias) All outcomes	Low risk	Intention‐to‐treat analysis was conducted using random‐effects regression
Selective reporting (reporting bias)	Unclear risk	It could not be determined if the report provided data on all expected outcomes
Other bias	Low risk	Pre‐treatment differences between groups: the study compared groups at pre‐treatment on a wide range of demographic variables and gambling outcomes (at pre‐treatment). Groups did not differ significantly on any demographic characteristic or measures of gambling or gambling problem severity. Measurement of treatment fidelity: therapy sessions were audiotaped and 37 randomly selected tapes were evaluated by 4 individuals according to a modification of the Yale Adherence Competence Scale. The raters scored CBT, MET and brief advice items on a 7‐point Likert scale (1 = none/poor, 3 = some/adequate, 7 = extensive/exceptional). The therapies were all found to be distinguishable.
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Research assistants conducted the follow‐up evaluations and were blind to treatment conditions

Petry 2009.

Methods	Design: RCT Study duration: interventions ranged from 10 minutes to 4 sessions Follow‐up: 9 months
Participants	Sample size: 117 individuals were randomised to groups Mean age: 20.5 years (SD = 2.0) for the assessment only group; 20.2 years (SD = 1.9) for the brief advice group; 20.5 years (SD = 1.4) for the MET group; 20.1 years (SD = 1.4) for the MET and condensed CBT group Gender: 18 (15.4%) women and 99 (84.6%) men Ethnicity: 100 (85.5%) European American, 10 (8.6%) Asian/Pacific Islander and 7 (6.0%) other Gambling modality: across groups, the most common preferred forms of gambling were cards/poker (43% to 62% across groups), sports (2% to 20%), scratch tickets/lottery (6% to 15%) and internet (0% to 15%) Inclusion criteria: 18 years or older; scores of 3 or more on the SOGS; spent USD 100 or more in wagering and gambled on 4 or more days in the past 2 months Exclusion criteria: past‐month suicidal intentions; psychotic symptoms; interest in receiving more intensive gambling treatment than provided in the study
Interventions	Group 1: MET ‐ n = 30. Therapy was the same as the MET intervention delivered in Petry 2008. 6 therapists (1 bachelor level, 2 master level, 2 clinical psychology doctoral students, 1 PhD psychologist) delivered this form of therapy, as well as the other interventions (see below). They received didactic training and close supervision of at least 1 case. Supervision consisted of reviews of therapy notes and audiotapes, and case discussion. Group 2: MET and Condensed CBT ‐ n = 21. Therapy was the same as the MET and Condensed CBT intervention delivered in Petry 2008. Group 3: Brief Advice ‐ n = 32. Therapy was the same as the brief advice intervention delivered in Petry 2008 Group 4: Assessment Only Control ‐ n = 34. After the baseline evaluation, research assistants informed participants that they would be re‐contacted for follow‐up evaluations in 6 weeks and 9 months.
Outcomes	NOD (at baseline only); SOGS (past‐month version); ASI‐G; TLFB (at baseline only); treatment service review
Notes	Participants were recruited from university campuses
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	From envelopes, participants selected slips of paper indicating group assignment
Allocation concealment (selection bias)	Low risk	From envelopes, participants selected slips of paper indicating group assignment
Incomplete outcome data (attrition bias) All outcomes	Low risk	Intention‐to‐treat analysis was conducted using hierarchical linear modelling
Selective reporting (reporting bias)	Unclear risk	It could not be determined if the report provided data on all expected outcomes
Other bias	Low risk	Pre‐treatment differences between groups: the study compared groups at pre‐treatment on a wide range of demographic variables and measures of gambling (at pre‐treatment). Groups did not differ significantly on any variable. Measurement of treatment fidelity: therapy sessions were audiotaped and 49 randomly selected tapes were evaluated by 5 individuals according to a modification of the Yale Adherence Competence Scale. The raters scored CBT, MET and brief advice items on a 7‐point Likert scale (1 = none/poor, 3 = some/adequate, 7 = extensive/exceptional). The therapies were all found to be distinguishable.
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Research assistants (some of whom were also therapists) conducted follow‐ups and were unaware of treatment assignment

Sylvain 1997.

Methods	Design: RCT Study duration: maximum of 30 hours of treatment Follow‐up: 6 and 12 months
Participants	Sample size: 58 individuals were initially recruited into the study; 18 refused treatment after the initial evaluation, and the remaining 40 were randomised to groups; 29 participants completed treatment (11 dropped out of the study) and were included in the analysis Mean age: 37.6 years (SD = 10.3) for the CBT group; 42.6 years (SD = 12.1) for the control group Gender: 100% male Ethnicity: 100% Caucasian Gambling modality: the majority of participants were described as video poker players; others were described as gambling on horse races or casino games Inclusion criteria: met DSM‐III‐R criteria for pathological gambling Exclusion criteria: none specified
Interventions	Group 1: CBT ‐ n = 14. Treatment comprised one or 2 weekly sessions, each lasting from 60 to 90 minutes (depending on patient goals). Treatment was maintained until patients developed an adequate perception of gambling and chance and ceased gambling. Participants received an average of 16.7 hours of treatment (SD = 5.7). Therapy included 4 components: (1) cognitive correction; (2) problem solving training; (3) social skills training; and (4) relapse prevention. 2 psychologists, with 4 and 5 years of clinical experience (respectively), administered the treatment. Group 2: Wait‐list control ‐ n = 15. After the initial evaluation, participants were contacted and told that their name would be placed on a waiting list. The therapist assured the participants that they would be called as soon as a place became available. It was expected that all participants would be offered treatment in the following 4 months. The therapist contacted them by telephone once a month.
Outcomes	DSM‐III‐R criteria; SOGS; perception of control, desire to gamble and self efficacy perception (each scored using 10‐point Likert scale response format); number of gambling sessions, number of hours spent gambling and total amount of money spent gambling during the previous week
Notes	2 patients in the wait‐list control condition could not wait longer and were immediately given treatment. It is unclear whether data from these patients were included in the analyses. Multiple measures of gambling symptom severity were available from this study. Data from the SOGS were included in the analyses.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	The study did not indicate whether an appropriate randomisation procedure was used
Allocation concealment (selection bias)	Unclear risk	The study did not indicate whether an appropriate method of allocation concealment was used
Incomplete outcome data (attrition bias) All outcomes	High risk	The participants who dropped out from the study were excluded from the analysis
Selective reporting (reporting bias)	Unclear risk	It could not be determined if the report provided data on all expected outcomes
Other bias	Low risk	Pre‐treatment differences between groups: the study compared groups at pre‐treatment on the dependent variables. Groups did not differ significantly on any variable. Measurement of treatment fidelity: all sessions were recorded and therapists completed a checklist of interventions after each session. An independent evaluator then listened to 20% of the interviews for each participant. All sessions rated by the evaluator had at least 1 required component of treatment. 48% of sessions contained information about gambling and the cognitive correction of beliefs. The development of skills pertaining to problem solving and relapse prevention were noted in 28% and 56% of the sessions, respectively.
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	The study did not indicate if outcome assessors were blinded

• ASI: Addiction Severity Index
• ASI‐G: gambling subscale of the Addiction Severity Index
• BAI: Beck Anxiety Inventory
• BDI: Beck Depression Inventory
• CB: cognitive‐behavioural
• CBGT: cognitive‐behavioural group therapy
• CBT: cognitive‐behaviour therapy
• CGI: Clinical Global Impression
• CI: control intervention
• DSM‐IV: Diagnostic and Statistical Manual of Mental Disorders, 4th edition
• GA: Gamblers Anonymous
• G‐SAS: Gambling Symptom Assessment Scale
• HAS: Hamilton Anxiety Scale
• HRSD: Hamilton Rating Scale for Depression
• LOCF: last observation carried forward
• MET: motivational enhancement therapy
• MI: motivational interviewing
• NODS: NORC DSM‐IV Screen for gambling problems
• PGSI: Problem Gambling Severity Index
• PGSI‐CPGI: Problem Gambling Severity Index of the Canadian Problem Gambling Index
• PG‐YBOCS: Yale Brown Obsessive Compulsive Scale adapted for Pathological Gambling
• QOLI: Quality of Life Inventory
• RCT: randomised controlled trial
• SCI‐PG: Structured Interview for Pathological Gambling
• SD: standard deviation
• SDS: Sheehan Disability Scale
• SOGS: South Oaks Gambling Screen
• STAI: State‐Trait Anxiety Inventory
• TLFB: Timeline Follow‐Back
• TSF: Twelve‐Step Facilitated (Group Therap

Characteristics of excluded studies [ordered by study ID]

Study	Reason for exclusion
Blanco 2010	The study compared 2 CBT interventions; it therefore lacked an appropriate control group for the purposes of this review
Carlbring 2008	The psychological therapy was delivered via the internet and not face‐to‐face
Cunningham 2009	The treatment was not delivered face‐to‐face by a practitioner
Dickerson 1990	Participants were not identified as pathological or problem gamblers using an accepted assessment strategy
Echeburua 1996	This was a quasi‐randomised controlled trial
Echeburua 2000	This was a quasi‐randomised controlled trial
Echeburua 2011	This was a quasi‐randomised controlled trial
Harvey 2010	The study compared 2 CBT interventions; it therefore lacked an appropriate control group for the purposes of this review
Hodgins 2001	The psychological therapy was delivered by telephone and not face‐to‐face
Hodgins 2007	The study compared 2 CBT interventions; it therefore lacked an appropriate control group for the purposes of this review
Hodgins 2009	The psychological therapy was delivered by telephone and not face‐to‐face
Korman 2008	The study compared CBT with a treatment‐as‐usual condition that also comprised significant cognitive and behavioural components; it therefore lacked an appropriate control group for the purposes of this review
McConaghy 1983	The study compared 2 CBT interventions; it therefore lacked an appropriate control group for the purposes of this review
McConaghy 1988	The study compared 2 CBT interventions; it therefore lacked an appropriate control group for the purposes of this review
McConaghy 1991	The study compared 2 CBT interventions; it therefore lacked an appropriate control group for the purposes of this review
Milton 2002	The study compared 2 CBT interventions; it therefore lacked an appropriate control group for the purposes of this review
Myrseth 2009	This was a quasi‐randomised controlled trial
Petry 2005c	The study compared 2 CBT interventions; it therefore lacked an appropriate control group for the purposes of this review
Petry 2006b	Participants were not identified as pathological or problem gamblers using an accepted assessment strategy
Petry 2008b	This study compared among 3 types of psychological interventions; it thus lacked an appropriate control group for the purposes of this review
Sani 2005	Participants were not identified as pathological or problem gamblers using an accepted assessment strategy
Toneatto 2007b	The study compared 2 CBT interventions; t therefore lacked an appropriate control group for the purposes of this review
Tse 2010	The study compared 2 CBT interventions; it therefore lacked an appropriate control group for the purposes of this review

CBT: cognitive‐behaviour therapy

Characteristics of studies awaiting assessment [ordered by study ID]

Anonymous 2003.

Methods	RCT
Participants	Individuals aged 17 years and over that meet the DSM‐IV criteria for pathological gambling
Interventions	Cognitive motivational behaviour therapy versus a Gamblers Anonymous control group
Outcomes
Notes	ClinicalTrials.gov Identifier: NCT00069420

Anonymous 2004.

Methods	RCT
Participants	Individuals aged between 17 and 24 years that score of 3 or more on the SOGS
Interventions	Brief personal feedback
Outcomes	Measures will include the NORC DSM Screen for Problem Gambling (NODS), the South Oaks Gambling Screen (SOGS), Gambling Quantity and Perceived Norms, Perceived Injunctive Gambling Norms, Gambling Problems Index, gambling frequency, attitudes and beliefs about gambling and self control, readiness to change, gambling expectancies, gambling motives, gambling risk perception, psychiatric symptoms, assertiveness, coping skills, substance use, alcohol‐related problems, self determination and social desirability
Notes	ClinicalTrials.gov Identifier: NCT00078273

Cunningham 2012.

Methods
Participants
Interventions
Outcomes
Notes

Petry 2008c.

Methods	RCT
Participants	Individuals aged 18 years and over and meet the DSM‐IV criteria for pathological gambling
Interventions	MET versus MET + CBT versus MET + CBT + aftercare treatment
Outcomes	The primary outcomes are self and collateral indices of gambling and gambling related problems
Notes	ClinicalTrials.gov Identifier: NCT00685724

DSM‐IV: Diagnostic and Statistical Manual of Mental Disorders, 4th edition 
 MET: motivational enhancement therapy 
 RCT: randomised controlled trial 
 SOGS: South Oaks Gambling Screen

Characteristics of ongoing studies [ordered by study ID]

Bellringer 2009.

Trial name or title	For callers to a problem gambling helpline are motivational interviewing‐based self help interventions as good as or better than standard care in reducing gambling involvement?
Methods	RCT
Participants	Individuals aged 18 years and over that perceive they have a gambling problem
Interventions	Single‐session motivational interview versus single‐session motivational interview + self help workbook versus single session motivational interview + self help workbook + 4 'booster' telephone calls versus standard treatment
Outcomes	The primary outcomes include number of days gambled per month and total dollars spent on gambling per month
Starting date	Unknown (date submitted 8 July 2009)
Contact information	Maria Bellringer (maria.bellringer@aut.ac.nz), Max Abbott (max.abbott@aut.ac.nz)
Notes	Main ID: ACTRN12609000560291

Ciesielski 2008.

Trial name or title	Brief therapies for problem gambling substance abusers
Methods	RCT
Participants	Individuals aged 18 years and older that meet the DSM‐IV criteria for alcohol, cocaine, opioid or marijuana abuse or dependence and score 3 or more on the SOGS
Interventions	Psycho‐education versus brief advice versus MET + CBT
Outcomes	The primary outcomes are dollars and days gambled per month
Starting date	July 2007
Contact information	Ellen Ciesielski (eciesielski@uchc.edu)
Notes	ClinicalTrials.gov Identifier: NCT00685048

DSM‐IV: Diagnostic and Statistical Manual of Mental Disorders, 4th edition 
 MET: motivational enhancement therapy 
 RCT: randomised controlled trial 
 SOGS: South Oaks Gambling Screen

Differences between protocol and review

There were a number of differences between the protocol and the review.

• The aims of the review have been restricted in scope. The aim of this review is not to consider the comparative efficacy of different psychological therapies, and it will not consider studies that compare a form of psychological therapy with another psychological or pharmacological intervention.

• The review has excluded from consideration 'minimal therapies' (i.e., therapies involving zero or less than 30 minutes of face‐to‐face contact with a trained therapist).

• The review has combined individual CBT and group CBT into a single category of therapy.

• Proposed subgroup and sensitivity analyses have been modified to reduce the number of analyses.

Contributions of authors

Sean Cowlishaw wrote the review and co‐ordinated contributions from Stephanie Merkouris, Shane Thomas, Nicki Dowling and Alun Jackson. Chris Anderson provided support in conducting the review and developing drafts of the report. All authors provided input and critical feedback on the final version of the review.

Sources of support

Internal sources

• Problem Gambling Research and Treatment Centre, Monash University, Australia.

  Staff, facilities and resources

External sources

• Victorian Government, Department of Justice, Australia.

  Financial

Declarations of interest

None.

New