Results of the search

We identified 23,901 records through database searching and 36 from other sources. After removing duplicates, we were left with 19,087 unique references for analysis. We excluded 18,951 on the basis of title and abstract. We retrieved 136 articles in full text for more detailed evaluation, 127 of which were excluded for not meeting the inclusion criteria.

We included seven studies (from nine articles) that satisfied all criteria required for inclusion in the review. See Figure 1.

Figure 1

---

---

Study flow diagram.

For substantive descriptions of studies see the Characteristics of included studies and Characteristics of excluded studies tables

Included studies

Seven studies enrolling a total of 608 participants were included (Beutler 2003; Brown 2006; Carpenter 2008; Johnson 2012; Lydecker 2010; Markowitz 2008; Rohde 2014).

Country: all included studies were conducted in the USA.

Participant characteristics: the studies were conducted with a variety of patient groups including veterans (Brown 2006; Lydecker 2010); adults (Beutler 2003; Carpenter 2008;Markowitz 2008); adolescents (Rohde 2014); and incarcerated females (Johnson 2012). The average age ranged from 16.4 to 48.8 years. All studies had samples with a majority of Caucasians (61% to 81%). The proportion of males ranged from 57% to 92%. One study enrolled females only (Johnson 2012).

While all study participants had comorbid DSM‐IV depression and substance use disorders there was considerable variability in the disorder types included. Several studies included participants with major depression disorders only (Brown 2006; Johnson 2012), two included participants with major depressive disorder/dysthymia only (Carpenter 2008; Lydecker 2010), one included dysthymia only (Markowitz 2008), and the others included any depression disorder (Beutler 2003; Rohde 2014).

Two studies included participants with any substance abuse or dependence (Johnson 2012; Rohde 2014), the two veterans studies included individuals with alcohol, cannabis and/or stimulant dependence (Brown 2006; Lydecker 2010), and the three remaining studies included alcohol abuse or dependence only (Markowitz 2008), stimulant (cocaine or methamphetamine) dependence only (Beutler 2003), or opiate dependence only (Carpenter 2008).

Settings: six studies were conducted in outpatient settings, including two in dual diagnosis treatment services for veterans (Brown 2006; Lydecker 2010). One study was conducted among incarcerated females (Johnson 2012).

Types of comparisons: no studies that compared psychological interventions (alone or combined with pharmacotherapy) with delayed treatment, treatment as usual or no treatment control conditions met inclusion criteria; nor did any studies comparing combined psychological and pharmacological interventions with other psychological interventions. All seven studies included for review compared different types of psychological treatments.

Four studies provided group therapy (Brown 2006; Carpenter 2008; Lydecker 2010;Rohde 2014), two individual therapy (Beutler 2003; Markowitz 2008), and one study provided group therapy followed by individual therapy (Johnson 2012).

Five studies compared CBT with other psychological interventions for comorbid depression and substance use disorders.

• Two studies compared 24 weeks (36 sessions) of group‐based integrated CBT (ICBT) versus Twelve Step Facilitation (TSF) in veterans accessing a dual diagnosis outpatient clinic with a DSM‐IV major depressive disorder and co‐occurring alcohol, cannabis and/or stimulant dependence (n = 90 Brown 2006; n = 206 Lydecker 2010).

• One study compared 24 weeks (24 sessions) of group‐based Behavioral Therapy for Depression in Drug Dependence (BTDD) versus 24 weeks of a group‐based structured Relaxation Intervention (REL) among 38 treatment‐seeking outpatients with DSM‐IV opiate dependence and major depressive disorder/dysthymia (Carpenter 2008).

• One study compared 1418 weeks (16‐20 sessions) of Cognitive Therapy (CT) with Narrative Therapy (NT) and Prescriptive Therapy (PT) among 40 patients seeking outpatient care with a DSM‐IV depression disorder and stimulant dependence (Beutler 2003).

• One study compared three methods for integrating CBT and Functional Family Therapy (FFT) in 170 adolescents with a comorbid DSM‐IV depression and substance use disorder seeking substance use treatment (Rohde 2014). Participants received 24 weeks (24 sessions) of either the adolescent Coping With Depression (CWD) followed by FFT (CWD/FFT), FFT followed by CWD (FFT/WD) or integrated CWD and FFT care (CWD+FFT).

Two studies compared IPT for depression (IPT‐D) with other psychological interventions.

• A pilot study compared 16 weeks (16‐18 sessions) of individual IPT‐D versus Brief Supportive Therapy (BST) in 26 adults seeking outpatient care with DSM‐IV dysthymia and secondary alcohol use disorder (Markowitz 2008).

• A second study compared eight weeks (24 sessions) of group IPT‐D with a psychoeducation group for co‐occurring mental health and substance use disorders (PSYCHOED) among 40 incarcerated females with DSM major depressive disorder and substance abuse or dependence. Both treatment groups also received three individual sessions at pre, mid and post the group intervention, as well as treatment as usual for substance use and mental health problems (Johnson 2012).

Outcomes

The depression and substance use outcomes reported differed across studies, as detailed in the Characteristics of included studies table. All but one study used a version of the Hamilton Depression Rating Scale (HDRS, Hamilton 1960), a clinician‐rated measure of the frequency of depression symptoms in the past week. The original version (Hamilton 1960) contained 17 items, but current versions range from 17 to 29 items (Hamilton 1980; Williams 1988). Items are scored on a 3‐or 5‐point scale. Total scores for depression are usually derived from the original 17 items on a 3‐point scale and range from zero to 51 points, with scores from 7 to 17 indicating mild depression, from 18 to 24 indicating moderate depression, and above 24 indicating severe depression.

Three studies also included the Beck Depression Inventory (BDI) (Beck 1961; Beck 1978; Beck 1996), a self‐report measure of the frequency of depression symptoms in the previous two weeks. Rohde 2014 used the Child Depression Rating Scale‐Revised (CDRS‐R, Poznanski 1995), a 17‐item clinician‐rated measure of depressive symptomatology in six‐ to 12‐year olds, which is also widely used in adolescents.

All studies used the Timeline Followback method (TLFB; Miller 1994; Sobell 1980; Sobell 1992) to assess substance use outcomes. This calendar‐based method assesses the frequency and quantity of substance use, including the number of days abstinent over a specified timeframe. Both the timeframe (30 to 90 days) and definition of substance use outcome measures (mean percentage days abstinent, proportion of days abstinent, square root of percentage of daily substance use) varied across studies. One study used the alcohol and drug index score of the Addiction Severity Index (ASI; Beutler 2003).

Definitions of treatment retention varied across studies. Four studies reported the mean number of treatment sessions attended (Brown 2006; Carpenter 2008; Lydecker 2010; Rohde 2014). Three studies reported the percentage of participants who dropped out of treatment but the definitions of dropout varied. Beutler 2003 defined treatment as missing three or more consecutive sessions out of the 20 sessions. Markowitz 2008 reported the percentage of participants who failed to complete treatment, but it is unclear how this was defined. Johnson 2012 defined treatment dropout as missing more than two out of 24 sessions by choice.

No studies reported information on any of the secondary outcome variables specified in the original Cochrane Review research protocol, including functioning, quality of life, anxiety symptoms or disorders or the global clinical severity of mental health disorders. No adverse effects linked to treatments delivered were reported.

Follow‐up

All studies reported post‐treatment outcomes. Three studies conducted three‐month follow‐ups (Brown 2006; Johnson 2012; Lydecker 2010), three conducted six‐month follow‐ups (Beutler 2003; Brown 2006; Lydecker 2010; Rohde 2014), and two conducted 12‐month follow‐ups (Lydecker 2010, Rohde 2014).

Excluded studies

We excluded 127 articles from 124 studies for not meeting the inclusion criteria. Overall 76 articles (from 75 studies) were excluded because they did not meet the depression criteria, 20 articles (from 19 studies) were excluded because they did not meet the substance use criteria, 23 articles (from 22 studies) were excluded because they did not meet both the depression and substance use criteria, two studies were excluded because they were not RCTs, one study was excluded because it focused depression and substance use in the context of Posttraumatic Stress Disorder and Trauma, and five studies were excluded because they did not meet psychotherapy criteria (three of these studies focused on the use of technology as an adjunct to therapy, not actual differences between therapies; and the remaining two studies randomised pharmacological treatments, without randomising psychological therapies) See Figure 1 for further details.

Allocation

Random sequence generation: all of the included studies were described as RCTs. Five studies were judged at low risk of selection bias: one study used concealed random draws (Beutler 2003) and four used independent computer‐based randomisation to generate the sequence allocation (Carpenter 2008; Johnson 2012; Markowitz 2008; Rohde 2014). Two studies used sequential allocation of consecutively‐admitted patients (Brown 2006; Lydecker 2010) and were judged to be at high risk of selection bias due to the use of a non‐random sequence generation process. All but one study (Beutler 2003), checked the success of randomisation by comparing the group allocations on baseline variables.

Allocation concealment: five studies were judged at low risk of bias ( Beutler 2003;Carpenter 2008; ,Johnson 2012; Markowitz 2008; Rohde 2014). Two studies were judged at high risk of bias (Brown 2006, Lydecker 2010) because a consecutive allocation was used.

Blinding

Performance bias: no study blinded therapists to the treatment allocation, as this is not possible in psychological trials. Only one study attempted to blind participants to their group allocation, and checked but did not report the success of this blinding process (Beutler 2003). Thus, all studies were rated at high risk of performance bias, due to the subjective nature of the depression and substance use outcomes. These studies were also rated at high risk for performance bias for the treatment retention/attendance variable, as despite the objective nature of this outcome, knowledge of the treatment allocation may influence participant's treatment engagement, participation and attendance.

Detection bias: five of the seven studies were at low risk of detection bias for interviewer‐rated outcomes, due to the use of outcome assessors blind to treatment allocation. Two studies which did not blind the outcome assessors were judged to be at high risk of detection bias (Carpenter 2008; Lydecker 2010). All four studies that reported self‐report outcomes were rated at unclear risk of detection bias (Beutler 2003; Carpenter 2008; Markowitz 2008; Rohde 2014). While participants were not blind to treatment allocation and self‐report measures may be impacted by self‐presentation bias or client insight, it is unlikely that any such risk of bias will vary by treatment condition. There was insufficient information to determine whether the individuals who collected information on treatment attendance/completion were blind to treatment allocation in all studies. However, this objective outcome measure is unlikely to be influenced by lack of blinding, and all seven studies were judged to be at low risk of detection bias on this outcome variable.

Incomplete outcome data

Treatment retention/attendance: less than 50% of participants completed treatment or attended half of the allocated treatment sessions (in at least one group) in four studies (Beutler 2003; Brown 2006; Carpenter 2008; Lydecker 2010). Between‐group differences in the number of sessions participants completed were reported in two studies (Beutler 2003; Rohde 2014).

Outcome assessments: high risk of attrition bias (defined as ≥50% in at least one group) was found in three studies at post‐treatment (Beutler 2003; Brown 2006; Carpenter 2008) and two studies at follow‐up (Beutler 2003; Brown 2006). One study in which one of the treatment groups had double the attrition rate at post‐treatment was considered high‐risk (Markowitz 2008). Attrition rates were typically between 20% and 30% in two other studies (Lydecker 2010; Rohde 2014) and one study among incarcerated females reported zero attrition (Johnson 2012).

Missing data/Intent‐to‐treat (ITT) analysis: three studies used the last observation carried forward (LOCF) method to manage missing data (Beutler 2003; Carpenter 2008; Markowitz 2008) and one imputed missing data, using Markov Chain Monte Carlo multiple imputation (Rohde 2014). All but one study usedITT analyses (Brown 2006). This study also excluded treatment dropouts (attended < 8 treatment sessions) and participants who missed two or more follow‐up assessments from the analyses and was judged to be at very high risk of attrition bias.

Selective reporting

Risk of selective reporting was unclear in all seven studies, as none of them had a published study protocol available. However, the results of all planned analyses using the outcome listed in the methods section are reported in the manuscripts. Two studies did not conduct or report the results of endpoint analyses at all follow‐up points. Trajectory, but not endpoint analyses were conducted (three‐month follow‐up (Brown 2006; Lydecker 2010) or nine‐month follow‐up (Lydecker 2010).

Other potential sources of bias

Fidelity ratings

All seven trials conducted fidelity ratings of the interventions, but only two were performed by independent raters (Johnson 2012; Markowitz 2008). Two studies controlled for therapist allegiance and contamination effects by using separate teams of therapists to deliver the different interventions (Beutler 2003; Markowitz 2008), but it was unclear if either study tested for contamination effects. One study routinely monitored therapist belief in and satisfaction with the treatment model selected and found this was comparable across the treatment groups (Beutler 2003). Risk of bias due to poor treatment fidelity was unclear in the other four studies, although Brown 2006 tested for therapist effects between and within treatment groups and reported it had no impact on the primary depression and substance use outcome variables. While all studies reported information on the number of treatment dropouts or treatment sessions attended, only one controlled for this variable in the analyses (Lydecker 2010).

Concurrent treatment

Only one study (Markowitz 2008) had concurrent pharmacotherapy as an exclusion criterion. None of the remaining studies reported on medication dose or adherence. For those that reported on pharmacotherapy, the main medication prescribed was antidepressants, with 22% to 98% of participants being prescribed an antidepressant during treatment (Brown 2006; Johnson 2012; Lydecker 2010; Rohde 2014), compared with only 1.1% to 2.7% of participants using medication to treat substance use (Johnson 2012; Lydecker 2010). The exception to this is the one study that included people with methadone‐maintained opiate dependence (Carpenter 2008). Few studies reported on the type or number of antidepressants prescribed, though when these were listed selective serotonin reuptake inhibitors (SSRls) or atypical antidepressants (Brown 2006) and lithium, aripiprazole, or quetiapine (Johnson 2012) were described as the most commonly used antidepressants. Only one study stratified the randomisation based on antidepressant use (Carpenter 2008). Five studies compared differences in antidepressant use between the treatment groups or controlled for antidepressant use in their analyses (Brown 2006; Carpenter 2008; Johnson 2012; Lydecker 2010; Rohde 2014), with two studies reporting significant differences between groups in use of antidepressants (Johnson 2012; Lydecker 2010).

One study had concurrent non‐study psychotherapy as an exclusion criterion (Markowitz 2008). A further two studies excluded people attending other psychotherapy beyond 12‐Step Community Addiction Meetings (Brown 2006; Lydecker 2010). It was unclear whether participants were permitted to attend adjunctive psychotherapy in two studies (Beutler 2003; Carpenter 2008), and the risk of other bias was high in the remaining two studies, as participants attended adjunctive therapy and attendance at this therapy was not consistent between the treatment arms (Johnson 2012; Rohde 2014).

1. Integrated Cognitive Behavioural Therapy (ICBT) versus Twelve Step Facilitation (TSF)

Two studies with 296 participants compared ICBT to TSF (Brown 2006; Lydecker 2010). Both studies recruited veterans experiencing alcohol, cannabinol and/or stimulant dependence and a major depressive diagnosis. Participants were predominately white males with a mean age of 48.3 (standard deviation (SD) = 7.8). Assessments were completed at post‐treatment and at three months, and six months later. Only one of the two studies assessed outcomes at 12 months post‐treatment (Lydecker 2010).

See: summary of findings Table for the main comparison.

Depression

See: Analysis 1.1 ;Analysis 1.2.

The meta‐analysis examining depression score at post‐treatment and at follow‐up consisted of 296 randomised participants.The meta‐analysis suggested that the ICBT group had higher depression scores, compared with the TSF group, at post‐treatment (mean difference (MD) 4.05, 95% confidence interval (CI) 1.43 to 6.66; 2 studies, 212 participants; Analysis 1.1), however, there did not appear to be any difference between the groups at six‐ to12‐month follow‐up (MD 1.53, 95% CI ‐1.73 to 4.79; 2 studies, 181 participants; Analysis 1.2; Figure 4) follow‐up. Heterogeneity was of no importance (I² = 0%, P = 0.91; I² = 0%, P = 0.44), respectively.

Figure 4

---

---

Comparison 1 Integrated CBT vs Twelve Step Facilitation, Outcome: Depression at 6‐12 months

Substance use

See: Analysis 1.3; Analysis 1.4.

The meta‐analysis consisting of two studies (296 participants) that compared ICBT to TSF, suggested no substantial difference between the groups in proportion of days abstinent in the past three months, as assessed by the Timeline Follow Back (TLFB), at immediately post‐treatment (MD ‐2.84, 95% CI ‐8.04 to 2.35; 2 studies, 220 participants; Analysis 1.3). At six‐to 12‐month follow‐up, the ICBT group experienced on average 10.76% more days abstinent (95% CI 3.10 to 18.42; 2 studies, 189 participants; Analysis 1.4) compared with the TSF group (Figure 5). Heterogeneity was of no importance in either analysis (I² = 0%; P = 0.39; I² = 0%;P = .65), respectively.

Figure 5

---

---

Comparison 1 Integrated CBT vs Twelve Step Facilitation, Outcome: Percentage of Days Abstinent at 6 to 12 months.

2. Interpersonal Psychotherapy for Depression (IPT‐D) versus other psychological treatments

Two studies with 64 participants compared IPT‐D with other psychological treatments, one examining Brief Supportive Psychotherapy, with a sample of patients experiencing major depression/dysthymia and secondary alcohol abuse/dependence drawn from a medical college (n = 26; Markowitz 2008); and the other examining psychoeducation, with a sample of participants experiencing major depression and a substance use disorder (including alcohol, cocaine, opiate, marijuana and sedatives/hypnotics) drawn from a female prison (n = 38; Johnson 2012). Both samples were predominantly female (63%‐ to ‐100%), mostly white, with an average age of 36.4 (SD = 9.9).

See: summary of findings Table 2.

Depression

See: Analysis 2.1; Analysis 2.2.

Two studies contributed 64 participants to end‐of‐treatment analyses (eight weeks/16 weeks), and one study (Johnson 2012) with 38 participants assessed three‐month post‐treatment effects. The common outcome measure used in both studies was the clinician‐rated Hamilton Rating Scale for Depression (HDRS) (17‐item and 24‐item scales were used). The end of treatment results, showed no important heterogeneity (I² = 0%; P = 0.54), and suggested participants receiving IPT‐D had lower depressive symptoms at post‐treatment than other treatment (Brief Supportive Psychotherapy/Psychoeducation; standardised mean difference (SMD) ‐0.54, 95% CI ‐1.04 to ‐0.04; I = 0%; 2 studies, 64 participants; Figure 6); however, there was no evidence that this effect was maintained at follow‐up (IPT‐D versus Psychoeducation; MD 3.80, 95% CI ‐3.83 to 11.43; 1 study, 38 participants).

Figure 6

---

---

Comparison 2 Interpersonal Psychotherapy for Depression (IPT‐D) vs Other Theraputic Interventions, Outcome: Depression at end of treatment.

3. Integrated Functional Family Therapy (FFT) andCoping With Depression (CWD) versus sequenced FFT‐CWD; Integrated FFT and CWD versus sequenced CWD‐FFT

One study with 170 adolescents compared integrated FFT and CWD to sequential FFT followed by CWD, or CWD followed by FFT (Rohde 2014). Adolescents were aged 13 to 18 years, were predominately white (61%) males (88%) experiencing a depressive disorder (54% major depression) and comorbid substance use disorder, which was predominately a cannabis use disorder (94%), comorbid with alcohol use disorder (65%). For ease of comparison the combined treatment was compared with the sequential treatment on primary outcomes. The first row in each analysis compares integrated FFT and CWD versus sequenced FFT‐CWD; and the second row compares integrated FFT and CWD versus sequenced CWD‐FFT. Assessments were completed at post‐treatment (20 weeks) and six and 12 months later.

4. Behavioral Therapy for Depression in Drug Dependence (BTDD) versus Relaxation Intervention (REL)

One study with 38 participants on methadone maintenance therapy for opiate dependence, who met criteria for a DSM‐IV depressive disorder, compared BTDD with REL (Carpenter 2008). Participants were predominately white (58%) males (82%) with a mean age of 40.1 years (SD = 10.7). Assessments were completed immediately post‐treatment (24 weeks).

5. Cognitive Therapy (CT) versus Narrative Therapy (NT) versus Prescriptive Therapy (PT)

One study with 40 patients experiencing comorbid DSM‐IV depression disorder and stimulant (cocaine or methamphetamine) dependence, compared CT, NT and PT (Beutler 2003). Participants were predominately white (75%) males (57%) with a mean age of 33.1 years (SD = 1.83). Assessments were completed immediately post‐treatment (20 weeks) and six months post‐treatment. Given the very small sample size, and the multiple comparisons that would be required to be conducted between treatment conditions, a narrative review of findings is presented below.