A systematic review and meta-analysis of psychological interventions for comorbid post-traumatic stress disorder and substance use disorder.

Background: The psychological treatment of comorbid post-traumatic stress disorder (PTSD) and substance use disorder (SUD) is clinically challenging, and outcomes are often poor. Objective: This paper describes a systematic review and meta-analysis which sought to establish the current efficacy for a number of established psychological approaches for adults and adolescents, in comparison to interventions for SUD alone, or other active approaches, following a pre-registered protocol. Method: This review followed PRISMA and Cochrane Collaboration guidelines. Data extraction and risk of bias judgements using Cochrane criteria were undertaken by all authors. Primary outcomes were PTSD severity and substance use post-treatment. The quality of findings was assessed using GRADE. Following a comprehensive search, conducted to 13 September 2021, 27 studies were included.
Results: We found a relatively high level of dropout across studies. In our main comparisons, we found no benefits for present-focused treatment approaches aimed at improving coping skills beyond those for SUD-only interventions. We found modest benefits for trauma-focused intervention plus SUD intervention post-treatment for PTSD (standardized mean difference (SMD) = -0.36, 95% confidence interval (CI) -0.64, -0.08), and at 6-13 months for PTSD (SMD = -0.48, 95% CI -0.81, -0.15) and alcohol use (SMD = -0.23, 95% CI -0.44, -0.02). There were no benefits for cognitive restructuring interventions as a group, but we found a modest effect for integrated cognitive behavioural therapy (ICBT) for PTSD post-treatment (SMD = -0.33, 95% CI -0.62, -0.04). There was evidence of some benefit for trauma-focused intervention over present-focused intervention for PTSD from a single study and for reduction in dropout for incentivized attendance for trauma-focused intervention from another single study. Most findings were of very low quality.
Conclusion: There is evidence that trauma-focused therapy and ICBT can improve PTSD for some individuals, but many patients do not fully engage with treatment and average treatment effects are modest. HIGHLIGHTS: For PTSD, evidence was strongest for trauma-focused CBT-based approaches, but effects were modest.There was little evidence of any added benefit on substance use, beyond that of standard addiction treatments, for any included intervention.Dropout from treatment was high.

Introduction

Post-traumatic stress disorder (PTSD) and substance use disorder (SUD) are both common disorders which cause significant functional impairment (e.g. Degenhardt et al., 2013; Grant et al., 2016; Kessler et al., 2017; World Health Organization, 2011). PTSD and SUD frequently co-occur. In a systematic review including 42 studies, Debell et al. (2014) found reported prevalence rates of comorbid alcohol abuse of 10–61% in those with PTSD, and comorbid rates of PTSD for those with alcohol misuse of 2–63%, with most studies reporting rates over 10%. Similarly high rates have been found for SUD both in epidemiological studies and among clinical samples. In an Australian national survey, nearly 6% of those with an SUD diagnosis met the diagnosis for PTSD; 6.5 times more likely than for those without an SUD. For specific substances, a third of those with an opiate use disorder had probable PTSD, with high rates also reported for those with sedative and amphetamine dependencies (Mills, Teeson, Ross, & Peters, 2006). High rates of PTSD diagnosis, of 20–42%, have been reported in a number of prevalence studies conducted with SUD service users (Gielen, Havermans, Tekelenburg, & Jansen, 2012; Najavits et al., 1998; Reynolds et al., 2005; Reynolds, Hinchliffe, Asamoah, & Kouimtsidis, 2011; Schäfer et al., 2010), with rates as high as 75% among military veterans (Schäfer & Lotzin, 2018; Seal et al., 2011).

For both disorders, well-established psychological and pharmacological interventions for management and treatment are available. For PTSD, psychological interventions include eye movement desensitization and reprocessing (EMDR), and various forms of trauma-focused cognitive behavioural therapy (CBT), including prolonged exposure (PE), cognitive therapy for PTSD, and cognitive processing therapy (CPT) (Lewis, Roberts, Andrew, Starling, & Bisson, 2020). There is uncertainty about the applicability of this literature for those with this comorbidity, as most randomized controlled trials (RCTs) of interventions for PTSD exclude participants based on substance use status (Leeman et al., 2017). For SUD, effective psychological interventions include CBT, motivational interviewing, behavioural couples therapy, relapse prevention, and contingency management (NICE, 2007; NIDA, 2020). Such interventions tend to be more effective when delivered alongside pharmacological substitutes (Jhanjee, 2014).

PTSD-SUD comorbidity poses many clinical challenges for treating clinicians. Individuals with PTSD-SUD present with a more severe clinical profile, often have significant case management needs and a higher incidence of childhood trauma, and tend to have poorer functioning and well-being, and inferior treatment outcomes and prognosis (Blakey et al., 2021; Roberts, Back, Mueser, & Murray, 2020; Schäfer & Lotzin, 2018; Schäfer & Najavits, 2007; Straus, Haller, Lyons, & Norman, 2018). Clinicians view this comorbidity as being substantially more difficult to treat than either disorder alone and there is often great uncertainty about whether and when to offer evidence-based trauma-focused intervention (Roberts et al., 2020). Most recent PTSD guidelines (e.g. American Psychological Association, 2017; Phoenix Australia, 2020; International Society for Traumatic Stress Studies, 2018; NICE, 2018) did not include scoping questions to address issues of comorbidity, and there are no widely accepted guidelines on how to manage and treat such individuals. The one exception is the US Department of Veterans Affairs/Department of Defense guidelines, which recommended that the presence of comorbidity should not prevent patients from receiving other guideline recommended interventions (VA/DoD, 2017).

The traditional model of treatment – often still practised in many places – is a sequential approach where addiction is treated first and trauma-focused therapy is only considered once abstinence or control is achieved (Roberts et al., 2020; van Dam, Vedel, Ehring, & Emmelkamp, 2012). This often necessitates referral to another clinician or service. For many individuals with PTSD-SUD comorbidity, drug and alcohol use is often at least partly a function of a perceived need to self-medicate to manage trauma symptoms and other negative emotional responses (Dvorak, Pearson, & Day, 2014; Kaysen, Stappenbeck, Rhew, & Simpson, 2014; Leeies, Pagura, Sareen, & Bolton, 2010). This is evidenced by the fact that PTSD symptom severity and symptom worsening often lead to increased substance use, while the reciprocal relationship of substance use being related to the development of subsequent PTSD symptoms is less strong (Langdon et al., 2016; Simpson, Stappenbeck, Luterek, Lehavot, & Kaysen, 2014; Tripp et al., 2020). This relationship between PTSD and SUD symptoms risks causing patients to drop out of treatment prematurely if their PTSD symptoms remain untreated, as they may find it more difficult to manage trauma symptoms, increasing the possibility that they might relapse or lose contact with services (Roberts et al., 2020; van Dam et al., 2012).

This understanding about the nature of the relationship between PTSD and SUD symptoms has led to the development of a number of integrated approaches where treatments for PTSD and SUD are delivered at the same time. Integrated treatment has been suggested as an effective treatment approach for PTSD and SUD that may be more advantageous than sequential treatment. Integrated treatment is delivered by one therapist or a team of therapists trained in both PTSD and SUD treatment. Care is delivered in a consistent manner, using a common disorder model. Several integrated treatment approaches are described in the literature. Differing terminology is sometimes used to describe these approaches (Hien et al., 2022); however, the distinction between present (or non-trauma-focused) and past (or trauma-focused) approaches is common and generally well understood. Present-focused approaches primarily seek to improve coping skills and self-compassion. Seeking Safety, a 25-session manualized intervention, which can be delivered in an individual or a group format (Najavits, 2002), is the most widely known and evaluated present-focused model. Other coping skills-focused models include the TARGET approach, which is also provided in groups of up to eight to nine sessions (Ford & Russo, 2006). A second type of present-focused interventions includes approaches that primarily aim to help individuals to challenge and manage negative PTSD and SUD-related cognitions and beliefs through cognitive behavioural strategies. The most commonly evaluated model of this type is integrated cognitive behavioural therapy (ICBT) (e.g. McGovern, Lambert-Harris, Alterman, Xie, & Meier, 2011). Past-focused approaches seek to actively process distressing trauma experiences by incorporating established evidence-based PTSD psychological interventions. The most well-established approach of this type is concurrent treatment of PTSD and substance use disorders using prolonged exposure (COPE) (Back et al., 2014). COPE offers PE alongside relapse prevention, delivered by the same therapist, usually over around 12 sessions.

Four systematic reviews of psychological interventions for PTSD and SUD have previously been published (Roberts, Roberts, Jones, & Bisson, 2016 – also summarized in Roberts, Roberts, Jones, & Bisson, 2015; Simpson, Lehavot, & Petrakis, 2017; Torchalla, Nosen, Rostam, & Allen, 2012; van Dam et al., 2012). Only one of these reviews (Roberts et al., 2015, 2016) involved a meta-analysis based on outcomes from RCTs. This review identified 14 studies, 13 of which included only adults and one including adolescents. The findings indicated some benefits for trauma-focused approaches for PTSD symptoms, when combined with treatment for SUD over control interventions, but effects were small. No benefits were found for present-focused approaches beyond those of control interventions. Dropout across studies evaluating both past- and present-focused approaches was high. Similar conclusions were drawn in the more recent narrative review of Simpson et al. (2017), although this review also noted that most of the included studies reported improvements in both experimental and control interventions. Since the publication of these reviews, a number of new trials have been completed. We therefore felt that it was timely to undertake an updated review of the literature. This work was also undertaken to support the development of expert treatment recommendations for the European Society for Traumatic Stress Studies (ESTSS).

Method

This review was based on the review by Roberts et al. (2016), which was previously led by the first author and followed guidelines outlined in the Cochrane Handbook (Higgins & Green, 2011). This protocol was registered with PROSPERO on 16 November 2020 and is available from: https://www.crd.york.ac.uk/prospero/display_record.php?ID = CRD42020207840.

Review questions

We established three overarching scoping questions which would be addressed through a number of separate comparisons. The comparisons below were not intended to be exclusive but were what we anticipated would be most meaningful based on our knowledge of the literature. The three scoping questions are numbered, with anticipated comparisons indicated by lettering.

For individuals with PTSD and comorbid SUD, do psychological treatments for PTSD or PTSD and SUD, when compared to treatment as usual (TAU) for SUD only, result in a clinically important reduction of PTSD and SUD symptoms, reduced presence of disorder, decreased dropout, or difference in reported adverse effects? Example comparisons:

Are present-focused treatments (also known as coping-based/non-trauma-focused treatments) plus TAU for SUD more effective than TAU for SUD only?

Are trauma-focused treatments plus TAU for SUD more effective than TAU for SUD only?

Are integrated cognitive restructuring-based interventions (without imaginal and in vivo exposure) for PTSD and SUD plus TAU for SUD more effective than TAU for SUD?

Is EMDR plus TAU for SUD more effective than TAU for SUD only?

For individuals with PTSD and comorbid SUD, do psychological treatments for PTSD and SUD, when compared to other psychological treatments for PTSD and SUD (head-to-head comparisons), result in a clinically important reduction of PTSD and SUD symptoms, reduced presence of disorder, decreased dropout, or difference in reported adverse effects? Example comparisons:

Are trauma-focused treatments more effective than present-focused treatments?

Are trauma-focused treatments more effective than integrated cognitive restructuring-based interventions (without imaginal and in vivo exposure)?

Are integrated cognitive restructuring-based interventions (without imaginal and in vivo exposure) more effective than present-focused treatments?

When compared to sequential treatments, do integrated treatments offered by one therapist OR simultaneous treatments for PTSD and SUD offered by different therapists result in a clinically important reduction of PTSD and SUD symptoms, reduced presence of disorder, decreased dropout, or difference in reported adverse effects?

Inclusion criteria

Participants

Participants were required to meet the formal diagnosis for PTSD using a structured clinical interview according to the International Classification of Diseases, 10th revision (ICD-10) or Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition (DSM-5), or subthreshold PTSD using predefined subthreshold criteria, which included re-experiencing symptoms. This decision was based on the fact that individuals with subthreshold PTSD experience similar levels of distress to those with full PTSD, and the high prevalence of subthreshold PTSD in SUD clinical populations (Jakupcak et al., 2007; Vujanovic, Smith, Green, Lane, & Schmitz, 2018). A minimum of 60% of participants were required to meet full PTSD diagnosis for a study to be included. Participants were also required to meet the formal diagnosis for SUD using a structured clinical interview according to ICD or DSM. Studies primarily aimed at evaluating PTSD and nicotine dependence were excluded. Studies were eligible for inclusion if a subgroup of participants in a study met the criteria for PTSD and SUD, and were randomly allocated across intervention, and subgroup data were made available. Five studies were included on this basis (Acosta et al., 2017; Frisman, Ford, Lin, Mallon, & Chang, 2008; Haller et al., 2016; Mueser et al., 2008; Possemato et al., 2019). We did not exclude on the basis of age but decided that studies including children and young people below the age of 18 years would be analysed and presented separately. Participants were not excluded on the basis of other comorbidity.

Type of study and intervention

To be considered for inclusion, studies had to be RCTs or cluster randomized trials, including studies with cross-over designs, evaluating one or more psychological intervention, in comparison to a control condition or an alternative psychological intervention. Interventions that were primarily pharmacological and interventions that were not based on a defined psychological model (e.g. yoga, acupuncture, neurofeedback, or transcranial magnetic stimulation) were not included. Control conditions included TAU for SUD only, waiting list or no treatment, attention control interventions, and other psychological treatments. Studies using a pharmacological intervention only as a control were excluded unless the intervention group also received this pharmacological intervention in addition to a psychological intervention. Studies were excluded if they did not evaluate either PTSD symptoms or substance use outcomes. Studies were not excluded on the basis of sample size or publication status. Controlled studies that did not allocate participants randomly were excluded. Studies not published in English were excluded. Unpublished studies were eligible for inclusion.

Outcomes

The primary outcomes were PTSD severity (giving primacy to clinician-administered outcomes, followed by self-report measures) and SUD severity (giving primacy to clinician-administered outcomes, followed by biological markers, followed by self-report measures). For SUD severity, alcohol and drug outcomes were considered separately. The primary outcome point was post-treatment. Additional outcome points were: 3–5 months post-treatment, 6–13 months post-treatment, and 13 months and beyond post-treatment. Other outcomes were PTSD and SUD diagnosis post-treatment, dropout/leaving treatment prematurely (using study-defined criteria), and adverse effects.

Searches

Review searches were based on an update to the search undertaken by Roberts et al. (2016). Searches were conducted from 2015 to the present date and were limited to English-language publications. Search sources included MEDLINE, PsycINFO, Embase, Cochrane CENTRAL, PTSDPubs, PTSD-Repository (https://ptsd-va.data.socrata.com/), ClinicalTrials.gov and the WHO International Clinical Trials Registry Platform. Searches included key words and terms related to the terms PTSD, SUD, alcohol use disorder, and RCTs. See Appendix 1 in the Supplementary material for the search stream applied to MEDLINE, PsycINFO, and Embase. We included studies identified and previously included in Roberts et al. (2016). However, in view of the fact that inclusion criteria related to PTSD diagnosis were slightly revised, we re-reviewed studies from this review that were previously excluded because of these criteria. We also reviewed the references of included studies and relevant systematic reviews for studies that we may have missed.

Procedure

Titles and abstracts were independently screened by two of the authors, as was the full manuscript of any study that either reviewer thought was potentially eligible for inclusion. Any differences of opinion about whether a study met the review inclusion criteria were discussed among all three authors. All three authors undertook data extraction and risk-of-bias judgements using a pre-established data-extraction template, which was then entered into Review Manager (RevMan) version 5.4.1 (RevMan, 2020). Extracted data included: study authors and publication date; study design; study setting; recruitment method; participant type; inclusion and exclusion criteria; baseline characteristics (e.g. number of participants diagnosed with full vs partial PTSD; type of drug/alcohol abuse); demographic features (e.g. age, gender, ethnicity); country where study was undertaken; all baseline and outcome measures; number of individuals screened, recruited, and randomized per arm; description of all experimental and control interventions, including length of treatment and number of sessions offered; outcome data for all review-identified outcomes for each time-point available; mean number of sessions attended; funding source; and key conclusions reached by the study authors.

Risk of bias was assessed using the risk of bias tool in the Cochrane Handbook for Systematic Reviews of Interventions, focusing on the following domains: sequence generation, allocation concealment, blinding of outcome assessors, incomplete outcome data, selective reporting, and other bias (Higgins & Green, 2011). Ordinarily, it is not possible to blind participants and therapists to the type of intervention allocation, so this domain was not assessed. Again, differences of opinion were settled following discussion among all authors.

Main and subgroup analyses

To address the review questions identified in Section 2.1, we grouped and analysed studies separately according to whether they were past focused or present focused and according to the model of intervention (e.g. coping focused, cognitive intervention). Comparisons considered under scoping question 1 included studies where the comparator was intervention for SUD only or SUD plus a minimal attention intervention. When an intervention had been evaluated in both an individual- and a group-based format, data were entered into the same comparison but grouped separately to enable possible differences in outcome to be examined.

Strategy for data synthesis

Data were synthesized using meta-analysis where sufficiently similar interventions and comparators had been investigated, using the comparisons described above. The synthesis was undertaken using RevMan 5.4.1 software (RevMan, 2020). Owing to the degree of clinical or statistical heterogeneity, data were pooled using a random-effects model. Analyses were evaluated using the standardized mean difference (SMD) for continuous outcomes and relative risk (RR) for dichotomous outcomes, and presented with 95% confidence intervals (CIs). We used intention-to-treat data where these were available. Completer-only data were included when these were the only data source available. Where data were incomplete, we made attempts to obtain additional data from the study authors. Heterogeneity was explored by observing the I2 statistic and chi-squared (χ2) test. We specified that we would consider the following possible causes of clinical heterogeneity for exploration of our post-treatment primary outcomes if sufficient data permitted: specified treatment intervention model (e.g. COPE, ICBT, Seeking Safety); participant subgroup (e.g. veterans, victims of sexual, physical, and domestic violence, childhood trauma); specific substances of misuse (e.g. alcohol, opioids, cocaine, amphetamines); and treatment setting (e.g. residential or inpatient treatment, outpatient treatment). We determined that publication bias would be investigated by generating funnel plots and examining them for signs of asymmetry if sufficient studies (≥ 10) were included in a comparison.

The quality of the findings was evaluated according to Grading of Recommendations Assessment, Development and Evaluation (GRADE) (e.g. Guyatt, Oxman, Schünemann, Tugwell, & Knottnerus, 2011; Langendam et al., 2013). GRADE has four levels of evidence: high, moderate, low, and very low. Following the International Society for Traumatic Stress Studies (ISTSS) PTSD Guidelines (Bisson et al., 2019), we decided to set thresholds for the clinical importance of findings for our primary outcomes. A threshold effect size greater than or equal to 0.4 was required for PTSD severity and 0.3 for SUD severity to indicate clinical importance for interventions compared against an SUD only/TAU comparator. We set a lower threshold for SUD severity because we thought that it was likely that participants would normally have received some SUD-based intervention prior to study enrolment, which would mean that it would be harder to detect symptom improvement. A threshold greater than or equal 0.2 was set for interventions compared against another active psychological intervention aimed at treating PTSD and SUD. We decided at the outset of the review that we would interpret these thresholds with caution, given that, to our knowledge, benchmarks for clinically important improvement have not been described in this population previously, and small effects might still have important clinical and functional significance.

Results

The final searches took place on 13 September 2021. After deduplication, 2219 new citations were identified, 66 references were read in full by two of the authors, and 13 additional studies were included, alongside the 14 studies identified previously (Roberts et al., 2015, 2016), giving 27 studies with a total of 2849 participants. Figure 1 shows a flow diagram of the selection process. Characteristics of the included studies can be found in Table 1. Twenty-six studies were conducted in adult populations and one in an adolescent population (Najavits, Gallop, & Weiss, 2006). Most studies were conducted in the USA, with two studies conducted in Australia (Mills et al., 2012; Sannibale et al., 2013) and one study in Germany (Schäfer et al., 2019). Nine studies with military veterans had mostly male participants (Acosta et al., 2017; Back et al., 2019; Boden et al., 2014; Capone et al., 2018; Haller et al., 2016; Kehle-Forbes et al., 2019; Najavits, Krinsley, Waring, Gallagher, & Skidmore, 2018; Norman et al., 2019; Possemato et al., 2019); samples were female only in six studies all evaluating Seeking Safety (Hien et al., 2009; Hien, Cohen, Miele, Litt, & Capstick, 2004; Myers, Browne, & Norman, 2015; Najavits et al., 2006; Schäfer et al., 2019; Zlotnick, Johnson, & Najavits, 2009) and mixed in the remaining studies. The majority of non-veteran studies recruited from community services, mainly for those with addiction problems. One study recruited from a residential addiction setting (Coffey et al., 2016) and one study in a prison setting (Zlotnick et al., 2009). Most studies were ethnically diverse, and where reported, most non-veterans’ studies included participants who were mostly from lower socio-economic groups. Studies invariably reported SUD type based on self-reported use at clinical interview. Six studies included only individuals with alcohol use disorder (AUD) (Coffey, Stasiewicz, Hughes, & Brimo, 2006; Foa et al., 2013; Myers et al., 2015; Norman et al., 2019; Sannibale et al., 2013; Stappenbeck et al., 2015), with an additional two studies recruiting on the basis of hazardous drinking, where we obtained subset data for those diagnosed with AUD (Acosta et al., 2017; Possemato et al., 2019). One study recruited exclusively based on opiate use (Schacht, Brooner, King, Kidorf, & Peirce, 2017). All other studies reported inclusion on the basis of polydrug abuse/dependence, or alcohol or polydrug abuse/dependence.

Figure 1.

Flow diagram of the systematic review. RCT, randomized controlled trial; PTSD, post-traumatic stress disorder; SUD, substance use disorder.

Table 1.

Characteristics of the included studies.

Source (country)	Participants and setting	SUD type	Age (years), mean (SD)	Gender	PTSD, alcohol, and substance use outcome measures	Experimental intervention	Control/ comparison intervention(s)	Number of participants, treatment/ control
Acosta et al. (2017) (USA)	Military veterans with PTSD and problematic alcohol misuse recruited from primary care	Alcohol misuse. A subset of participants met diagnosis for AUD	34 (8.1)	93% male; 7% female	PCL-M; TLFB	Web-based self-management programme based on CBT	Primary care TAU	29/32a
Back et al. (2019) (USA)	Treatment-seeking veterans primarily recruited from newspaper and internet advertisements	Alcohol and substance use disorder	40.4 (10.7)	90.1% male; 9.9% female	CAPS, PCL-M; TLFB, ASI	COPE	Response prevention	54/27
Boden et al. (2014) (USA)	Military veterans recruited from a Veterans Affairs outpatient substance use disorder clinic	Polydrug use	54.0 (9.6)	100% male	IES-R; ASI	Group-based Seeking Safety + TAU	Group-based TAU for SUD	54/55
Capone et al. (2018) (USA)	Military veterans recruited from speciality PTSD, substance use, and returning veteran clinics	Polydrug use	34.2 (8.9)	95.5% male; 4.5% female	CAPS; ASI, TLFB, toxicology	Individual and group integrated CBT for PTSD and SUD + TAU	TAU for SUD	21/23
Coffey et al. (2006) (USA)	Laboratory based: participants recruited from outpatient SUD treatment programmes	Alcohol dependence	37.5 (8.0)	33% male; 67% female	IES-R	Individual trauma-focused exposure therapy + TAU	Imagery-based relaxation + TAU	16/15
Coffey et al. (2016) (USA)	Participants recruited from residential SUD treatment facility	Alcohol dependence and polydrug use	33.7 (10.3)	53.3% male; 46.7% female	CAPS, IES-R; TLFB	Individual trauma-focused exposure therapy + TAU	Healthy lifestyle sessions + TAU	82/38
Foa et al. (2013) (USA)	Treatment-seeking participants recruited by professional referral and advertisement	Alcohol dependence	42.7 (9.7)	65.5% male; 34.5% female	PSS-I; TLFB	Individual PE + supportive counsellinga	Supportive counsellingb	80/85
Frisman et al. (2008) (USA)	Participants recruited from outpatient SUD clinics	Polydrug use	37.5 (8.4)	39.0% male; 61% female	GAIN subscales for substance use frequency, percentage drinking to intoxication, percentage using any drugs, and percentage abusing drugs or alcohol	Group-based coping skill-focused therapy + trauma-sensitive usual care	Trauma-sensitive usual substance use care	119/63a
Haller et al. (2016) (USA)	Veterans recruited from referrals to an outpatient dual-diagnosis treatment programme	Alcohol or substance dependence	47.26 (11.97)	89% male; 11% female	PCL-C; TLFB	Individual cognitive processing therapy for PTSD and SUD	Individual CBT for SUD and depression only	51/47a
Hien et al. (2004) (USA)	Outpatients recruited through substance use treatment programmes and advertisement	Polydrug use	36.2 (9.0)	100% female	CAPS, IES-R; SUI, CGI	Individual Seeking Safety + TAU	Relapse prevention + TAU	41/34c
Hien et al. (2009) (USA)	Outpatients recruited from community-based substance abuse treatment programmes	Polydrug use	39.2 (9.2)	100% female	CAPS, PSS-SR; SUI, toxicology	Group-based Seeking Safety + TAU	Group-based women’s health education + TAU	176/177
Kehle-Forbes et al. (2019) (USA)	Veterans recruited through a variety of channels, including provider referrals and advertisements in veterans’ medical centres	Alcohol or drug abuse. Sample was mainly alcohol dependent	44.1 (16.4)	92.3% male; 7.6% female	PSS-I; TLFB	Phased delivery of motivation enhancement therapy followed by PE	Integrated delivery of motivation enhancement therapy and PE at the same time	88/95
McGovern et al. (2011) (USA)	Participants recruited from community intensive outpatient or methadone maintenance programmes	Polydrug use	37.7 (10.7)	43.4% male; 56.6% female	CAPS; ASI, toxicology	Individual integrated CBT for PTSD and SUD + TAU	Individual addiction counselling + TAU	32/21
McGovern et al. (2015) (USA)	Participants recruited from addiction treatment agencies	Polydrug use, primarily mixed alcohol and opioid use	35.3 (10.4)	40.7% male; 59.3% female	CAPS, PCL-C; TLFB, ASI, toxicology	Individual integrated CBT for PTSD and SUD + TAU	Individual addiction counselling + TAU; TAU only	73/75/73
Mills et al. (2012) (Australia)	Participants recruited from substance use treatment services, advertisement, and practitioner referrals	Substance-dependent polydrug users	33.7 (7.9)	37.9% male; 62.1% female	CAPS; CIDI	COPE + TAU	TAU	55/48
Mueser et al. (2008) (USA)	Individuals with severe mental illness recruited from community mental health centres	Polydrug use. Participants met diagnosis for SUD but not substance dependence	44.2 (10.6)	20.5% male; 79.5% female	CAPS	Individual non-trauma-focused CBT for PTSD	TAU	17/27a
Myers et al. (2015) (USA)	Participants were victims of IPV recruited through flyers in community agencies serving IPV victims and in primary care and psychiatry clinics	AUD	42.8 (9.3)	100% female	CAPS, PCL-C; TLFB	Adapted group-based Seeking Safety + TAU	12-Step supportive group	31/9
Najavits et al. (2006) (UA)	Outpatient adolescents recruited from local clinics, hospitals, and schools, and through clinicians and posted flyers	Polydrug use. Most participants met diagnosis for substance dependence	16.1 (1.2)	100% female	TSCC; PEI	Individual Seeking Safety + TAU	TAU	18/15
Najavits et al. (2018) (USA)	Veterans recruited through clinicians, flyers, and word of mouth	Polydrug use. Most participants met diagnosis for substance dependence	48.8 (10.8)	73.1% male; 26.9% female	PCL; ASI, toxicology	Individual Creating Change + TAU	Individual Seeking Safety + TAU	26/26
Norman et al. (2019) (USA)	Veterans recruited through veterans’ mental health service clinics	AUD	41.6 (12.6)	89.9% male; 10.1% female	CAPS-5; TLFB	COPE	Seeking Safety	63/56
Possemato et al. (2019) (USA)	Military veterans with PTSD and problematic alcohol misuse recruited from primary care	Alcohol misuse. A subset of participants met diagnosis for AUD	39	93% male; 7% female	PCL-M; TLFB	Peer-supported web-based self-management programme based on CBT	Web-based self-management programme based on CBT	4/6a
Ruglass et al. (2017) (USA)	Outpatients recruited through advertisements and clinic referrals	Polydrug and alcohol dependence	44.6 (9.3)	63.6% male; 36.4% female	CAPS, MPSS-SR; ASI, SUI	COPE + TAU for SUD	Relapse prevention; activity monitoring	39/43d
Sannibale et al. (2013) (Australia)	Participants recruited from a range of services and seen on an outpatient basis	AUD	41.2 (11.9)	47% male; 53% female	CAPS, PDS; TLFB, SADQ-C	Individual integrated trauma-focused CBT for PTSD and AUD	CBT for AUD + supportive counselling	33/29
Schacht et al. (2017) (USA)	Participants recruited from an outpatient methadone maintenance clinic	Opioid users with PTSD	37.4 (11.3)	21% male; 79% female	CAPS; ASI	Prolonged exposure + contingency incentives + TAU for SUD	Prolonged exposure + TAU for SUD	28/30
Schäfer et al. (2019) (Germany)	Participants recruited via substance abuse and other psychosocial counselling agencies, substance abuse and mental health clinics, psychotherapists in private practice and in the community	Alcohol and polydrug use	40.9 (11.4)	100% female	PSS-I, PDS; ASI, substance-free days in the past month	Seeking Safety + TAU	Relapse prevention + TAU; TAU only	115/111/117
Stappenbeck et al. (2015) (USA)	Outpatients recruited through newspaper adverts and flyers	Alcohol dependence	44.3 (11.5)	51.3% male; 48.7% female	Adapted version of PCL-C; daily monitoring of number of standard drinks consumed per day	Cognitive restructuring	Experiential acceptance; attention control	31/27/20
Zlotnick et al. (2009) (USA)	Treatment was conducted in a minimum security prison. Participants recruited from a voluntary residential substance abuse treatment programme	Polydrug users. 88% met criteria for alcohol dependence prior to imprisonment	34.6 (7.4)	100% female	CAPS, TSC-40; TLFB, ASI	Group-based SeekingSafety + TAU	TAU	27/22

Data were obtained for a subset of participants, from these studies, who met the inclusion criteria for the review.

Study used a 2 × 2 design. Participants received experimental and control psychological interventions in combination with naltrexone or a placebo. Data were combined for the psychological intervention arms.

We did not include data from a TAU arm, as participants in this group were non-randomized.

We did not include data from the activity monitoring arm, owing to significant baseline differences from the other two arms for PTSD and depression.

ASI, Addiction Severity Index; AUD, alcohol use disorder; CAPS, Clinician Administered PTSD Scale; CAPS-5, Clinician Administered PTSD Scale for DSM-5; CBT, cognitive behavioural therapy; CGI, Clinical Global Interview; CIDI, Composite International Diagnostic Interview; COPE, concurrent treatment of PTSD and SUD using PE; GAIN, Global Appraisal of Individual Needs; IES-R, Impact of Events Scale – Revised; IPV, interpersonal violence; MPSS-SR, Modified PTSD Symptom Scale Self-Report; PCL, PTSD Checklist; PCL-C, PTSD Checklist – Civilian Version; PCL-M, PTSD Checklist – Military Version; PDS, Posttraumatic Diagnostic Scale; PE, prolonged exposure; PEI, Personal Experiences Inventory; PSS-I, PTSD Symptom Scale – Interview Version; PSS-SR, PTSD Symptom Scale – Self-Report Version; PTSD, post-traumatic stress disorder; SADQ-C, Severity of Alcohol Dependence Questionnaire; SUD, substance use disorder; SUI, Substance Use Inventory; TAU, treatment as usual; TLFB, Timeline Follow Back; TSCC, Trauma Symptom Checklist for Children; TSC-40, Trauma Symptom Checklist-40.

Synthesis of results for adults

Findings of the meta-analyses for all available outcomes are provided in Table 2.

Table 2.

Summary of meta-analysis of results for each comparison for an experimental psychological intervention vs a substance use disorder (SUD)-based or treatment as usual (TAU) control.

Comparison outcome	Follow-up point	Contributing studies	Studies (n)	Sample (n)	RR (95% CI)	SMD (95% CI)	Grade rating
Present-focused treatments + TAU for SUD vs TAU for SUD only
PTSD severity	Post-treatment	Boden et al. (2014), Hien et al. (2004, 2009), Myers et al. (2015), Schäfer et al. (2019), Zlotnick et al. (2009)	6	814		−0.02 (−0.16, 0.12); I2 = 0%	Low
	3–5 months	Boden et al. (2014), Hien et al. (2004, 2009), Myers et al. (2015), Schäfer et al. (2019), Zlotnick et al. (2009)	6	800		−0.01 (−0.15, 0.13); I2 = 0%	Low
	6–13 months	Hien et al. (2004, 2009), Myers et al. (2015), Schäfer et al. (2019), Zlotnick et al. (2009)	5	726		0.04 (−0.18, 0.26); I2 = 40%	Low
PTSD diagnosis	Post-treatment	Myers et al. (2015), Zlotnick et al. (2009)	2	77	1.01 (0.66, 1.54); I2 = 0%		Very low
Alcohol use	Post-treatment	Boden et al. (2014), Myers et al. (2015), Schäfer et al. (2019)	3	337		−0.21 (−0.67, 0.25); I2 = 66%	Very low
	3–5 months	Boden et al. (2014), Myers et al. (2015), Schäfer et al. (2019), Zlotnick et al. (2009)	4	372		−0.11 (−0.31–0.10); I2 = 0%	Very low
	6–13 months	Myers et al. (2015), Schäfer et al. (2019), Zlotnick et al. (2009)	3	298		0.44 (−0.43–0.51); I2 = 58%	Very low
Substance use	Post-treatment	Boden et al. (2014), Hien et al. (2004, 2009), Myers et al. (2015), Schäfer et al. (2019)	5	765		0.13 (−0.42, 0.16); I2 = 67%	Very low
	3–5 months	Boden et al. (2014), Frisman et al. (2008), Hien et al. (2004, 2009), Myers et al. (2015), Schäfer et al. (2019), Zlotnick et al. (2009)	7	925		−0.02 (−0.19–0.15); I2 = 29%	Low
	6–13 months	Frisman et al. (2008), Hien et al. (2004, 2009), Myers et al. (2015), Schäfer et al. (2019), Zlotnick et al. (2009)	6	874		−0.00 (−0.13, 0.13); I2 = 0%	Low
Dropout		Hien et al. (2004, 2009), Myers et al. (2015), Schäfer et al. (2019)	4	694	1.06 (0.92, 1.22); I2 = 0		Very low
Adverse events		Hien et al. (2009)	1	353	1.03 (0.71, 1.50)		Very low
Trauma-focused treatments + TAU for SUD vs TAU for SUD only
PTSD severity	Post-treatment	Back et al. (2019), Coffey et al. (2006, 2016), Foa et al. (2013), Mills et al. (2012), Ruglass et al. (2017), Sannibale et al. (2013)	7	544		−0.36 (−0.64, −0.08)*; I2 = 54%	Very low
	3–5 months	Back et al. (2019), Coffey et al. (2016), Ruglass et al. (2017)	3	213		−0.97 (−2.10, 0.16); I2 = 91%	Very low
	6–13 months	Back et al. (2019), Coffey et al. (2016), Foa et al. (2013), Mills et al. (2012), Sannibale et al. (2013)	5	469		−0.48 (−0.81, −0.15)*; I2 = 62%	Very low
PTSD diagnosis	Post-treatment	Back et al. (2019), Coffey et al. (2016)	2	201	0.62 (0.46, 0.84)*; I2 = 30%		Very low
Alcohol use	Post-treatment	Back et al. (2019), Foa et al. (2013), Sannibale et al. (2013)	3	234		0.05 (−0.21, 0.31); I2 = 0%	Very low
	3–5 months	Back et al. (2019), Coffey et al. (2016)	2	153		−0.07 (−0.41, 0.27); I2 = 0%	Very low
	6–13 months	Back et al. (2019), Coffey et al. (2016), Foa et al. (2013), Sannibale et al. (2013)	4	363		−0.23 (−0.44, −0.02)*; I2 = 0%	Very low
Substance use	Post-treatment	Back et al. (2019), Mills et al. (2012), Ruglass et al. (2017)	3	170		0.16 ( −0.63, 0.95); I2 = 81%	Very low
	3–5 months	Back et al. (2019), Coffey et al. (2016), Ruglass et al. (2017)	3	205		0.09 (−0.28. 0.46); I2 = 29%	Very low
	6–13 months	Back et al. (2019), Coffey et al. (2016), Mills et al. (2012)	3	256		0.01 (−0.35, 0.38); I2 = 44%	Very low
Dropout		Back et al. (2019), Coffey et al. (2006, 2016), Foa et al. (2013), Sannibale et al. (2013)	5	459	0.86 (0.71, 1.04); I2 = 48%		Very low
Adverse events		Back et al. (2019), Foa et al. (2013), Mills et al. (2012), Ruglass et al. (2017), Sannibale et al. (2013)	5	493	0.85 (0.40, 1.82); I2 = 0%		Very low
ICBT + TAU for SUD vs TAU for SUD only
PTSD severity	Post-treatment	Capone et al. (2018), Haller et al. (2016), McGovern et al. (2011, 2015)	4	349		−0.24 (−0.51, 0.03); I2 = 33%	Very low
	3–5 months	Capone et al. (2018), Haller et al. (2016), McGovern et al. (2011)	3	156		−0.12 (−0.44, 0.20); I2 = 0%	Very low
	6–13 months	McGovern et al. (2011)	1	43		0.07 (0.94, 1.30)	Very low
PTSD diagnosis	Post-treatment	McGovern et al. (2011)	1	53	0.94 (0.68, 1.30)		Very low
Alcohol use	Post-treatment	Capone et al. (2018), Haller et al. (2016), McGovern et al. (2011, 2015)	4	352		0.02 (−0.19, 0.24); I2 = 0	Very low
	3–5 months	Capone et al. (2018), Haller et al. (2016), McGovern et al. (2011)	3	165		−0.10 (−0.41, 0.21); I2 = 0	Very low
	6–13 months	Haller et al. (2016)	1	75		0.19 (−0.27, 0.64)	Very low
Substance use	Post-treatment	Capone et al. (2018), Haller et al. (2016), McGovern et al. (2011, 2015)	4	352		−0.08 (−0.30, –0.13); I2 = 0	Very low
	3–5 months	Capone et al. (2018), Haller et al. (2016), McGovern et al. (2011)	3	165		−0.01 (−0.31, 0.30); I2 = 0	Very low
	6–13 months	Haller et al. (2016)	1	75		0.19 (−0.27, 0.64)	Very low
Web-based CBT + TAU vs TAU
PTSD severity	Post-treatment	Acosta et al. (2017)	1	49		−0.07 (−0.63, 0.49)	Very low
	3–5 months	Acosta et al. (2017)	1	49		−0.29 (−0.85, 0.28)	Very low
Alcohol use	Post-treatment	Acosta et al. (2017)	1	61		0.11 (−0.39, 0.61)	Very low
	3–5 months	Acosta et al. (2017)	1	61		0.19 (−0.31, 0.69)	Very low
CBT for PTSD vs TAU only
PTSD severity	Post-treatment	Mueser et al. (2008)	1	44		−0.22 (−0.83, 0.39)	Very low
	3–5 months	Mueser et al. (2008)	1	44		−0.25 (−0.86, 0.36)	Very low

CBT, cognitive behavioural therapy; CI, confidence interval; ICBT, integrated cognitive behavioural therapy; PTSD, post-traumatic stress disorder; RR, relative risk; SMD, standardized mean difference.

RR (of categorical data): RR = 1, same as control; RR < 1, intervention better; RR > 1, control better.

SMD (of continuous symptom score): SMD = 0, no difference between intervention and control; SMD < 0, intervention better; SMD > 0, control better.

*Statistically significant difference at the p < 0.05 level.

Psychological intervention for PTSD, or PTSD and SUD versus intervention for SUD only

Present-focused interventions

Present-focused therapies were compared against SUD-only interventions in seven studies (Boden et al., 2014; Frisman et al., 2008; Hien et al., 2004, 2009; Myers et al., 2015; Schäfer et al., 2019; Zlotnick et al., 2009). In all but one of these studies, the present-focused approach was based on Seeking Safety, with the number of sessions offered ranging from eight to 25 sessions. Seeking Safety was delivered through groups, apart from in Hien et al. (2004). The other present-focused approach evaluated was TARGET (Frisman et al., 2008), a group-based coping skills-focused therapy. Where reported, participants attended between 36% (Frisman et al., 2008) and 62% (Zlotnick et al., 2009) of present-focused therapy sessions available to them. Combining data from four studies (Hien et al., 2004, 2009; Myers et al., 2015; Schäfer et al., 2019), 51% of participants receiving present-focused intervention were considered completers, based on study-specified definitions. Hien et al. (2004) was a three-arm trial comparing Seeking Safety, relapse prevention, and TAU for SUD only. Participants in the TAU arm were not recruited randomly and data from this arm were not included in meta-analysis. In meta-analysis, we found no benefit beyond that of intervention for SUD only at any time-point for PTSD severity, alcohol use, or substance use. No differences were observed for treatment dropout or adverse effects. Adverse events were not reported for Frisman et al. (2008) or Zlotnick et al. (2009).

All post-treatment analyses of present-focused therapies were based on evaluations of Seeking Safety, and other prespecified study characteristics were too similar for us to consider other meaningful subgroup analyses.

Trauma-focused interventions

Seven studies compared trauma-focused interventions plus SUD intervention with SUD-only interventions (Back et al., 2019; Coffey et al., 2006, 2016; Foa et al., 2013; Mills et al., 2012; Ruglass et al., 2017; Sannibale et al., 2013). The trauma-focused approach in all of these studies was based on PE. Three studies were based specifically on the COPE model, where relapse prevention was offered alongside PE (Back et al., 2019; Mills et al., 2012; Ruglass et al., 2017). The number of available sessions ranged from six to 18. Trauma-focused therapy was delivered individually in all studies. Where reported, participants attended between 35.2% (Foa et al., 2013) and 73.3% (Back et al., 2019) of trauma-focused therapy sessions available to them. Combining data from five studies (Back et al., 2019; Coffey et al., 2006, 2016; Foa et al., 2013; Sannibale et al., 2013), 62% of participants receiving trauma-focused intervention were considered to be treatment completers, based on study-specified definitions. Foa et al. (2013) used a 2 × 2 study design to test the benefits of PE and naltrexone against supportive addiction counselling and placebo medication. As reported in Roberts et al. (2015), we combined the two arms for those receiving PE plus addiction counselling to allow comparison with the two arms receiving supportive addiction counselling. Following Roberts et al. (2015), we also combined the two active treatment arms from Coffey et al. (2016), where PE and PE plus motivational interviewing were compared against an attention control intervention. The trial conducted by Ruglass et al. (2017) had three arms: COPE, relapse prevention, and a waiting list group. There were significant baseline differences between the waiting list arm and the other two arms for PTSD and depression, and we therefore excluded the waiting list arm from our analysis.

In meta-analysis, we found a significant difference in favour of trauma-focused intervention plus SUD intervention for PTSD severity post-treatment (K = 7; N = 544; SMD −0.36, 95% CI −0.64, −0.08; I2 = 54%; GRADE very low). This effect did not reach our threshold for clinical importance. Findings based on three studies were not significant at 3–5 months but were significant, in favour of trauma-focused intervention, at 6–13 months (K = 5; N = 469; SMD −0.48, 95% CI −0.81, −0.15; I2 = 62%; GRADE very low). There was also a difference in favour of trauma-focused intervention for PTSD diagnosis post-treatment based on outcomes from two studies (K = 2; N = 201; RR 0.62, 95% CI 0.46, 0.84; I2 = 30%; GRADE very low). No differences were observed for alcohol or drug use, beyond those achieved for SUD-only intervention, apart from a small effect for alcohol use in favour of trauma-focused intervention at 6–13 months (K = 4; N = 363; SMD −0.23, 95% CI −0.44, −0.02; I2 = 0%; GRADE very low). No differences were observed for treatment dropout or adverse effects. Adverse events were not reported for Coffey et al. (2006) and Coffey et al. (2016).

In view of the limited number of studies, we only investigated subgroup differences by intervention model. For studies evaluating COPE, there were no post-treatment differences for PTSD (K = 3; N = 190; SMD −0.31, 95% CI −0.93, 0.32; I2 = 74%; GRADE very low), alcohol use (K = 1; N = 23; SMD 0.07, 95% CI −0.76, 0.91; GRADE very low), or drug use (K = 3; N = 170; SMD −0.16, 95% CI −0.63, 0.95; I2 = 81%; GRADE very low).

Cognitive restructuring-based interventions

Four studies tested integrated cognitive restructuring-based approaches for PTSD and SUD (without imaginal exposure) with interventions for SUD only (Capone et al., 2018; Haller et al., 2016; McGovern et al., 2011, 2015). Capone et al. (2018) and McGovern et al. (2011, 2015) tested up to 12 sessions of ICBT for PTSD and SUD, while the approach tested by Haller et al, (2016) was based on a form of CPT, where a narrative account of the index trauma experience was excluded. All participants in this study completed 12 sessions of group CBT for SUD and depression before randomization to either 12 individual sessions of CPT for PTSD and SUD, or individual CBT for SUD and depression only. Intervention was offered individually in McGovern et al. (2011, 2015), whereas Capone et al. (2018) offered the first four sessions individually and subsequent sessions were group based. Where reported, participants attended between 45.7% (Capone et al., 2018) and 56.3% (Haller et al., 2016) of integrated cognitive restructuring sessions available to them. Based on data from two studies (Capone et al., 2018; McGovern et al., 2015), 39.4% of participants were considered to be treatment completers. McGovern et al. (2015) included both a relapse prevention and an SUD-only TAU arm. Data from both arms were entered separately into meta-analysis and we followed the guidance in the Cochrane Handbook (Higgins & Green, 2011) to ensure that data from the integrated cognitive restructuring arm were not double counted. We found no benefit beyond that of intervention for SUD at any time-point for integrated cognitive interventions for PTSD severity, PTSD diagnosis, alcohol use, or substance use. Outcomes were not reported for treatment dropout or adverse events.

We undertook a subgroup analysis of post-treatment outcomes to investigate the potential benefits of ICBT, excluding Haller et al. (2016) as this study was based on CPT. For PTSD severity, there was a positive effect in favour of ICBT (K = 3; N = 263; SMD −0.33, 95% CI −0.62, −0.04; I2  = 21%; GRADE very low). Effects for alcohol use (K = 3; N = 263; SMD −0.01, 95% CI −0.26, 0.24; I2 = 45%; GRADE very low) and drug use (K = 3; N = 263; SMD −0.16, 95% CI −0.42, 0.09; I2 = 0%; GRADE very low) remained non-significant.

Other interventions

No studies tested interventions based on EMDR or any other trauma-focused approach. Data were available for a subset of participants with PTSD and SUD in a study evaluating CBT (without SUD intervention) versus a waiting list control in one study targeted at patients with PTSD and serious mental illness (Mueser et al., 2008). The extent to which participants in this study were also receiving intervention specifically for SUD is unknown, so we did not include it in the meta-analyses of integrated cognitive restructuring-based interventions described in Section 3.1.1.3. There was no difference between CBT and the control condition for this SUD subgroup for PTSD severity at any time-point or for PTSD diagnosis. Our other outcomes were not reported. The authors found that across the whole sample, PTSD severity was significantly reduced for those receiving CBT, in comparison to the control.

Data were also available for a subset of participants with PTSD and AUD in a study testing a 12 week web-based CBT intervention (Thinking Forward) plus TAU versus TAU for veterans with PTSD and harmful drinking (Acosta et al., 2017). CBT in this intervention did not include a trauma processing or imaginal exposure component. There were no differences for this subgroup on PTSD severity or alcohol use at any time-point. Other outcomes were not reported. The authors reported that across the whole sample (those with AUD and hazardous drinking) there were no significant effects for PTSD, but there were benefits for the web-based CBT for alcohol use and binge drinking.

Active psychological intervention versus other active psychological intervention

Five studies compared two active psychological interventions aimed at treating both PTSD and SUD.

Present-focused versus trauma-focused intervention

Norman et al. (2019) compared trauma-focused intervention (COPE) with present-focused intervention (Seeking Safety). Participants received up to 16 sessions of treatment. There was a significant difference in favour of trauma-focused intervention plus SUD intervention for PTSD severity post-treatment (N = 119; SMD −0.49, 95% CI −0.86, −0.13) and at 6–13 months (N = 119; SMD −0.44, 95% CI −0.81, −0.08), but not at 3–5 months. There were no differences for alcohol use, dropout, or adverse events.

Present-focused versus integrated CBT intervention

Najavits et al. (2018) tested an integrated past-focused CBT-based approach not including imaginal exposure (Creating Change) with present-focused intervention (Seeking Safety), with participants receiving up to 17 sessions. There were no differences between the two interventions for PTSD severity, or alcohol use at any time-point, or in terms of adverse events.

Other active comparisons

Stappenbeck et al. (2015) tested the effectiveness of teaching two coping skills: cognitive restructuring versus experiential acceptance versus attention control. Participants received one session of training followed by up to four coaching calls in the following week. There were no differences between the two interventions for PTSD severity, or alcohol use post-treatment, or treatment dropout. Adverse events were not reported.

Possemato et al. (2019) tested whether there was any benefit in adding peer support to a web-based CBT intervention (Thinking Forward), compared to self-directed web-based CBT, in a small pilot trial aimed at individuals with PTSD and problematic drinking. Data were obtained for a small subset of individuals who met the diagnosis for AUD. There were no differences between the two groups for PTSD or alcohol use outcomes for this subset.

Schacht et al. (2017) tested whether PE where attendance was incentivized with vouchers was superior to PE without incentivization. Participants received up to 12 sessions of treatment. Those in the incentivized arm attended 59.3% of available sessions, with those in the standard PE arm attended 15.0% of sessions. There were no differences between the two interventions for PTSD severity, or alcohol use at any time-point. However, there was a difference in favour of the incentivized approach for treatment dropout (N = 58; RR 10.71, 95% CI 1.46, 78.39).

Sequential psychological intervention versus integrated intervention

Only one study attempted to test sequential versus integrated treatment. Kehle-Forbes et al. (2019) tested the delivery of four sessions of motivational interviewing before 12 sessions of PE against 16 sessions of integrated motivational interviewing and PE. Participants attended 57.3% of available sessions for the sequential approach and 52.2% of sessions for the integrated approach. There were no significant differences between the two approaches for PTSD severity or alcohol use at any time-point (Table 3). Dropout, in favour of the sequential approach, approached but did not quite reach significance (N = 183; RR 1.57, 95% CI 0.99, 2.48).

Table 3.

Summary of findings for each comparison for two active interventions.

Comparison outcome	Follow-up point	Contributing studies	Studies (n)	Sample (n)	RR (95% CI)	SMD (95% CI)	Grade rating
Trauma-focused treatments + TAU for SUD vs present-focused treatments + TAU for SUD
PTSD severity	Post-treatment	Norman et al. (2019)	1	119		−0.49 (−0.86, −0.13)*	Very low
	3–5 months	Norman et al. (2019)	1	119		−0.30 (−0.66, 0.06)	Very low
	6–13 months	Norman et al. (2019)	1	119		−0.44 (−0.81, −0.08)*	Very low
PTSD diagnosis	Post-treatment	Norman et al. (2019)	1	119	0.83 (0.69, 1.01)		Very low
Alcohol use	Post-treatment	Norman et al. (2019)	1	119		0.13 (−0.23, 0.49)	Very low
	3–5 months	Norman et al. (2019)	1	119		−0.08 (−0.44, 0.28)	Very low
	6–13 months	Norman et al. (2019)	1	119		0.06 (−0.30, 0.42)	Very low
Dropout		Norman et al. (2019)	1	119	0.48 (0.32, 0.72)		Very low
Adverse events		Norman et al. (2019)	1	119	No participants were discharged from the study owing to SAEs		Very low
Creating Change + TAU for SUD vs present-focused treatments + TAU for SUD
PTSD severity	Post-treatment	Najavits et al. (2018)	1	52		−0.18 (−0.72, 0.37)	Very low
	3–5 months	Najavits et al. (2018)	1	52		0.23 (−0.32, 0.77)	Very low
Alcohol use	Post-treatment	Najavits et al. (2018)	1	52		0.25 (−0.29, 0.80)	Very low
	3–5 months	Najavits et al. (2018)	1	52		−0.39 (−0.94, 0.16)	Very low
Substance use	Post-treatment	Najavits et al. (2018)	1	52		−0.15 (−0.70, 0.39)	Very low
	3–5 months	Najavits et al. (2018)	1	52		0.18 (−0.37, 0.72)	Very low
Adverse events		Najavits et al. (2018)	1	52	No AEs related to either treatment		Very low
Brief cognitive restructuring training vs brief experiential acceptance training
PTSD severity	Post-treatment	Stappenbeck et al. (2015)	1	58		0.24 (−0.28, 0.76)	Very low
Alcohol use	Post-treatment	Stappenbeck et al. (2015)	1	58		−0.46 (−0.98, 0.06)	Very low
Dropout		Stappenbeck et al. (2015)	1	60	1.07 (0.96, 1.21)		Very low
Web-based CBT + peer support vs web-based CBT
PTSD severity	Post-treatment	Possemato et al. (2019)	1	8		0.09 (−1.34, 1.52)	Very low
	3–5 months	Possemato et al. (2019)	1	8		−0.61 (−2.11, 0.88)	Very low
Alcohol use	Post-treatment	Possemato et al. (2019)	1	10		0.35 (−0.93, 1.63)	Very low
	3–5 months	Possemato et al. (2019)	1	10		0.46 (−0.84, 1.75)	Very low
Incentivized trauma-focused treatment + TAU for SUD vs standards trauma-focused treatment + TAU for SUD
PTSD severity	Post-treatment	Schacht et al. (2017)	1	50		−0.42 (−0.99, 0.14)	Very low
	3–5 months	Schacht et al. (2017)	1	47		−0.54 (−1.13, 0.04)	Very low
Alcohol use	Post-treatment	Schacht et al. (2017)	1	52		−0.29 (−0.94, 0.36)	Very low
	3–5 months	Schacht et al. (2017)	1	33		−0.21 (−0.91, 0.48)	Very low
Dropout		Schacht et al. (2017)	1	58	10.71 (1.46, 78.39)*		Very low
Adverse events					Randomized participants had 10 SAEs: 9 hospitalizations and 1 death. None of the SAEs was determined to be related to study participation		Very low
Sequential trauma-focused treatment + TAU for SUD vs integrated trauma-focused treatment + TAU for SUD
PTSD severity	Post-treatment	Kehle-Forbes et al. (2019)	1	157		0.03 (−0.28, 0.35)	Very low
	3–5 months	Kehle-Forbes et al. (2019)	1	141		−0.18 (−0.51, 0.15)	Very low
Alcohol use	Post-treatment	Kehle-Forbes et al. (2019)	1	159		−0.28 (−0.59, 0.04)	Very low
	3–5 months	Kehle-Forbes et al. (2019)	1	145		−0.19 (−0.51, 0.14)	Very low
Dropout		Kehle-Forbes et al. (2019)	1	183	1.57 (0.99, 2.48)		Very low

AE, adverse event; CBT, cognitive behavioural therapy; CI, confidence interval; PTSD, post-traumatic stress disorder; RR, relative risk; SAE, serious adverse event; SMD, standardized mean difference; SUD, substance use disorder; TAU, treatment as usual.

RR (of categorical data): RR = 1, same as control; RR < 1, intervention better; RR > 1, control better.

SMD (of continuous symptom score): SMD = 0, no difference between intervention and control ; SMD < 0, intervention better; SMD > 0, control better.

*Statistically significant difference at the p < 0.05 level.

Synthesis of results for children and young people

Only one pilot study conducted with young people met our inclusion criteria (Najavits et al., 2006). This study evaluated an individual present-focused therapy (Seeking Safety + TAU for SUD) in comparison to TAU for SUD in adolescent girls. The study did not provide data usable for secondary analysis post-treatment for PTSD severity or alcohol or drug use. The authors report that outcomes were better for those allocated to Seeking Safety on seven out of 10 substance abuse-related subscales on the Personal Experiences Inventory, which was the study’s primary outcome. Participants in the Seeking Safety group attended a mean of 9.7 (38.7%) of the available treatment sessions. Dropout and adverse events were not reported.

Methodological quality of included studies

Risk-of-bias judgements for each study are shown in Supplementary Table 1. Eighteen studies were judged to have a low risk of bias for the method of allocation and no studies were judged to be of high risk. Twelve studies provided a clear description of allocation concealment, indicating low risk of bias, and two studies were judged to be at high risk in this domain (Najavits et al., 2018; Stappenbeck et al., 2015). Adequate blinding of assessment of outcomes measures was indicated in 15 studies, with two studies judged to be at high risk (McGovern et al., 2011; Myers et al., 2015). Seventeen studies were judged to be at low risk for incompleteness of data, with five studies judged to be at high risk (Coffey et al., 2006; Hien et al., 2004; Mueser et al., 2008; Myers et al., 2015; Ruglass et al., 2017). Twelve studies, all published since 2011, had registered protocols and were judged to be at low risk of selective reporting, and five studies were judged to be of high risk (Coffey et al., 2006; Kehle-Forbes et al., 2019; McGovern et al., 2015; Najavits et al., 2006; Possemato et al., 2019). We considered seven studies to be at low risk of other bias and 16 studies to be at high risk. Reasons for these studies being judged as high risk included possible investigator allegiance to one of the interventions being evaluated (Acosta et al., 2017; Back et al., 2019; Foa et al., 2013; Najavits et al., 2006, N2018; Possemato et al., 2019; Ruglass et al., 2017), potentially imbalanced use of other interventions (Capone et al., 2018; Schäfer et al., 2019; Zlotnick et al., 2009), possible contamination of treatments (Frisman et al., 2008; McGovern et al., 2015; Zlotnick et al., 2009), unclear reporting of the number of participants randomized (McGovern et al., 2015), lack of a study intervention manual (McGovern et al., 2015), significant differences in one or more potentially important baseline characteristics (McGovern et al., 2011; Mills et al., 2012; Myers et al., 2015; Najavits et al., 2006; Ruglass et al., 2017; Stappenbeck et al., 2015), differences in the number of available treatment sessions (McGovern et al., 2011), divergence from time of planned collection of outcomes (Mills et al., 2012), small sample size (Myers et al., 2015; Possemato et al., 2019), and the use of measures with possibly inadequate validation (Stappenbeck et al., 2015). We were unable to investigate for potential publication bias through funnel plots, as there were insufficient studies in each of the analyses to allow us to create these.

Discussion

Main findings

Twenty-seven studies with 2849 participants were included in this review, evaluating a number of mainly CBT-based approaches. For adults, we found evidence that trauma-focused intervention delivered alongside treatment for SUD led to improvements in PTSD symptom severity and diagnosis post-treatment when compared against intervention for SUD only, although the observed effect was below our predefined threshold for a clinically important finding. There was some evidence that PTSD improvements were maintained, and of delayed benefits in reduced alcohol use beyond that of SUD-only intervention, with effects above our threshold of clinical importance for PTSD at 6–13 months, although follow-up data were not available from all trauma-focused studies. A subgroup analysis of studies evaluating the COPE model showed no benefit on any of our primary outcomes over SUD-based intervention alone. However, there was evidence, from one head-to-head study (Norman et al., 2019), that trauma-focused intervention in the form of COPE was more effective than present-focused intervention (Seeking Safety) for PTSD severity post-treatment, again with a mixed picture at subsequent follow-up. Although studies evaluating present-focused approaches, mainly evaluating Seeking Safety, reported improvements for PTSD severity and alcohol and drug abuse for participants receiving these interventions, there was no evidence that these approaches led to improvements beyond those found for SUD-only interventions. Although our evaluations of outcomes for all studies evaluating cognitive restructuring-based interventions for PTSD and SUD were not significant, we did find a benefit for ICBT, specifically in improving PTSD severity in a prespecified subgroup analyses (excluding Haller et al., 2016). This effect was below our threshold for a clinically important finding. We note that the Haller study was unique in that all participants received 12 sessions of group ICBT for SUD and depression prior to randomization to individual ICBT for SUD and depression only or individual CPT for PTSD and SUD. It is arguable that the two trial interventions in this study were not sufficiently different to be able to demonstrate a meaningful difference between cohorts. There was no evidence of differential benefit for any other intervention evaluated in this review.

Only one study attempted to evaluate the order of intervention delivery (Kehle-Forbes et al., 2019). In this study, motivational interviewing was either offered before prolonged exposure or delivered in an integrated way alongside PE. There were no significant differences in outcome. While this study is a welcome addition to the literature, we felt that it did not really address whether the sequential delivery of treatment, as seen in routine clinical practice is any more or any less effective than integrated or combined intervention delivery. Participants in the sequential arm of the study received only four sessions of motivational enhancement therapy (MET) before progressing to PE with the same therapists. They also continued to have their substance use monitored by their therapist, with further MET if needed. It therefore remains very unclear which approach is more suitable in routine clinical practice.

Only one small pilot RCT involving children and adolescents met our inclusion criteria (Najavits et al., 2006). This study reported some promising changes in substance abuse-related outcomes and trauma-related beliefs, in favour of Seeking Safety.

GRADE judgements for all positive findings were very low, indicating that we are very uncertain about these estimated effects, which are very likely to change as further studies are reported. Findings received very low ratings for a number of different reasons. Some findings came from single small or moderately powered studies with noted risk of bias concerns. In addition, findings based on meta-analyses of multiple studies tended to be heterogeneous in terms of study populations and study design features, and we could not rule out the possibility of publication bias. Findings in relation to trauma-focused interventions were also rated very low, in part because reported effects were not consistent across the included studies.

The increase in studies from 14 identified by Roberts et al. (2015, 2016) to 27 in the current review indicates that this remains an important area of clinical and academic concern for many in the field. The findings from this review in favour of trauma-focused intervention over present-focused approaches are largely consistent with, and reinforce those of earlier reviews (e.g. Roberts et al., 2015, 2016; Simpson et al., 2017; Torchalla et al., 2012; van Dam et al., 2012), but are based on a larger number of trials. However, to our knowledge, no review has previously reported on small positive outcomes for ICBT for PTSD. Alcohol and substance use reduced across all studies for both experimental and SUD-only treatments, and we saw no evidence of any differential effect for either condition (apart from the small effect in favour of trauma-focused intervention at 6–13 months) across the interventions that we investigated. This is perhaps not surprising given that all participants were able to access some SUD-based intervention, regardless of treatment allocation, and, as participants were normally recruited from addiction services, they were likely to have achieved some reduction in alcohol and drug use consumption before study recruitment occurred. However, we note that the dose of SUD-focused treatment in the experimental integrated treatments that we identified was often significantly less than that offered in the equivalent SUD-only arm, suggesting that on average participants were improving comparably and meaningfully in both conditions. The finding that there was a small difference in alcohol use at 6–13 months in favour of trauma-focused intervention over SUD intervention only is of interest. This difference was seen alongside a similar significant difference in PTSD severity and may indicate a reduced need to self-medicate (Dvorak et al., 2014; Kaysen et al., 2014; Leeies et al., 2010) in those who had received trauma-focused intervention. However, we did not observe a similar improvement in substance use and, given the increased difficulty in obtaining longer-term outcome data, this finding needs to be treated with caution.

Study-defined criteria for a treatment completer were generally modest. As reported previously (Roberts et al., 2015, 2016; Simpson et al., 2017), there was evidence of a high level of dropout across studies and of a relatively low uptake of available treatment sessions, regardless of intervention. Studies also tended to report quite large standard deviations, relative to outcome means, suggesting quite a wide variation in outcome responses. There was evidence from one study (Schacht et al., 2017) that dropout was reduced when attendance was incentivized. Patients are known to drop out of treatment for a variety of reasons (Lewis, Roberts, Gibson, & Bisson, 2020). There is some evidence from secondary evaluations of some of the larger trials included in this review that a number of factors may influence dropout, including trauma type, rate of PTSD and addiction symptom improvement, between-session alcohol use, educational status, anxiety sensitivity, and more severe baseline PTSD (Belleau et al., 2017; Kline, Panza, et al., 2021; López-Castro, Zhao, Fitzpatrick, Ruglass, & Hien, 2021; Szafranski et al., 2017; Zandberg, Rosenfield, Alpert, McLean, & Foa, 2016), although PTSD severity is not a consistent finding (Belleau et al., 2017). Session attendance has also been found to be influenced by drug-use severity, abstinence status, and employment status (Grundmann et al., 2021), and more frequent days of heavy drinking and greater AUD severity (Kline, Straus, et al., 2021). There is also some indication that patients who drop out of treatment prematurely and those who do not engage in optimal attendance may still make important improvements (Hien et al., 2012; Szafranski et al., 2019). However, evidence from the Norman et al. (2019) study indicates that differences in PTSD outcomes in favour of trauma-focused over present-focused therapy intervention were only evident for participants who completed all or most of their treatment sessions (Straus et al., 2022).

Strengths and limitations

This systematic review was undertaken according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) (Moher, Liberati, Tetzlaff, & Altman, 2009) and Cochrane Collaboration guidelines (Higgins & Green, 2011) and included registration of a review protocol. We followed a thorough search, screening, and study identification strategy. We searched for both published and unpublished studies, and accessed data from several studies where a subgroup of participants met our inclusion criteria. The methodological quality of included studies was reviewed by all authors and the quality of included studies was comparable to those included in analogous reviews (e.g. Lewis, Roberts, Andrew, et al., 2020). Our main methodological concerns were the fact that some studies were conducted by groups with allegiance to interventions that they were investigating and loss to follow-up. Study authors generally took steps to impute missing data through appropriate statistical methods, although reported outcomes for several studies were only based on available follow-up data. An updated Cochrane Collaboration risk-of-bias tool is now available. Our use of the previous version was to maintain continuity with Roberts et al. (2016) and may be viewed as a limitation.

There was significant clinical and statistical heterogeneity in the studies included in this review, and this should be held in mind in interpreting the reported findings. Studies were conducted with a number of different populations, including veterans, victims of domestic abuse, prisoners, and community treatment-seeking patients. Some studies included mainly male samples, while others were female only. Participants also differed in the severity and nature of their addiction use and type of trauma exposure. We grouped interventions for comparison in what we felt was a clinically meaningful way according to our prespecified plan, but within these comparisons there was significant variability in how experimental interventions were delivered and the number of treatment sessions available. There were also significant differences in the nature of control SUD-only interventions that were offered, with some control interventions being more active than others. There was sizeable statistical heterogeneity in some comparisons, but because of the small number of studies contributing to each comparison, we were only able to undertake limited sensitivity/subgroup analyses and we chose to focus on treatment model as we judged that this would be of greatest interest to the reader. Owing to the small number of studies included in each analysis, we were unable to investigate the possibility of publication bias.

Clinical implications

The number of studies that have evaluated psychological interventions for PTSD and SUD has increased appreciably over the past 5 years, and this review gives a clear indication of interventions that appear to be helpful, particularly in reducing PTSD symptoms, with the strongest findings for trauma-focused interventions. As indicated by Roberts et al. (2015), clinicians should continue to exercise caution when considering whether to provide the interventions identified in this review, as it is unlikely that these interventions will be appropriate for everyone and individual treatment planning will need to be guided by an individual formulation, which takes account of the predictors of outcome, alongside patient-related priorities, risk factors, and preferences (Roberts et al., 2020). The studies included in this review mostly continued to exclude on the basis of cognitive impairment, current suicide risk, and severe mental illness, and our findings should not be generalized to individuals with these types of presentation. Going forward, initiatives such as Project Harmony (Saavedra et al., 2021), which plan to undertake meta-analysis of individual patient data, are likely to greatly improve our understanding of key patient outcome predictors.

Implications for research

Given the limited treatment gains described in this review, there is a clear continuing need to develop and evaluate psychological interventions for PTSD and SUD for both adults and younger people. Among children and adolescent populations, we only included one pilot study with female adolescents (Najavits et al., 2006). We identified one other adolescent-focused RCT evaluating a family therapy-based intervention (Danielson et al., 2020) that did not fulfil our diagnostic inclusion criteria, and a protocol for one study which is currently under way (Mills et al., 2020).

There is a need to develop a better understanding of the predictors of symptom improvement and of treatment dropout. It seems of high importance to test whether other non-trauma-focused interventions may be beneficial to patients with PTSD and SUD. The trauma-focused approaches included in this review were all based on prolonged exposure; and there is a need to test whether other established trauma-focused approaches, such as EMDR, cognitive therapy for PTSD, and CPT, can be adapted for this population. There is still considerable uncertainty about how SUD- and PTSD-related interventions should be coordinated and whether sequential, integrated, or combined delivery is most effective or acceptable. Finally, from the PTSD field, several studies have reported on the efficacy of intensive trauma-focused therapies, delivered in a concentrated way over several days (e.g. De Jongh et al., 2020; Ehlers et al., 2014; Voorendonk, De Jongh, Rozendaal, & Van Minnen, 2020). Findings reported for a multicomponent inpatient programme for PTSD patients with significant comorbidity and other complexities have been encouraging in several studies and may warrant evaluation for those with comorbid PTSD (e.g. De Jongh et al., 2020; Voorendonk et al., 2020; Zoet, Wagenmans, van Minnen, & de Jongh, 2018).

Click here for additional data file.