How changes in post‐traumatic stress disorder (PTSD) severity mediate substance use disorder (SUD) severity during and after treatment for co‐occurring PTSD and SUD: Results from Project Harmony

Abstract

Background and Aims

Post‐traumatic stress disorder (PTSD) commonly co‐occurs with substance use disorders (SUD). Comorbid PTSD and SUD (PTSD+SUD) is associated with greater severity and impairment and poorer treatment outcomes. Several interventions exist to treat PTSD, SUD and PTSD+SUD; however, research has yet to elucidate the indirect pathways underlying treatment for PTSD+SUD. The present study examined how changes in PTSD severity relate to changes in SUD severity across treatment types during and post‐treatment.

Methods

Observational study using data collected as part of Project Harmony, a virtual clinical trial employing integrative data analysis to compare treatment effectiveness of PTSD+SUD interventions from 36 randomized controlled trials for PTSD+SUD (n = 4046). Multilevel mediated linear growth modeling was used to examine potential outcomes mediation. Each of the eight active treatments was compared to treatment as usual (TAU) for both alcohol and drug use outcomes.

Results

Alcohol use severity outcomes were fully or partially mediated by changes in PTSD severity for trauma‐focused psychotherapy + AUD medication [ab = −0.16 (95% confidence interval = −0.30 to −0.04)]; other treatments with mediation effects included trauma‐focused integrated psychotherapy, AUD medications and PTSD medications. Drug use severity outcomes were fully or partially mediated by changes in PTSD severity for trauma‐focused psychotherapy + AUD medication [ab = −0.08 (−0.18 to −0.001)]; other treatments with mediation effects on drug use severity included trauma‐focused integrated psychotherapy, AUD medications, PTSD medications and placebo medications.

Conclusions

Among people with co‐occurring post‐traumatic stress disorder (PTSD) and substance use disorders (SUD), reductions in alcohol and drug use severity appear to be mediated by reductions in PTSD during treatment. For those with drug use disorders, PTSD reductions appear to mediate further SUD reductions after treatment.

INTRODUCTION

Post‐traumatic stress disorder (PTSD) and substance use disorders (SUD) frequently co‐occur, with 20% to 60% of individuals with PTSD also having a SUD [1, 2]. Compared to adults without such diagnoses, adults with a past‐year drug use disorder had 1.6 to 2.2 greater odds of having PTSD [3], and adults with lifetime alcohol use disorder (AUD) had 1.3 to 1.4 greater odds of having PTSD [4]. In comparison to PTSD or SUD alone, co‐occurring PTSD and SUD (PTSD+SUD) presents a more severe clinical profile, marked by greater psychosocial problems, additional psychiatric comorbidity, suicide attempts, legal issues, housing concerns and medical problems [5, 6]. Moreover, individuals with PTSD+SUD are less likely to complete or respond to treatment [7, 8, 9]. Given the prevalence and deleterious consequences of this comorbidity, it is imperative to understand, which treatments work best for individuals with PTSD+SUD, and the intermediate endpoints [10, 11] through which different treatments reduce symptoms of both PTSD and SUD.

Comparative efficacy of PTSD+SUD treatments

Existing PTSD+SUD treatments can be classified by intervention type (e.g. psychotherapy or pharmacotherapy), whether the treatment is integrated (i.e. targeting PTSD and SUD simultaneously) or non‐integrated (i.e. targeting only PTSD or only SUD), whether intervention‐related content is ‘trauma‐focused’ (TF) or not [12], and the phasing of treatment (i.e. addressing PTSD and SUD concurrently or sequentially). TF treatments address trauma‐related symptoms through exposure to trauma‐related memories and cues [13], restructuring trauma‐related beliefs [14] and processing trauma‐related emotions. In contrast, non‐TF treatments teach adaptive coping skills to manage symptoms and stressors (e.g. problem‐solving, assertiveness [15]), but do not explicitly address trauma memories.

Extant meta‐analyses have yielded disparate findings regarding, which PTSD+SUD psychotherapies work best and for whom. Some syntheses demonstrate that TF treatments outperformed non‐TF interventions [8, 16], while others found less robust differences across treatment types and even some advantage for non‐integrated SUD‐focused psychotherapies for SUD symptoms [17, 18]. Evidence syntheses of pharmacological interventions have largely been limited to systematic reviews [19]. Importantly, prior evidence syntheses had been cited as ‘low quality’ [8, 16] because they used suboptimal risk of bias criteria and narrow inclusion criteria. Further, many interventions were not compared against a ‘common’ comparator condition, which can advantage effect sizes for treatments with weaker comparators within individual trials [20].

To address these limitations, the Project Harmony (PH) group completed a systematic review and network meta‐analysis (NMA) of 39 randomized controlled trials (RCTs) of PTSD+SUD treatments [21]. The NMA demonstrated that TF and integrated TF psychotherapies were superior for reducing PTSD symptoms compared to non‐TF or SUD‐focused pharmacotherapy. TF psychotherapy, particularly when combined with pharmacotherapies for AUD, was also effective at improving alcohol use severity, compared to placebo and PTSD pharmacotherapy. PH also conducted a combined meta‐analysis of individual patient data and integrative data analysis (MIPD/IDA), which has several methodological advantages over prior meta‐analyses [22, 23], including comparability in measurement of outcomes under IDA [24, 25] and the evaluation of head‐to‐head comparative effectiveness of diverse interventions in a pooled data set. Individual‐level data were harmonized across 36 RCTs (n = 4046) to test the relative efficacy of psychotherapy, pharmacological and combination interventions for PTSD+SUD compared to treatment as usual (TAU) [26]. As with the NMA [21], TF psychotherapies, in combination with pharmacotherapies for SUD, showed the largest relative reductions in alcohol use and PTSD severity at post‐treatment [27]. However, no treatment types examined were inefficacious, suggesting that each type has some utility [18].

Present study

A critical next step in optimizing PTSD+SUD care is understanding how treatments exert change on each set of symptoms—particularly whether SUD improvements are direct treatment effects and/or mediated by PTSD improvements [28, 29, 30, 31]. The self‐medication model provides a theoretical framework for conceptualizing a mediation pathway between PTSD and SUD symptoms; that is, individuals use substances to cope with PTSD symptom related distress [32, 33]. As such, untreated PTSD in this comorbid population can increase the likelihood of continued use and relapse [34]. In line with the self‐medication model, TF approaches show greater symptom reductions than other treatment modalities [27], suggesting that PTSD improvement is a necessary intermediate endpoint [10, 35] underlying SUD symptom reduction in this comorbid population, however, no PTSD+SUD evidence synthesis has assessed mediation effects [26].

The PH dataset provides an opportunity to examine mechanistic questions with a large sample incorporating counterfactual mediation modeling [10, 36] of both during treatment (pre‐ to post‐treatment) and after treatment (post‐treatment through follow‐up) changes in PTSD severity and their direct and indirect effect on SUD outcomes. This secondary analysis of the PH dataset tested the mediating effects of change in PTSD on two key outcomes: alcohol use severity and drug use severity. We examined these effects for two different phases: in‐treatment and post‐treatment. Given the overall superiority of TF treatments on direct outcomes for PTSD and alcohol use, we hypothesized that PTSD symptom changes would specifically mediate the effect of TF psychotherapies on both alcohol use and drug use severity in‐treatment and post‐treatment [37, 38, 39]. This work helps to illuminate treatment processes that can advance tailoring efforts to improve interventions for individuals with PTSD+SUD.

METHOD

We analyzed data from 36 RCTs included in the PH dataset. Studies included in PH were indexed from a systematic review of PTSD+SUD RCTs published between 1995 and 2020 and trials registered on Clinicaltrials.gov. In line with previous work using PH data, we modeled PTSD and SUD severity using data obtained from baseline, treatment mid‐point, treatment end‐point, 3‐, 6‐, 9‐ and 12‐month follow‐ups (see Figure 1). Covariates included age, gender, race/ethnicity, education level, marital status, population status (i.e. civilian, Veteran, incarcerated), proportion of intended treatment dosage, depression status and concomitant medications. For Project Harmony search criteria, see Appendix S1. For further details on harmonization across timepoints or information on covariates, please see supplemental material from prior work [27]. Details of PH have been published elsewhere [21, 26].

FIGURE 1.

Multi‐level mediated linear growth model examining potential outcomes mediation. This diagram illustrates a multi‐level mediated linear growth model examining the mediating effect of treatment [9 conditions; 8 active conditions vs. treatment as usual (TAU)] on the relationship between change in post‐traumatic stress disorder (PTSD) severity and change in substance use severity across six timepoints (i.e. baseline, mid‐treatment, end‐treatment, 3‐month, 6‐month, 9‐month and 12‐month follow‐ups). Octagons reflect the latent scale scores estimated previously under the moderated non‐linear factor analysis framework. Latent growth factors were constructed to model piecewise linear trajectories (i.e. slopes) with random intercepts. The first trajectory estimated change from baseline through the end‐of‐treatment and the second trajectory estimated change from end‐treatment through 12‐month follow‐up. Indicators for the random intercept growth factors were set to ‘1’. Indicators for the first slope factor were set to ‘0’, ‘0.5’, ‘1’, ‘1’, ‘1’, ‘1’ and ‘1’ for each timepoint, respectively. Indicators for the second slope factor were set to ‘0’, ‘0’, ‘0’, ‘1’, ‘2’, ‘3’ and ‘4’ for each timepoint, respectively. Age, gender, race/ethnicity, education level, marital status, population status (i.e. civilian, veteran, incarcerated), proportion of intended treatment dosage, depression status and concomitant medications are visualized as a global set of covariates. Propensity score weights allowed for balancing of covariates across treatment classes to account for differences across individual studies and to minimize the influence of confounds. Direct and indirect paths were estimated and are visualized as four sets of paths: (1) a1 and a2 pathways reflect the effect of treatment condition (vs. TAU) on changes in PTSD severity during‐ and post‐treatment; (2) b1, b2, b3 reflect the associations between changes in PTSD severity during‐ and post‐treatment and change in substance use disorder (SUD) severity during‐ and post‐treatment; (3) c1’ and c2’ reflect the effect of treatment condition (vs. TAU) on changes in SUD severity during‐ and post‐treatment; and (4) h1, h2, h3 reflect the interaction between treatment condition (vs. TAU) and changes in PTSD severity on changes in SUD severity, during‐ and post‐treatment.

MEASURES

Treatment assignment

All treatment arms across the 36 trials were coded based on whether the treatment was behavioral, pharmacological or combined; TF or not; integrated or not; or considered TAU. This non‐mutually exclusive coding yielded nine treatment types (Table 1), as described in Hien et al. [27].

TABLE 1.

Descriptions and examples of RCT treatment types.

Treatment type	Description	Original RCT arms
Integrated, trauma focused psychotherapy	Intervention content integrates direct processing of trauma memories with cognitive‐behavioral strategies for treating SUD	COPE Modified CPT PE + virtual reality Craving reduction intervention Integrated Motivational Enhancement Therapy/PE
Non‐integrated, trauma focused psychotherapy	Intervention content focuses on direct processing of trauma memories but does not include strategies for treating SUD	PE CPT
Integrated, non‐trauma focused psychotherapy	Intervention content integrates cognitive or behavioral strategies for treating PTSD symptoms (without direct processing of trauma content) with strategies for treating SUD	Seeking Safety Integrated CBT Vet change
SUD psychotherapy	Intervention content features supportive or cognitive‐behavioral strategies for treating SUD	Relapse Prevention Individual addiction counseling CBT for alcohol support CBT for SUD
PTSD pharmacotherapy	Medication intended to target PTSD symptoms	Prazosin Topiramate
AUD pharmacotherapy	Medication intended to target AUD symptoms	N‐acetylcysteine Naltrexone Paroxetine Prazosin Sertraline Topiramate Zonisamide
Placebo medication	Non‐active medication (mechanism of action does not directly target PTSD or SUD symptoms)
Combined trauma focused psychotherapy + AUD pharmacotherapy	Intervention pairs trauma‐focused treatment (integrated or non‐integrated) with medication intended to target AUD symptoms	CPT + zonisamide PE + naltrexone EMDR + SSRI
Treatment‐as‐usual psychotherapy	Treatment does not directly target PTSD or SUD symptoms	Community mental health treatment Waitlist control Non‐evidence based behavioral treatment

Note: For our purposes, ratings were based on the intended target of the intervention described in the manuscript, regardless of whether or not they are considered evidence‐based efficacious agents at present.

Abbreviations: AUD, alcohol use disorder; COPE, Concurrent Treatment of PTSD and Substance Use Disorders Using Prolonged Exposure; CPT, Cognitive Processing Therapy; EMDR, Eye Movement Desensitization and Reprocessing; PE, Prolonged Exposure; PTSD, post‐traumatic stress disorder; RCT, randomized controlled trials; SSRI, selective serotonin reuptake inhibitors; SUD, substance use disorder.

Latent PTSD severity

Latent PTSD severity scores estimated previously under the moderated non‐linear factor analysis (MNLFA) framework [40, 41] for other PH analyses [27, 42, 43] were used. MNLFA scale scores were estimated using 42 harmonized PTSD indicators (21 clinical interview symptoms and 21 self‐report symptoms), with the item mix including (1) 16 PTSD symptoms common to both the Diagnostic and Statistical Manual of Mental Disorders, fourth edition (DSM‐IV) and DSM‐5 diagnostic systems; (2) one symptom (sense of foreshortened future) unique to DSM‐IV; and (3) four symptoms unique to DSM‐5.

Latent alcohol use

A latent alcohol use variable was estimated under MNLFA using two past‐30‐day indicators: number of days with any alcohol use and number of days with alcohol use to intoxication.

Latent drug use severity

Latent drug use severity scores previously estimated under MNLFA [27] were used. Binary indicators of any past‐30‐day use of the following substances were used to support a six‐indicator continuous latent drug use variable: cocaine, heroin, opioids (excluding heroin), sedatives, other stimulants (excluding cocaine) and hallucinogens.

Data analytic plan

Analyses were conducted in SAS 9.4 [44] (SAS Institute, 2014) and Mplus 8.7 [45]. This study is similar in spirit to emulated target trials (ETTs) [46], where observational studies of interventions are under study. Here, treatment conditions were combined across RCTs, which can compromise within‐trial randomization in MIPDs [26, 47, 48], in a manner similar to ETTs. ETTs have five elements: (1) observational study eligibility criteria that mimic an RCT (in our case, these were RCTs); (2) sequential ignorability of treatment assignment (and of the mediator; MacKinnon et al. [10]; (3) consistent specification of the beginning of treatment (i.e. a common baseline); (4) consistent specification of outcomes; and (5) a comparison of intent‐to‐treat versus per‐protocol analysis. Elements 1 and 3 are easily achieved as functions of the designs of the original studies. Element 2 was addressed using average treatment effect propensity score weights (ATE‐PSWs) estimated in previous PH outcome analyses [27] using SAS Proc GLIMMIX and the potential outcomes mediation framework (see below). Element 4 was achieved through harmonization of mediators/outcomes described elsewhere [21]. Element 5 was achieved between the intent‐to‐treat analysis described in Hien et al. [21] versus a dose–response/per protocol analysis described in Morgan‐López et al., (in press) [49]. We used the multi‐level imputation R package ‘mice’ [50] to address missingness on covariates for estimation of propensity score weights, this procedure is recommended for data with a multi‐level structure and a mix of continuous and categorical variables for which normality cannot be assumed. Covariates accounted for in the propensity weighting included age, sex, race and ethnicity, education, marital status, veteran status, non‐study concomitant psychotropic medication prescription status and baseline severity of PTSD symptoms, alcohol use and non‐cannabis drug use. Readers interested in a complete description of our imputation approach are referred to the detailed PH primary outcomes paper supplement [27].

Propensity score‐weighted multi‐level mediated linear growth modeling using robust maximum likelihood estimation to examine potential outcomes mediation (see Figure 1). Models were specified to allow for examination of during‐ and post‐treatment changes in PTSD symptom severity as a mediator between treatment type and during‐ and post‐treatment SUD symptom severity changes. SUD outcomes were modeled using random individual‐level intercepts and slopes using a parallel growth process [51] and random intercepts at the study‐level to address study‐level heterogeneity. Using this longitudinal framework, we examined ‘a’ paths from the treatment type to the random slope for during‐ and post‐treatment changes in PTSD severity (via piecewise linear time structure) [27], ‘b’ from changes in PTSD severity to the SUD outcomes and ‘h’ paths that capture treatment by mediator interactions (required for testing of potential outcomes mediation) [10] to examine the mediating effect of changes in PTSD severity on effects of treatment type on SUD outcomes. We presented separate estimates of both the pure natural indirect effect (i.e. the expected mediation effect for participants in TAU) and total natural indirect effect (i.e. the mediation effect for those in an active treatment) only if treatment by mediator interactions were significant [10]. Tests for significance of mediation effects were estimated using tests for Bayesian potential outcomes mediation effects outlined in Miočević et al. [35]. Separate models were fit for alcohol outcomes and drug use outcomes. For each model, the treatment types were compared to TAU. For models with significant full or partial mediation, the fully standardized indirect effect is presented following the recommendations of Miočević et al. [52]:

$${\mathrm{ab}}_{\mathrm{fs}}=\mathrm{ab}\left({\mathrm{S}}_{\mathrm{X}}\right)/{\mathrm{S}}_{\mathrm{Y}}$$

Where (ab_fs represented the fully standardized mediational effect, ab = the indirect effect, S_x= the standard deviation of treatment type and S_Y = the standard deviation of the outcome (i.e. alcohol or drug use).

RESULTS

The sample was 53% male, 65% White, 25% Black or African American and 7% Hispanic. The mean age was 39.0 years. A detailed description of the study sample characteristics can be found in Hien et al. [27]. Results are presented for alcohol and drug use outcomes type by direct effects of treatment type on PTSD severity, treatment type and mediation effects in Table 2. ¹ Significant coefficients are presented in‐text and mediation model estimates are presented in Table 2. A descriptive summary of direct and mediation effects can be found in Table 3.

TABLE 2.

Mediation effects of intervention conditions on substance use outcomes as transmitted through changes in PTSD severity.

	Alcohol use severity (during treatment)			Drug use severity (during treatment)			Drug use severity (post‐treatment)
Treatment type	b	LCL	UCL	b	LCL	UCL	b	LCL	UCL
Integrated, trauma focused psychotherapy
a	−0.36	−0.71	0.00	−0.35	−0.70	−0.00	0.23	0.04	0.40
b	0.21	0.10	0.30	0.10	0.02	0.19	0.46	0.23	0.72
c′	0.21	−0.14	0.55	0.01	−0.40	0.44	−0.02	−0.13	0.08
ab	−0.10	−0.22	0.00	−0.05	−0.12	0.01	0.03	−0.09	0.16
TE	−0.28	−0.69	0.15	0.07	−0.50	0.62	−0.06	−0.24	0.13
Non‐integrated, trauma focused psychotherapy
a	−0.18	−0.42	0.05	−0.18	−0.42	0.05	−0.08	−0.22	0.07
b	0.21	0.10	0.30	0.10	0.02	0.19	0.46	0.23	0.72
c′	−0.35	−0.54	−0.16	0.12	−0.25	0.51	−0.05	−0.12	0.01
ab	−0.04	−0.09	0.01	−0.02	−0.05	0.08	−0.04	−0.11	0.03
TE	−0.39	−0.59	−0.19	0.09	−0.28	0.48	−0.09	−0.19	0.01
Integrated, non‐trauma focused psychotherapy
a	0.04	−0.13	0.21	0.04	−0.13	0.21	−0.09	−0.18	−0.01
b	0.21	0.10	0.30	0.10	0.02	0.19	0.46	0.23	0.72
c′	−0.05	−0.14	0.05	−0.01	−0.12	0.09	−0.01	−0.07	0.05
ab	0.01	−0.03	0.05	0.00	−0.01	0.03	−0.04	−0.09	0.00
CDE	−0.05	−0.14	0.05	−0.01	−0.12	0.09	−0.01	−0.07	0.05
TE	−0.04	−0.14	0.06	−0.01	−0.11	0.10	−0.06	−0.17	0.01
PNIE							−0.04	−0.09	0.00
TNIE							−0.05	−0.12	0.00
SUD psychotherapy
a	−0.04	−0.26	0.17	−0.04	−0.26	0.17	−0.05	−0.13	0.10
b	0.21	0.10	0.30	0.10	0.02	0.19	0.46	0.23	0.72
c′	−0.12	−0.31	0.09	0.07	−0.14	0.28	−0.13	−0.14	0.01
ab	−0.01	−0.06	0.04	0.00	−0.03	0.02	−0.01	−0.06	0.05
TE	−0.12	−0.32	0.09	0.06	−0.15	0.28	‐0.07	‐0.16	0.03
PTSD medication
a	−0.43	−0.64	−0.21	−0.42	−0.61	−0.21	0.27	−0.29	0.85
b	0.21	0.10	0.30	0.10	0.02	0.19	0.46	0.23	0.72
c′	0.12	−0.20	0.44	0.23	0.02	0.43	−0.04	−0.40	0.30
ab	−0.09	−0.15	−0.03	−0.04	−0.08	−0.01	0.12	−0.15	0.43
TE	0.03	−0.29	0.37	0.18	−0.03	0.38	0.08	−0.37	0.56
AUD medication
a	−0.31	−0.50	−0.11	−0.31	−0.49	−0.12	−0.25	−0.37	−0.11
b	0.21	0.10	0.30	0.10	0.02	0.19	0.46	0.23	0.72
c′	−0.58	−0.94	−0.21	0.15	−0.09	0.41	−0.02	−0.11	0.07
ab	−0.06	−0.11	−0.01	−0.03	−0.06	−0.00	−0.12	−0.21	−0.03
TE	−0.64	−1.01	−0.28	0.12	−0.14	0.37	−0.13	−0.26	−0.01
Placebo medication
a	−0.28	−0.54	−0.13	−0.27	−0.53	−0.01	−0.06	−0.26	0.14
b	0.21	0.10	0.30	0.10	0.02	0.19	0.46	0.23	0.72
c′	−0.27	−0.49	−0.04	0.21	0.03	0.41	−0.13	−0.10	0.08
ab	−0.06	−0.12	0.00	−0.03	−0.06	0.00	−0.03	−0.13	0.07
TE	−0.32	−0.57	−0.10	0.19	−0.01	0.38	−0.04	−0.18	0.09
Combined trauma focused psychotherapy + AUD medication
a	−0.33	−0.69	0.06	−0.32	−0.65	0.04	0.03	−0.17	0.21
b	0.21	0.10	0.30	0.10	0.02	0.19	0.46	0.23	0.72
c′	0.37	0.00	0.74	0.03	−0.39	0.44	0.06	−0.05	0.17
ab	−0.16	−0.30	−0.05	−0.09	−0.18	−0.01	−0.14	−0.29	0.01
TE	−0.73	−1.29	−0.15	0.21	−0.40	0.85	−0.15	−0.36	0.08

Notes: Bold indicates significance. a = treatment to mediator path. b = mediator to outcome path. c′ = direct effect. There were no statistically significant effects for alcohol use severity at post‐treatment.

Abbreviations: LCL, lower confidence limit; PNIE, pure natural indirect effect; TE, total effect, TNIE, total natural indirect effect; UCL, upper confidence limit.

TABLE 3.

Summary of mediation effects of reductions in PTSD severity on alcohol and drug use severity during and post‐treatment.

Treatment type	Mediation effects
	Alcohol use severity		Drug use severity
	During treatment	Post‐treatment	During treatment	Post‐treatment
Integrated, trauma focused psychotherapy	Full	NS	Full	NS
Non‐integrated, trauma focused psychotherapy	NS	NS	NS	NS
Integrated, non‐trauma focused psychotherapy	NS	NS	NS	NS
SUD psychotherapy	NS	NS	NS	NS
PTSD pharmacotherapy	Full	NS	Partial	NS
AUD pharmacotherapy	Partial	NS	Full	Full
Placebo medication	NS	NS	Partial	NS
Combined trauma focused psychotherapy + AUD pharmacotherapy	Partial	NS	Full	NS
Combined trauma focused psychotherapy + PTSD pharmacotherapy	NS	NS	NS	NS

Note: Treatment types are described further in Hien et al. [21, 27]. Mediation refers to whether reductions in PTSD severity partially or fully mediated the effect of treatment on alcohol or drug use severity during or at post‐treatment. Effects referenced here were statistically significant at P < 0.05.

Abbreviations: AUD, alcohol use disorder, PTSD, post‐traumatic stress disorder; NS, indicates that the mediation effect was not statistically significant; SUD, substance use disorder.

Alcohol use severity

During‐treatment effects

When testing the a path, effects of treatment type on PTSD symptoms were found for AUD pharmacotherapy, PTSD pharmacotherapy and placebo medications in models examining alcohol use severity outcomes during treatment, such that there was a greater reduction in PTSD symptoms during treatment for these treatment types compared to TAU (b = −0.28 to −0.43; SE = 0.10 to 0.14; t = −2.20 to −3.94; P < 0.001–0.042). TF non‐integrated therapies, AUD pharmacology and placebo medication led to greater reductions in alcohol use severity that were not mediated through changes in PTSD (i.e. c’ path direct effects; b = −0.27 to −0.59; SE = 0.10–0.19; t = −2.28 to −3.62; P < 0.001 to 0.023). For the b path, changes in PTSD symptoms predicted changes in alcohol use severity such that during‐treatment reductions in PTSD symptoms were associated with reductions in alcohol use severity (b = 0.20; SE = 0.05; t = 3.82; P < 0.001). Last, several significant h paths were observed. Reductions in PTSD symptoms fully mediated the pathway from treatment type to alcohol use severity for TF integrated therapies (ab _fs  = −0.034) and PTSD pharmacotherapies (ab _fs  = −0.022) and partially mediated the pathways for AUD pharmacotherapies (ab _fs  = −0.016) and TF therapy + AUD pharmacotherapy (ab _fs  = −0.020).

Post‐treatment effects

At post‐treatment, one significant a path emerged such that there was an effect of treatment type on PTSD severity for TF integrated class such that reductions in PTSD symptoms were greater for TAU compared to TF integrated therapies (b = 0.23; SE = 0.10; t = 2.40; P = 0.016). No other significant effects of treatment type emerged at post‐treatment. There was not a significant effect of PTSD symptoms on alcohol symptoms from end of treatment through post‐treatment or evidence of any additional total, mediated nor direct effects. This indicated that the effects observed during treatment persisted through post‐treatment.

Drug use severity

During‐treatment effects

The active treatment conditions on the a path resulting in greater PTSD symptom reduction during treatment compared to TAU were the AUD pharmacotherapy (b = −0.31; SE = 0.10; t = −3.27; P = 0.001), PTSD pharmacotherapy (b = −0.42; SE = 0.10; t = −4.14; P < 0.001) and placebo medication classes (b = − 0.27; SE = 0.13; t = −2.04; P = 0.041). There was an effect of PTSD symptom reductions on drug use severity on the b path such that greater PTSD symptom reductions during treatment were associated with greater drug use severity reduction (b = 0.11; SE = 0.04; t = 2.43; P = 0.015). Following the h path, PTSD severity reduction fully mediated pathways from treatment class to drug severity for AUD pharmacotherapy (ab _fs  = −0.008) and TF psychotherapy + AUD pharmacotherapy classes (ab _fs  = −0.010) and partially mediated during‐treatment drug use severity effects for PTSD pharmacotherapy (ab _fs  = −0.011).

Post‐treatment effects

At post‐treatment, AUD pharmacotherapy (b = −0.25; SE = 0.07; t = −3.77; P < 0.001) resulted in a significant a path, with greater PTSD severity reduction in this class compared to TAU above‐and‐beyond the changes observed in‐treatment. The a path for TF integrated therapy saw increases in PTSD during the post‐treatment period compared to TAU (b = 0.23; SE = 0.09; t = 2.43; P = 0.015) that negated the reductions observed during the during‐treatment period. Further, a treatment‐by‐mediator interaction was observed, where the links between changes in PTSD severity and changes in drug use severity were stronger for TF psychotherapy + AUD pharmacotherapy than TAU (b = 0.16; SE = 0.08; t = 1.96; P = 0.05). Examining the b path, greater post‐treatment reductions in PTSD severity were associated with greater reductions in drug use severity at post‐treatment (b = 0.46; SE = 0.13; t = 3.61; P < 0.001), with post‐treatment mediation effects (above‐and‐beyond during‐treatment mediation) observed for AUD pharmacotherapies.

DISCUSSION

Given numerous treatment options available for PTSD+SUD, this secondary analysis of PH data examined how changes in PTSD symptoms mediate the effects of behavioral, pharmacotherapeutic and combination therapies on AUD and SUD severity. Following from overall findings from the PH primary comparative effectiveness analysis [21, 27], which showed total effects on PTSD and alcohol use severity for many of the treatment classes, the first two questions we posed broadly examined whether PTSD reductions would mediate improvements in alcohol and drug use severity at two different stages—during and/or after treatment. Findings revealed complex and differential effects of the mediating role of PTSD reductions for those with AUDs in comparison to those with drug use disorders.

In response to these research questions, and in keeping with the literature, the present analyses demonstrated mediation between PTSD reductions and alcohol use severity improvements, but these effects were only observed during the in‐treatment phase and not after. Mediation effects of PTSD change on alcohol use severity during treatment were shown by a subset of treatment classes: [TF integrated and PTSD pharmacotherapy (full mediation), AUD pharmacotherapy and TF non‐integrated + AUD pharmacotherapy (partial mediation)]. TF psychotherapies reduced PTSD symptoms, as well as showed a possible indirect pathway by which PTSD symptom reductions exert subsequent changes in drug use severity [37, 39, 53] particularly when integrated with behavioral SUD interventions and/or AUD pharmacotherapies.

Next, we examined mediation between PTSD improvements and reductions in drug severity across specific PTSD+SUD treatments (drawn from the literature) [37, 38, 39, 53, 54]. Here, we examined whether there were mediation effects between PTSD improvement and drug use severity reductions during and/or after treatment. Because the original PH primary comparative effectiveness analyses had not shown any total effects of PTSD+SUD treatments on drug use severity outcomes for any of the treatment classes, and in line with the literature suggesting that drug use severity changes are often dependent on PTSD reductions and may occur temporally later than the PTSD changes (e.g. Hien et al.) [29, 37], we specifically expected that the trauma focused psychotherapies with and without AUD pharmacotherapy (the treatment classes, which PH had shown to be most effective in reducing PTSD and alcohol use severity outcomes overall), would evidence a mediation effect on drug use severity during and after treatment delivery. The findings supported that PTSD symptom reductions lead to changes in drug severity [27, 53, 55, 56, 57]. Importantly, for the drug severity outcomes, the present study findings provided evidence of mediation effects for PTSD changes on drug use severity outcomes both during (similar to alcohol use severity findings) and additional decreases after treatment delivery (in contrast to alcohol use severity findings). PTSD symptom reductions fully mediated pathways to drug use severity improvements for AUD pharmacotherapy, and TF psychotherapy + AUD pharmacotherapy classes during treatment, and AUD pharmacotherapies also showed post‐treatment mediation. This set of analyses extends the comparative effectiveness results of PH [21, 27] by providing a more nuanced understanding of the overall effectiveness of treatments, as well as indirect effects of certain treatments on SUD outcomes via PTSD improvements during and after treatment delivery. We are intrigued that the alcohol effects were partially mediated during treatment, but not after treatment, in comparison to the drug use effects that were fully mediated both during and after treatment. These findings, paired with the overall findings from the original PH analysis, which did not show as many comparative effects for drug use outcomes as for alcohol use outcomes, suggest that drug use severity of all kinds may operate differently than alcohol use among those with PTSD. Additionally, harm reduction approaches are often used in treatment and many impact post‐treatment alcohol use more than drug use. However, more research will be needed to understand if the mechanisms of action are themselves different or if those with drug use have more severe conditions that require more time in treatment to be responsive to its effects. For example, recent PH analyses suggest that when treatment dosage is treated as a post‐treatment covariate/moderator, TF treatments (whether integrated or not) show larger effects for drug use outcomes as treatment dosage increases [49].

For AUD pharmacotherapies, with or without TF psychotherapy, decreases in alcohol and drug use severity during treatment were either partially or fully mediated by PTSD reductions after treatment, and these effects were fully mediated by PTSD reductions during treatment. For drug use severity, these mediation effects were also present in the post‐treatment analysis. Although AUD pharmacotherapy, with or without TF psychotherapy, also showed effects on PTSD and alcohol use severity in the comparative effectiveness analysis [27], none showed significant total effects (i.e. zero‐order effects independent of mediation) on drug use severity in the previous report. Following other findings [e.g. Hien et al. and Morgan‐Lopez et al. (in press)] [29, 37, 49], those with drug use disorders specifically may need either a longer or more intensive course of treatment targeting their PTSD to support their subsequent drug use improvements. These findings provide strong support for the benefits of TF models, which can target PTSD symptom reduction. It appears that the use and/or addition of AUD pharmacotherapy synergizes to support these needed PTSD symptom reductions and may also support the psychotherapy itself. Accordingly, present findings continue to underscore the importance of addressing and processing trauma‐related memories and stimuli during PTSD+SUD treatment. The addition of pharmacotherapy may enhance these outcomes. Further research to examine these mechanisms is warranted.

Changes in PTSD symptoms also fully mediated the effects of PTSD pharmacotherapy on alcohol use severity during treatment, supporting the connection between improvements in PTSD symptoms and subsequent reductions in alcohol use severity. However, the impact of PTSD pharmacotherapy and placebo medication on drug use severity outcomes differed from the three treatment types described above (AUD pharmacotherapy, TF psychotherapy and their combination). If these treatments improved PTSD symptoms, a similar pathway of mediation was observed. However, for patients with PTSD pharmacotherapy and placebo whose PTSD symptoms did not improve their drug use severity outcomes worsened. This raises questions about the factors influencing this contrast. For PTSD pharmacotherapies, if the treatment is ineffective in improving PTSD symptoms (considered ‘treatment‐resistant’), participants tended to use more substances compared to those in TAU. One interpretation of these findings is that patients may experience distress or hopelessness related to a lack of symptom reduction, which leads to increased drug use consistent with the self‐medication model [32]. This may be more likely to occur when receiving pharmacotherapies, which may lack non‐specific behavioral therapeutic factors present in TAU or even create side effect related stress. In the primary outcome analysis [27], participants who received non‐TF or SUD psychotherapies, on average, neither deteriorated nor improved relative to TAU. Compared to AUD pharmacotherapy, TF integrated psychotherapy and their combination, there were no mediation effects found for the non‐TF or SUD psychotherapies on either alcohol or drug use severity. Findings from the present study continue to demonstrate that non‐TF and SUD‐only focused interventions have modest impact compared to TF integrated treatments. It is possible that the effects derived from these intervention types do not necessarily involve PTSD symptom reductions as a potential mediator on substance use outcomes. Additionally, the discrepant findings related to the mediation effects of PTSD for non‐TF treatment and their combination with AUD pharmacotherapy suggest that AUD medications may be needed to activate mechanisms of change for non‐TF PTSD+SUD treatment. Specifically, preliminary evidence supports the addition of medications that target substance use craving (naltrexone, zonisamide) [58, 59] and may point to craving as a candidate mechanism for PTSD+SUD treatment outcomes. However, more research is warranted.

There are methodological limitations to this work. PH studies were categorized based on primary substance of choice. That is, many adults enrolled in PTSD+AUD trials endorsed alcohol as their primary substance of choice while also reporting regular use of other substances or meeting criteria for another SUD. Therefore, findings regarding alcohol use outcomes derived from this harmonized sample may not generalize to populations with PTSD+AUD reporting no other drug use. Further, PH only included pharmacotherapy studies that targeted alcohol. Future research is needed to examine how the results demonstrated here might generalize to other pharmacotherapies for SUD and pharmacotherapy–psychotherapy combinations that target PTSD+SUD, particularly opioid use disorder.

Moreover, the target of the pharmacotherapies was determined by the intended target of the treatment (e.g. reduce PTSD severity, alcohol or drug use severity) as described in the publication. Therefore, not all pharmacotherapies investigated are United States Food and Drug Administration‐approved or considered efficacious agents at this time. Additionally, we specified binary indicators for the six substances that comprised the latent substance use outcome. However, as many psychotherapies for SUD conceptualize treatment goals and progress using a harm reduction framework that captures greater complexity in treatment response, future research is needed to examine how change in PTSD symptoms impacts change in substance use using ordered categorical substance use indicators, particularly as evidence demonstrates that even modest shifts in alcohol use are associated with positive health outcomes [60, 61]. TAU, although classified as treatments that did not have evidence bases for impacting PTSD or substance use, had conditions that varied across studies and included waitlist, behavioral therapies and community mental health treatment. Although this heterogeneity is expected across this large number of harmonized studies, such heterogeneity is addressed with random effects across studies. The present analysis cannot address treatment specific differences in how PTSD symptoms are reduced (e.g. by addressing guilt‐related cognitions, extinction learning, social cognitions), and causal mediation would need to be examined with a different set of analyses, which could model mediator and outcome at non‐contemporaneous timepoints (e.g. studies that include a longer series of during‐treatment timepoints could help address this limitation). Finally with regards to generalizability of these findings, because the original universe of studies (completed before the opioid crisis, pre‐2019) included in PH did not include: (1) any of the newly conducted clinical trials with medications for opioid use disorder; (2) specific populations such as female veterans, civilian and/or incarcerated men; and (3) the kinds of treatments available differed depending on veteran or civilian populations, conclusions applied to these missing populations should be made with some caution.

Future research directions and clinical and policy implications

The current study carries implications for PTSD+SUD clinical care and research. Understanding the mechanisms by which certain treatments ameliorate both PTSD and SUD severity allows for more targeted interventions and individualized standards of care. Findings may inform provider considerations regarding potentially integrating TF PTSD+SUD approaches and monitoring treatment response. The effectiveness of the integrated TF models supports the value of addressing both PTSD and substance use within the same treatment protocol. This integrated TF approach may lead to more comprehensive and sustained improvements in both PTSD symptoms and substance use severity—both alcohol and drug use severity. Findings underscore the importance of considering pharmacotherapies in the management of PTSD+SUD and the need for more research to identify effective pharmacotherapy approaches. However, the modest drug use severity response observed for PTSD medications calls for careful consideration and further exploration. The analysis sheds light on the complex relationship between placebo medications, treatment resistance and substance use outcomes. Understanding why improvements in PTSD symptoms with certain treatments do not consistently lead to positive substance use outcomes is crucial for refining treatment strategies. Identifying factors contributing to treatment non‐responsiveness can guide clinicians in adjusting therapeutic approaches. Future work should aim to examine differences in treatment response as a function of socio‐demographic characteristics (e.g. veteran status, sex) and clinical presentation (e.g. trauma type, primary substance) to further elucidate, which interventions are most efficacious and for whom. Clinicians should be vigilant in monitoring changes in PTSD symptoms and substance use severity throughout treatment.

CONCLUSION

This study delineates the differential effects of various treatment types on PTSD and SUD severity during and post‐treatment, providing insights into the mediating roles of PTSD symptom changes in influencing alcohol and drug use treatment outcomes. The findings highlight the need for continued research exploring the mechanisms underlying treatment responses to inform the development of more effective interventions and contribute to the ongoing refinement of evidence‐based practices for PTSD+SUD. By understanding the mechanisms of treatment effectiveness, clinicians can enhance their ability to provide tailored interventions, ultimately improving outcomes for individuals facing these dual challenges.

AUTHOR CONTRIBUTIONS

Denise Hien: Conceptualization (lead); funding acquisition (lead); investigation (equal); methodology (equal); project administration (equal); resources (equal); writing—original draft (lead); writing—review and editing (lead). Jordan A. Gette: Formal analysis (supporting); visualization (equal); writing—original draft (equal); writing—review and editing (equal). Shannon Blakey: Formal analysis (supporting); methodology (supporting); visualization (equal); writing—original draft (equal); writing—review and editing (equal). Marilyn Piccirillo: Visualization (equal); writing—original draft (equal); writing—review and editing (equal). Sudie E. Back: Conceptualization (equal); funding acquisition (equal); investigation (equal); project administration (equal); writing—original draft (equal); writing—review and editing (equal). Alexandria Bauer: Conceptualization (equal); writing—original draft (equal); writing—review and editing (equal). Chantel Tannaz Ebrahimi: Data curation (equal); project administration (equal); writing—review and editing (equal). Robyn Ellis: Writing—original draft (equal); writing—review and editing (equal). Therese K. Killeen: Conceptualization (equal); funding acquisition (equal); investigation (equal); methodology (equal); project administration (equal); writing—original draft (equal); writing—review and editing (equal). Elizabeth A. Lehinger: Writing—original draft (equal); writing—review and editing (equal). Teresa López‐Castro: Conceptualization (equal); funding acquisition (equal); investigation (equal); project administration (equal); writing—review and editing (equal). Sonya Norman: Conceptualization (equal); investigation (equal); methodology (equal); project administration (equal); writing—original draft (equal); writing—review and editing (equal). Lesia M. Ruglass: Conceptualization (equal); funding acquisition (equal); investigation (equal); methodology (equal); project administration (equal); writing— original draft (equal); writing—review and editing (equal). Tanya Saraiya: Writing—original draft (equal); writing—review and editing (equal). Lissette Saavedra: Conceptualization (equal); data curation (equal); funding acquisition (equal); investigation (equal); methodology (equal); project administration (equal); writing—review and editing (supporting). Antonio A. Morgan‐López: Conceptualization (equal); data curation (lead); formal analysis (lead); funding acquisition (equal); investigation (equal); methodology (equal); project administration (equal); writing—review and editing (equal).

DECLARATION OF INTERESTS

All other authors report no conflicts of interest.

Supporting information

Data S1. Supplementary Information.

Hien DA, Gette JA, Blakey SM, Piccirillo ML, Back SE, Bauer AG, et al. How changes in post‐traumatic stress disorder (PTSD) severity mediate substance use disorder (SUD) severity during and after treatment for co‐occurring PTSD and SUD: Results from Project Harmony. Addiction. 2025;120(11):2245–2257. 10.1111/add.70126

DATA AVAILABILITY STATEMENT

The data that support the findings of this study are available from the corresponding author upon reasonable request.