Phase-based versus non-phase-based psychological interventions for complex PTSD: a systematic review and meta-analysis

ABSTRACT

Background: Since complex post-traumatic stress disorder (CPTSD) was incorporated to the International Classification of Diseases-11th revision (ICD-11), the construct has drawn increasing attention in clinical research due to its debilitating effects across multiple domains of functioning. However, whether stabilization (Phase 1) is necessary before exposure to traumatic memories (Phase 2) remains controversial.

Objective: We examined whether phase-based interventions are more effective than non-phase-based interventions in reducing symptoms of CPTSD.

Methods: On 17 June 2025, we searched the Cochrane Library, Embase, Medline, PTSDpubs, Scopus, and Web of Science as well as South Korean databases, including the Korean Citation Index, the Research Information Sharing Service, DataBase Periodical Information Academic (DBPia), and ScienceON. Randomized controlled trials of psychological interventions for individuals with clinically significant CPTSD symptoms or PTSD with at least two of the ‘disturbances in self-organization’ (DSO) symptom clusters were eligible. We used Hedges’ g to pool effect sizes and performed subgroup analyses using Cochran’s Q test. We assessed risk of bias using the Cochrane Risk-of-Bias Assessment 2 tool.

Results: Across most outcomes, we observed no significant differences between phase- and non-phase-based interventions, between multi-phase and single-phase interventions, or between exposure and non-exposure interventions. However, multi-phase interventions had superior effects on PTSD, while phase-based, multi-phase, and exposure interventions showed greater improvements in DSO – affect regulation compared to their counterparts.

Conclusions: This study was limited by the small number of available trials, single time-point, and heterogeneity in outcome measures, which constrained both the detection of subgroup differences and the precision of effect size estimates. Our findings provide empirical evidence for the current discourse, suggesting that non-phase-based and non-exposure-based interventions can be as effective as structured approaches in many contexts, while some outcomes may still benefit from phase-, multi-modular-, or exposure-based designs.

HIGHLIGHTS

• This systematic review examined whether interventions that include a first phase of stabilization followed by a second phase of exposure to traumatic memories are more effective than other interventions.

• Across most outcomes, there were no clear differences between phase- and non-phase-based interventions, or between exposure and non-exposure interventions.

• For post-traumatic stress symptoms, interventions with more than one phase were more effective than those with a single-phase treatment. For difficulties with emotion regulation, multi-phase interventions that included stabilization before exposure showed greater benefits.

1. Introduction

Since the International Classification of Diseases-11th revision (ICD-11; WHO, 2022) – which includes diagnostic criteria for complex post-traumatic stress disorder (complex PTSD or CPTSD) – was released, the construct of CPTSD has garnered attention in both clinical practice and academic research. Researchers have advocated for a differentiated classification of ‘complex’ PTSD regarding its lifelong nature and the impairment of self-regulatory abilities compared to single-incident traumas (Cloitre et al., 2013; Herman, 1992a).

According to ICD-11 criteria, in addition to PTSD symptoms (i.e. re-experiencing, avoidance, and a sense of threat), CPTSD diagnosis requires symptoms from each of the three ‘disturbances in self-organization’ (DSO) symptom clusters (i.e. affect dysregulation, negative self-concept, and disturbed relationships) (WHO, 2022).

A recent meta-analysis estimated the global prevalence of CPTSD to be 6.2% (Huynh et al., 2025). In addition to its considerable prevalence, CPTSD has negative ramifications across multiple functional domains. Compared to PTSD, CPTSD has a unique relationship with higher suicide risk (Chong et al., 2024; Jannini et al., 2023), less perceived social support (Simon et al., 2019), and lower working capacity (Brenner et al., 2019). Previous studies have called for greater clinical efforts tailored to the specific features of CPTSD. However, how and when to intervene remain controversial in the clinical field. It is still unclear whether evidence-based PTSD interventions can be generalized to CPTSD, particularly to DSO symptom clusters. Would evidence-based interventions for PTSD be sufficient for treating CPTSD? Or would some other ‘complex’ treatment work for a ‘complex’ form of PTSD?

Previous meta-analyses have investigated the efficacy of psychological interventions for CPTSD (Choi et al., 2020; Darby et al., 2023; Hu et al., 2025; Karatzias et al., 2019). According to these studies, evidence-based interventions such as cognitive processing therapy (CPT), trauma-focused cognitive behavioural therapy (TF-CBT), narrative exposure therapy (NET), prolonged exposure therapy (PE), and eye-movement desensitization and reprocessing (EMDR) have moderate-to-large effect sizes on trauma-related symptoms of CPTSD (Hedge’s g or Cohen’s d ≈ 0.7–1.3). Improvements in DSO symptoms have also been observed; however, the number of studies is relatively small.

Herman (1992b) first mentioned the need for a phase-based approach to CPTSD when the concept was not yet standardized or clarified. In 2012, the ‘ISTSS Expert Consensus Guidelines for CPTSD’ were released, presenting clinical and empirical knowledge on CPTSD (Cloitre et al., 2012). These guidelines recommend the following three phases. The initial safety and stabilization phase (Phase 1) usually encompasses psychoeducation, stress management, and emotion regulation strategies. Without stabilization, the affected individual might experience disengagement (e.g. dissociation, dropping out) or deterioration in well-being (Cloitre et al., 2020). Phase 2 involves the direct reappraisal and reprocessing of traumatic memories. Typical exposure-based therapies (e.g. PE) are considered in Phase 2. Phase 3 consists of reintegration and reconnection with one’s future life and society. This phase-based intervention has been effective for people with complex trauma, especially those with a history of childhood abuse (Cloitre et al., 2010; Dauber et al., 2015) and veterans (Wiedeman et al., 2020).

Some studies have claimed that it may be ineffective (and even unethical) to delay or restrict Phase 2 (De Jongh et al., 2016). Randomized controlled trials (RCTs) comparing phase-based interventions (e.g. TF-CBT, Skills Training in Affective and Interpersonal Regulation and Narrative Therapy [SNT]) with non-phase-based interventions (e.g. PE and EMDR) have not yet distinguished the superior clinical outcomes of phase-oriented therapy. If there is no strong evidence of conducting stabilization phase, this unnecessary process may hinder patients from receiving effective, evidence-based, trauma-focused interventions (De Jongh et al., 2016).

To address this controversy, it is essential to compare the effectiveness of these two types of interventions in supporting clinical decision-making.

Darby et al. (2023) conducted a systematic review of phase-based interventions related to CPTSD using the ICD-11 criteria. They found that phase-based therapies are generally more effective than treatment-as-usual (TAU) or waitlist conditions and, in some cases, are better than even unimodal exposure-based therapies. However, the relative efficacy of phase-based versus non-phase-based approaches remains unclear. Moreover, the results of prior studies may reflect selection bias since most of them excluded potential participants with comorbid borderline personality disorder (BPD), substance abuse, bipolar disorder, or high suicide risk. Thus, the study population may have already possessed sufficient emotional regulation capacity to tolerate exposure-based interventions in Phase 2 without preliminary stabilization.

Corrigan et al. (2020) evaluated the effectiveness of phase-based treatment models for PTSD and DSO symptoms. The 13 phase-based therapies included in this study demonstrated a large overall effect size, with a mean Cohen’s d of 1.77 and a 95% confidence interval of [1.46, 3.08] (p < .001). This large effect size was also observed in the subgroup analysis of high-complexity samples, such as individuals with childhood trauma. However, the authors emphasized that the definition of ‘complex trauma’ and specific phase components should be further clarified.

Dyer and Corrigan (2021) also pointed that there is currently no standardized definition of ‘phase-oriented therapy,’ and that considerable heterogeneity of treatment elements exists across studies. Other studies and guidelines also claim that this issue warrants further investigation (Corrigan et al., 2020; NICE, 2018). The lack of consensus may hinder precise and rigorous evidence of the effectiveness of phase-based interventions. Furthermore, it is necessary to move beyond an all-or-none distinction between phase- and non-phase-based approaches and specify which aspects of a given phase (e.g. the multiplicity of phases, sequencing Phase 1 before Phase 2, or the exposure effect) have contributed to treatment effects.

To our knowledge, this is the first review to directly compare the effectiveness of phase- and non-phase-based interventions for CPTSD using the clearly defined ICD-11 criteria. The present systematic review and meta-analysis aimed to

• compare the efficacy of phase-based versus non-phase-based interventions for ICD-11 CPTSD symptoms, including PTSD and DSO;

• determine whether specific treatment elements – namely the number of phases (multi-phase vs. single-phase) and the inclusion of exposure components (exposure vs. non-exposure) – influence treatment outcomes; and

• provide empirical evidence to inform clinical decision-making regarding optimal treatment elements and sequencing for CPTSD.

2. Methods

2.1. Protocol registration

We registered this study with PROSPERO (CRD420251089272) on 22 July 2025, after completing the literature search (17 June 2025) but before extracting and analysing data.

2.2. Search strategy and study selection

The databases used in the search encompassed the Cochrane Library, Embase, Medline, PTSDpubs, Scopus, and Web of Science (Wos) as well as South Korean databases including the Korean Citation Index, the Research Information Sharing Service, DataBase Periodical Information Academic (DBpia), and ScienceON. The core search strings were as follows: (complex post-traumatic stress disorder OR complex posttraumatic stress disorder OR complex post traumatic stress disorder OR complex PTSD OR cPTSD OR CPTSD) AND (therapy OR psychological therapy OR psychological intervention OR intervention OR treatment) AND (randomized, or randomised, or randomized controlled trial OR randomized controlled trial OR RCT OR clinical trial). The search terms were adapted for each database and the search was performed on 17 June 2025.

Studies were eligible if the participants had been diagnosed with CPTSD using validated measures such as the International Trauma Questionnaire (ITQ) or the Structured Interview for Disorders of Extreme Stress (SIDES, clinician version or self-report). However, given that CPTSD has not been diagnosed according to official diagnostic criteria for a long time, most studies have not employed such measures. Hence, we adopted the methodology proposed by Karatzias et al. (2019). That is, if the population in a study had been diagnosed with PTSD according to the ICD or DSM (Diagnostic and Statistical Manual of Mental Disorders) and the research reported clinically significant levels in at least two DSO symptom clusters-affect dysregulation and disturbances in relationships – with alternative validated measures, the study was deemed eligible. Negative self-esteem was not a mandatory criterion.

We made this decision based on two reasons. First, most of the studies did not report outcomes related to self-esteem. Thus, requiring all three DSO symptom clusters would have yielded only a small number of studies for the meta-analysis. Second, research on DSO symptom clusters has revealed that negative self-esteem has a relatively weak correlation with PTSD symptoms and low factor loadings compared with other clusters (Kindred et al., 2025; Melamed et al., 2024). Nevertheless, a given population may differ from one that meets the full criteria for CPTSD, as described in detail in the limitations of the present study.

Other eligibility criteria included RCTs published in English. We allowed for comorbidities such as BPD and substance abuse because they are highly prevalent in cases of CPTSD (Darby et al., 2023). In terms of treatment modalities, group and online therapy were also included. For online therapy studies, we included only if they were structured, protocolized, and provided sufficient details about the intervention to a degree comparable to that of face-to-face interventions.

Exclusion criteria comprised studies focusing on neurodegenerative disorders, acquired and/or traumatic brain injury, or primary substance use disorder; studies with insufficient information regarding CPTSD assessment; case reports; uncontrolled trials; crossover trials; and grey literature.

Two researchers independently assessed the studies’ eligibility. The inter-rater reliability, evaluated using Cohen’s kappa value, was substantial (k = 0.66; Landis & Koch, 1977). Any disagreements were resolved via consensus.

2.3. Outcomes and data extraction

The primary outcome was PTSD symptoms, and the secondary outcomes were the DSO symptom clusters (i.e. affect dysregulation, disturbances in relationships, and negative self-concept) as well as total CPTSD symptoms. As the studies reported different follow-up time points, we included only pre- and post-treatment outcomes. This will be further discussed in the limitations section. We extracted and stored data related to the study characteristics and statistics for the meta-analysis in a spreadsheet. We resolved any disagreements during this process through discussion.

2.4. Analysis

We performed all analyses using statistical software for data science (Stata Now)/MP 19.5. Hedge's g was calculated using the post-intervention scores of the intervention and comparator groups. When studies reported multiple time points, we prioritized the immediate post-treatment assessment. Follow-up analyses were not conducted because not all studies reported follow-up data, and the time points varied across studies. We followed general conventions to interpret individual and pooled effect sizes (0.2 = small, 0.5 = moderate, 0.8 = large, Cohen, 1988). Each study used and reported different outcome variables; for example, some studies reported only the total CPTSD score, while others reported all relevant scores (PTSD, DSO). As proposed by Borenstein et al. (2009), to address the issue of multiple outcomes within a study, we calculated composite effect sizes by combining PTSD and any available DSO scores. We computed Hedge’s g by combining the standardized effect sizes of each outcome; we estimated between-outcome correlations by referring to other meta-analytic studies. For PTSD-DSO, we identified Kindred et al. (2025); for PTSD-affect dysregulation, Jones et al. (2018). For PTSD-disturbances in relationships, we identified Cloitre et al. (2021). For PTSD-negative self-concept, we turned to Melamed et al. (2024).

Since each study employed different types of comparators – such as a waitlist, TAU, or an active control group – pooling all of them together could have created bias against the active control conditions. Thus, we planned two sets of comparisons: (1) intervention versus waitlist/TAU and (2) a ‘head-to-head comparison’ where all intervention groups were phase-based and all active control groups were non-phase-based. In studies with more than three arms requiring multiple comparisons, we divided the comparator group sample size proportionally to avoid double counting (Higgins & Green, 2011).

Publication bias was analysed by visually inspecting funnel plots, using Egger’s test (Egger et al., 1997), and employing the trim-and-fill method (Duval & Tweedie, 2004). We conducted statistical tests for publication bias only when at least ten studies were available for a given outcome (Higgins & Green, 2011).

The heterogeneity of each outcome was examined using the I² statistic and followed the criteria proposed by the Cochrane Handbook (<40% = low, 30–60% = moderate, 50–90% = substantial, and 75–100% = considerable) (Higgins & Green, 2011).

2.5. Assessment of study quality

We assessed the study quality using the Cochrane Collaboration Risk of Bias 2 (RoB 2) tool (Higgins & Green, 2011). Since strict blinding is not feasible in psychotherapy trials, we rated studies as blinded if the study had active or TAU control groups, ensuring comparable expectations regarding intervention effectiveness across conditions (Munder & Barth, 2018). Two authors independently assessed study quality, and we resolved disagreements through ongoing discussions.

2.6. Subgroup analysis

To determine the heterogeneity in effect size estimates between subgroups of comparison categories (phase vs. non-phase, multi-phase vs. single-phase, and exposure vs. non-exposure), we used Cochran’s Q test. We performed subgroup analyses only when at least two studies were available for each subgroup.

The definitions used to distinguish each subgroup are as follows:

2.6.1. Phase-based or non-phase-based interventions

In order to classify interventions as phase-based or non-phase-based, we adopted the definition established by Cloitre et al. (2012). We thoroughly reviewed intervention protocols and coded studies as phase-based if they included a stabilization phase (Phase 1) prior to trauma exposure (Phase 2), with or without a subsequent reintegration phase (Phase 3). All others were classified as non-phase-based.

2.6.2. Multi-phase or single-phase intervention

Treatments that included more than one phase (i.e. Phase 1 +Phase 2, Phase 2 + Phase 3, or Phase 1 + Phase 3) were coded as multi-phase, whereas treatments consisting of only a single phase were coded as single-phase.

The distinction between phase- and non-phase-based interventions is rooted in whether the intervention follows the prescribed sequence of Phase 1 (stabilization) followed by Phase 2 (trauma-focused treatment). In contrast, the classification of multi-phase versus single-phase interventions depends solely on the number of phases included. Accordingly, non-phase-based interventions may be either multi-phase (e.g. Phase 1 + Phase 3 or Phase 2 + Phase 3) or single-phase (Phase 1, Phase 2, or Phase 3 alone). This distinction allowed us to explore whether the treatment effect of multi-phase, phase-based interventions arises from the prescribed sequence of phases or merely from the number of phases included.

2.6.3. Exposure or non-exposure interventions

Treatments were classified as ‘exposure-based’ if re-experiencing or reappraisal of traumatic memories was included in the procedure. For instance, PE therapy and EMDR are considered exposure-based interventions.

3. Results

3.1. Study selection

Of the 1,642 studies identified through the database search, we removed 503 duplicates and excluded 847 based on title and abstract screening, resulting in 292 studies for full-text review. Of these, we excluded 271, primarily because they did not report CPTSD-related outcomes. We excluded five additional articles due to invalid outcomes (k = 1), improper population (k = 3), and improper comparators (k = 1). As a result, total 16 studies met the inclusion criteria. Figure 1 presents the flow of including and excluding studies as well as the reasons for each instance of exclusion.

Figure 1.

PRISMA flowchart.

Tables 1 and 2 outline the study characteristics. We included 16 studies on psychological interventions targeting CPTSD. Eleven studies were conducted in European countries (the Netherlands, Germany, Norway, England, Lithuania, Austria, Sweden, and Portugal), two in North America, and three in East Asia (South Korea). Five studies included female participants only. Female outnumbered male in most studies, except for one that targeted veterans.

Table 1.

Study Characteristics (Intervention vs. Waitlist and Treatment-as-usual comparison).

Study	Country	Population	Psychiatric history/medication	Measurement	Treatment Group					Comparator Group			Context
					Type of Treatment	Age M(SD)	N (%Female)	Session	Phase	Comparator	Age M(SD)	N (%Female)
Jeong and Kim (2022)	Korea	Current child-rearing mothers with childhood abuse history	N.A.	CTQ, CTI, Posttraumatic Contextualized Memory Questionnaire, Traumatized Identity Questionnaire, DERS, IPBI, Posttraumatic Emotion Scale, Dysfunctional Parenting Thoughts, Iowa Parent Behaviour Inventory	Trauma-focused Cognitive-behavioural therapy (TF-CBT)	36.33 (4.49)	6 (100)	12	Phase 1 + Phase 2 + Phase 3	No Treatment	35.67 (4.11)	6 (100)	Community
Park and Park (2023)	Korea	University students with complex trauma and emotional dysregulation	N.A.	CTI, FI, DERS, Anxiety Control Questionnaire, Relationship Change Scale	Emotion-focused group therapy	N.A.	9 (77.8)	10	Phase 1 + Phase 2	Waitlist	N.A.	9 (77.8)	University
You and Son (2018)	Korea	University Students with childhood abuse history	N.A.	Child abuse Scale, K-PTGI, AAQ-II, SIDES-SR	Acceptance and Commitment Therapy (ACT)	N.A.	7 (85.7)	8	Phase 1 + Phase 3	No Treatment	N.A.	7 (42.9)	University
Classen et al. (2020)	Canada	Women with childhood abuse experience	N.A.	CTQ-SF, LSCL-R, SBC, PCL-C, SDQ-20, DES, BDI-II, BAI, PHLMS, IIP-32, SRS	Body-oriented group therapy	43.51 (10.01) a	14 (100)	20	Phase 1 + Phase 3	WLC	43.51 (10.01) a	14 (100)	Hybrid (Hospital programme + community)
Cloitre et al. (2002)	USA	Women with PTSD related to childhood abuse	Axis I disorders and Axis II borderline personality disorder.	SCID-I, NMR, Ax/Ex, TAS-20, DISS, BDI, STAI-S, IIP-32, SAS-SR, ISEL, WAI	Skills Training in Affective and Interpersonal Regulation – modified Prolonged Exposure (STAIR-modified PE)	34 (7.22) a	31 (100)	16	Phase 1 + Phase 2	WLC	34 (7.22) a	27 (100)	Community
Dorrepaal et al. (2012)	Netherlands	People with experience of sexual and/or physical abuse before the age of 16	DSM-IV axis I disorders, comorbid personality disorders	Davidson Trauma Scale, SIDES, BPDSI, DES	Stabilizing group treatment	40.3 (10.7)	38 (100)	20	Phase 1 + Phase 2 + Phase 3	TAU	37.1 (10.3)	33 (100)	Outpatient
Dumarkaite et al. (2021)	Lithuania	University students exposed to various traumatic experiences	N.A.	ITQ, PHQ-9, GAD-7, Positive Mental Health Scale, User Satisfaction and Programme Usability	Mindfulness-Based Internet Intervention	23.06 (2.85)	31 (77.4)	8	Phase 1 + Phase 3	WLC	23.56 (3.32)	39 (94.9)	University
Knefel et al. (2022)	Austria	Dari-speaking Afghan asylum seekers or refugees	N.A.	GHQ-28, PMLDC, ITQ, WHOQOL-BREF, PSYCHLOPS, Immigrant Integration Index	Adapted version of Problem Management Plus (aPM+)	29.9 (10.7)	26 (38.5)	6	Phase 1 + Phase 3	TAU	38.8 (14.9)	25 (60)	Specialized facility
Karatzias et al. (2024)	England	Veterans	N.A.	ITQ, PHQ-9, GAD-7, PHQ-15, AUDIT-C, Adverse Events Questionnaire	Enhanced Skills Training in Affective and Interpersonal Regulation (ESTAIR)	47.14 (12.24)	28(3.6)	25	Phase 1 + Phase 2 + Phase 3	TAU	46.32 (10.47)	28(10.7)	Outpatient
Moreira et al. (2022)	Portugal	Intimate partner violence victims	N.A.	ITQ, PHQ-9, CVES Research Version, LEC-5	Cognitive-Narrative Therapy	37.00 (13.57)	8 (100)	4	Phase 2	TAU (but without any specific intervention targeting those symptoms)	41.87 (10.98)	9 (100)	Special protection shelter house (SH)
Raabe et al. (2022) 1	Netherlands	People with a history of repeated childhood abuse before the age of 15.	N.A.	CAPS-4, PDS, BDI-II, DISS, DERS, IIP-32	ImRs	35.4 (10.7)	21 (85.71)	16	Phase 2	WLC	35.5 (11.8)	20 (90)	Outpatient
Raabe et al. (2022) 2	Netherlands	People with a history of repeated childhood abuse before the age of 15.	N.A.	CAPS-4, PDS, BDI-II, DISS, DERS, IIP-32	STAIR	36.8 (10.3)	20 (90)	8	Phase 1	WLC	35.5 (11.8)	20 (90)	Outpatient
Rudstam et al. (2022)	Sweden	Women with PTSD/CPTSD symptoms	N.A.	PCL-5, PCL-5-DSO (Proxy measure), DES, SDQ-20, HSCL-25, POSOMS	Trauma-focused group music and imagery therapy (TFGrpMI)	45.2 (10.7)	22 (100)	12	Phase 1 + Phase 2 + Phase 3	WLC	42.2 (9.10)	23 (100)	Community
Wagner et al. (2024)	Germany	Survivors of institutional childhood abuse	N.A.	ITQ, PHQ-9, GAD-7	Internet-based written imagery rescripting intervention	54.52 (8.25)	38 (76.3)	6	Phase 2 + Phase 3	WLC	54.60 (9.15)	34 (58.8)	Community

AAQ-II: Acceptance & Action Questionnaire-II, AAQ-II; AUDIT-C: Alcohol Use Disorder Identification Test; Ax/Ex: Affective Expressiveness/alexithymia; BAI: Beck Anxiety Inventory; BDI: Beck Depression Inventory; BDPSI: Borderline Personality Disorder Severity Index; CAPS: Clinician-Administered PTSD Scale; CTI: Complex Trauma Inventory; CTQ:  Childhood Trauma Questionnaire; DES: Dissociative Experiences Scale; DERS: Difficulties in Emotional Regulation Scale; DISS: Dissociation Questionnaire; FI: Functional Impairment; GAD-7: General Anxiety Disorder-7; GHQ-28: General Health Questionnaire; HSC-25: Hopkins Symptom Checklist-25; IBPI: Irrational Performance Beliefs Inventory; IIP: Inventory of Interpersonal Problems; ImRs: Imagery Rescripting; ISEL: Interpersonal Support Evaluation List; ITQ: International Trauma Questionnaire; K-PTGI: Korean Version Posttraumatic Growth Inventory; LEC-5: Life Events Checklist for DSM-5; LSCL-R: Life Stressor Checklist-Revised; NMR: Negative Mood Regulation Expectancies Scale; PHQ-9: Patient Health Questionnaire; PCL-5: PTSD Checklist for DSM-5; PE: Prolonged Exposure; PHLMS: Philadelphia Mindfulness Scale; PMLDC: Post-Migration Living Difficulties Checklist; POSOMS; Positive State of Mind Scale; PSYCHLOPS: Psychological Outcome Profiles; PTGI: Post-traumatic Growth Inventory; SAS-SR: Social Adjustmetn Scale; SBC: Scale of Body Connection; SCID: Structured Clinical Interview; SIDES-SR: Structured Interview of Disorder of Extreme Stress–Self Report; STAI-S: State-Trait Anxiety Inventory; SDQ: Somatoform Dissociation Questionnaire; SRS: Soothing Receptivity scale; STAIR: Skills Training in Affective and Interpersonal Regulation; TAS-20: Toronto Alexithymia Scale – 20 items; TAU: Treatment as usual; WAI: Working Alliance Inventory; WLC: Waitlist Control; WHOQOL-BREF: Brief version of WHO Quality of Life scale. ^aThese studies did not report each group’s mean and S.D. age but the total statistics.

Table 2.

Study Characteristics (Head-to-head comparison).

Study	Country	Population	Psychiatric history/medication	Measurement	Treatment 1					Treatment 2					Treatment 3					Context
					Type	Age M(SD)	N (%)	Session	Phase	Type	Age M(SD)	N (%)	Session	Phase	Type	Age M(SD)	N (%)	Session	Phase
Sele et al. (2023)	Norway	People exposed to childhood trauma	Comorbid axis I diagnoses	CAPS, ITQ, PCL-5, BDI, IIP-64	SNT	44.3 (9.8)	30	18	Phase 1 + Phase 2	PE	42.5 (8.8)	32	8∼16	Phase 2	STAIR	41.7 (9.8)	30	10	Phase 1	Outpatient
Oprel et al. (2021)	Netherlands	People with a history of childhood abuse before the age of 18	N.A.	CAPS-5, PCL-5, DERS, IIP-32, Rosenberg Self-esteem Scale	PE	34.52 (11.05)	48 (77.1)	16	Phase 2	iPE	38.87 (11.57)	51 (74.5)	14	Phase 2	STAIR + PE	37.07 (12.39)	50 (78.0)	16	Phase 1 + Phase 2	Outpatient
Raabe et al. (2022)	Netherlands	People with a history of repeated childhood abuse before the age of 15.	N.A.	CAPS-4, PDS, BDI-II, DISS, DERS, IIP-32	ImRs	35.4 (10.7)	21 (85.71)	16	Phase 2	STAIR + ImRs	36.8 (10.3)	20 (90)	24	Phase 1 + Phase 2	WLC	35.5 (11.8)	20 (90)	-	-	Outpatient
van Vliet et al. (2021)	Netherlands	victim of repeated sexual and/or physical abuse before the age of 18	N.A.	CAPS, SIDES-R, IIP, DERS, PTCI, DES-II, BSI	STAIR-EMDR	N.A.	57 (33.1)	24	Phase 1 + Phase 2	EMDR	N.A.	64 (35.5)	16	Phase 2	-	-	-	-	-	Outpatient

BDI: Beck Depression Inventory; BSI: Brief Symptom Inventory; CAPS: Clinician-Administered PTSD Scale; DERS: Difficulties in Emotional Regulation Scale; DISS:  Dissociation Questionnaire; EMDR: Eye-Movement Desensitization and reprocessing; SIDES: Structured Interview of Disorder of Extreme Stress; SNT: STAIR Narrative Therapy; STAIR:  Skills Training in Affective and Interpersonal Regulation; IIP: Inventory of Interpersonal Problems; ImRs: Imagery Rescripting; iPE: Intensified Prolonged Exposure; PCL-5: PTSD Checklist for DSM-5; PE = Prolonged Exposure.

To measure CPTSD, six studies used the ITQ, which directly reflects the ICD-11 CPTSD criteria. Three studies employed the SIDES, which were published before the ICD-11 was released. To gauge DSO symptoms, one study (Rudstam et al., 2022) relied on an indirect measure by selecting some items from the PTSD Checklist-5 (PCL-5). Some studies have assessed affective dysregulation and disturbances in relationships using alternative measurement tools such as the Difficulties in Emotional Regulation Scale and the Inventory of Interpersonal Problems.

Six studies included a STAIR or a modified STAIR arm, three had a PE arm, and two had an imagery rescription arm. Other interventions included EMDR, TF-CBT, emotion focused therapy, acceptance and commitment therapy (ACT), body-oriented group therapy, stabilization therapy, mindfulness-based therapy, problem management, cognitive narrative therapy, and trauma-focused group music and imagery therapy. Four studies directly compared the efficacy of phase-based and non-phase-based therapies. Regarding treatment modality, two studies involved internet-based therapy and four were group-based.

Of the 16 included studies, 12 contributed to the intervention versus waitlist/TAU subgroup analyses and three contributed to head-to-head comparisons. One study employed a three-arm design, including two active control arms and one waitlist arm, and was therefore included in both analyses. For the subgroup analyses, we categorized six arms as phase-based, whereas we classified eight as non-phase-based. Three arms were single-phase, and the remaining arms were multi-phase. Nine arms included trauma exposure, while five were non-exposure-based.

3.2. Quality assessment

In the supplemental material, Figure S1 shows the results of the risk of bias. We rated more than half of the studies as low risk in the randomization process domain (56.3%). Most were rated as low risk in the missing outcomes domain (87.5%) and in the measurement of the outcomes domain (87.5%). In the deviations from the intended interventions domain, approximately 30% of the studies were rated as high risk, mainly because strict blinding was not possible for psychological treatments. In the selection of the reported results domain, 50% of the studies had some concerns and 12.5% had a high risk, primarily owing to a lack of registration.

3.3. Meta-analytic outcomes

3.3.1. Waitlist and TAU comparison

Table 3 displays the effect sizes, heterogeneity, and subgroup analyses for the waitlist/TAU comparison. Table 4 presents the effect sizes and heterogeneity of the head-to-head comparison. Figures S2 and S3 show the corresponding forest plots (see the supplementary material).

Table 3.

Subgroup analysis (Waitlist and TAU comparison).

Table 4.

Effect Sizes (Head-to-head Comparison).

Outcome	Total k	Hedge’s g (95% CI)	Heterogeneity
			Q, p-value	${\mathit{I}}^2$(%)
PTSD (CAPS)	4	0.04 (−0.23, 0.31)	4.62, .20	30.89
PTSD (CAPS + PCL)	4	0.03 (−0.22, 0.28)	4.21, .24	22.03
DSO (AD)	3	−0.15 (−0.44, 0.15)	3.05, .22	28.17
DSO (DR)	4	0.06 (−0.17, 0.29)	3.66, .30	9.87
DSO (NSC)	2	−0.02 (−0.30, 0.26)	0.09, .77	0.00
PTSD + DSO (AD)	3	−0.10 (−0.37, 0.17)	3.31, .21	32.36
PTSD + DSO (DR)	4	0.05 (−0.20, 0.30)	5.12, .16	38.97
PTSD + DSO (NSC)	2	0.00 (−0.21, 0.22)	0.68, .41	0.00
PTSD + DSO (AD + DR)	3	−0.07 (−0.36, 0.22)	3.72, .16	45.50
PTSD + DSO (DR + NSC)	2	0.04 (−0.19, 0.27)	0.86. .35	0.00
PTSD + DSO (AD + NSC)	2	0.00 (−0.23, 0.23)	0.53, .47	0.00
PTSD + DSO (AD + DR + NSC)	2	0.02 (−0.20, 0.25)	0.79, .37	0.00

AD: Affect Dysregulation; CAPS: Clinician-Administered PTSD Scale; DSO: Disturbances in Self-Organization; DR: Disturbances in Relationships; NSC: Negative Self-Concept; PCL: PTSD Checklist; PTSD: Post-Traumatic Stress Disorder.

3.3.1.1. PTSD

We excluded two studies due to an outlier effect size and a lack of homogeneity in the baseline assessment. Overall, the pooled effect on PTSD symptoms was moderate-to-large $(k=9,g=-0.58;95\text{\%}\mathrm{CI}=-0.82\mathrm{to}-0.35)$, with low heterogeneity (I² = 23.55%).

• Phase-based versus non-phase-based interventions: Phase-based interventions $(k=2,g=-0.94;$ $95\text{\%}\mathrm{CI}=-1.58\mathrm{to}-0.31)$ had larger effects than non-phase-based interventions $(k=7,g=-0.50;$ $95\text{\%}\mathrm{CI}=-0.73\mathrm{to}-0.27)$, but between-group heterogeneity was not significant ($Q=1.68,p=.20$).

• Multi-phase versus single-phase interventions: Multi-phase interventions $(k=6,g=-0.70;$ $95\text{\%}\mathrm{CI}=-0.93\mathrm{to}-0.48)$ had significantly larger effect sizes than single- phase interventions $(k=3,g=-0.15;95\text{\%}\mathrm{CI}=-0.62\mathrm{to}0.33;$ $Q=4.28,p=.04).$

• Exposure versus non-exposure interventions: Exposure $(k=5,g=-0.61;95\text{\%}\mathrm{CI}=-1.01$ to −0.22) and non-exposure $(k=4,g=-0.54;$ $95\text{\%}\mathrm{CI}=-0.87\mathrm{to}-0.20)$ interventions both revealed moderate-to-large effects, with no significant differences observed ($Q=0.09,p=.77$).

Overall, only the multi- vs. single-phase comparison reached significance in favour of multi-phase interventions, although non-phase-based (moderate effect size) and non-exposure (moderate-to-large effect size) interventions also had meaningful effects on PTSD symptoms.

3.3.1.2. CPTSD

We excluded one study due to a lack of homogeneity in the baseline assessment. In total, the studies had a moderate-to-large effect on CPTSD symptoms $(k=7,g=-0.60;95\text{\%}\mathrm{CI}=-0.83$ to −0.36). Heterogeneity across studies was low (I² = 17.25%).

• Phase-based versus non-phase-based interventions: Phase-based interventions $(k=3,g=-0.80;$ $95\text{\%}\mathrm{CI}=-1.87\mathrm{to}0.27)$ and non-phase-based interventions $(k=4,g=-0.67;95\text{\%}\mathrm{CI}=-0.93$ $\mathrm{to}-0.41)$ did not differ significantly ($Q=0.05,$ $p=.82$).

• Multi-phase versus single-phase interventions: All studies involved multi-phase interventions.

• Exposure versus non-exposure interventions: Exposure $(k=4,g=-0.65;95\text{\%}\mathrm{CI}=-1.13$ $\mathrm{to}-0.16)$, and non-exposure $(k=3,g=-0.62;$ $95\text{\%}\mathrm{CI}=-0.94\mathrm{to}-0.29)$ interventions also showed no significant difference ($Q=0.01,$ $p=.93$).

In summary, no subgroup differences emerged, and both non-phase-based and non-exposure interventions demonstrated moderate-to-large effects on CPTSD symptoms.

3.3.1.3. Disturbances in self-organization (DSO)

Two studies were excluded because of an outlier and baseline imbalance. In total, the studies had a moderate-to-large effect on DSO symptoms $(k=4,g=-0.62;95\text{\%}\mathrm{CI}=-0.93\mathrm{to}-0.31)$ with low heterogeneity (I² = 28.86%).

• Phase-based versus non-phase-based interventions: Only one study was phase-based.

• Multi-phase versus single-phase interventions: All studies were multi-phase.

• Exposure versus non-exposure interventions: Exposure $(k=2,g=-0.82;95\text{\%}\mathrm{CI}=-1.30$ $\mathrm{to}-0.33)$ and non-exposure $(k=2,g=-0.43;95\text{\%}\mathrm{CI}=-0.79$ $\mathrm{to}-0.07)$ interventions did not differ significantly ($Q=1.58,p=.21$). Non-exposure interventions also had small-to-moderate effects on DSO symptoms.

3.3.1.4. Disturbances in self-organization (DSO) – affect dysregulation (AD)

All studies had a large effect on AD symptoms $(k=7,g=-0.89;95\text{\%}$ $\mathrm{CI}=-1.47\mathrm{to}-0.31)$. Heterogeneity across the studies was considerable (I² = 78.47%).

• Phase-based versus non–phase-based interventions: Phase-based interventions $(k=3,g=$ $-1.61;95\text{\%}\mathrm{CI}=-2.21\mathrm{to}-1.01)$ had significantly larger effect sizes compared to non-phase-based interventions ($k=4,g=-0.43;95\text{\%}\mathrm{CI}=-0.94\mathrm{to}-0.08;$ $Q=8.54,p<.01$).

• Multi-phase versus single-phase interventions: Multi-phase interventions $(k=5,g=-1.19;95\text{\%}$ $\mathrm{CI}=-1.83\mathrm{to}-0.55)$ had significantly larger effect sizes than single-phase interventions ($k=2,$ $g=-0.10;95\text{\%}\mathrm{CI}=-0.65\mathrm{to}0.45$; $Q=6.48,p=.01$).

• Exposure versus non-exposure interventions: Exposure interventions $(k=5,g=-1.22;$ $95\text{\%}\mathrm{CI}=-1.78\mathrm{to}-0.66)$ also displayed significantly larger effect sizes than non-exposure interventions ($k=2,g=-0.17;$ $95\text{\%}\mathrm{CI}=-0.57\mathrm{to}0.23;$ $Q=8.99,p<.01$).

In summary, phase-based, multi-phase, and exposure interventions had significantly higher effect sizes on AD than non-phase-based, single-phase, and non-exposure interventions. However, non-phase-based interventions also had a small-to-moderate effect on AD symptoms.

3.3.1.5. Disturbances in self-organization (DSO) – disturbances in relationships (DR)

The pooled effect on DR symptoms was moderate-to-large $(k=6,g=-0.72;95\text{\%}\mathrm{CI}=-0.97\mathrm{to}-0.47)$, with low heterogeneity (I² = 0.00%).

• Phase-based versus non-phase-based interventions: Phase-based interventions $(k=2,g=-1.18;$ $95\text{\%}\mathrm{CI}=-1.87\mathrm{to}-0.48)$ and non-phase-based interventions $(k=4,g=-0.60;95\text{\%}\mathrm{CI}=-0.89$ $\mathrm{to}-0.31)$ did not differ significantly ($Q=2.28,$ $p=.13$).

• Multi-phase versus single-phase interventions: Multi-phase $(k=4,g=-0.81;95\text{\%}\mathrm{CI}=-1.10$ $\mathrm{to}-0.53)$ and single-phase $(k=2,g=-0.35;$ $95\text{\%}\mathrm{CI}=-0.90\mathrm{to}0.20)$ interventions also showed no difference ($Q=2.14,p=.14$).

• Exposure versus non-exposure interventions: Exposure $(k=4,g=-0.86;95\text{\%}\mathrm{CI}=-1.18$ $\mathrm{to}-0.54)$ and non-exposure $(k=2,g=-0.49;$ $95\text{\%}\mathrm{CI}=-0.90\mathrm{to}-0.09)$ interventions likewise exhibited no significant difference ($Q=1.96,$ $p=.16$).

Overall, none of the subgroup analyses revealed significant differences between the subgroups. Both non-phase- and non-exposure interventions produced moderate-to-large and small-to-moderate effects on DR symptoms, respectively.

3.3.1.6. Disturbances in self-organization (DSO) – negative self-concept (NSC)

All studies had a moderate-to-large effect on NSC symptoms $(k=2,g=-0.69;95\text{\%}\mathrm{CI}=-1.03\mathrm{to}-0.36)$. Heterogeneity across studies was low (I² = 0.00%). Because there were only two studies, a subgroup analysis was not feasible.

3.3.2. Head-to-head comparison

In studies directly comparing phase- and non-phase-based interventions, none of the outcome measures demonstrated a significant advantage with either approach.

3.4. Reporting bias

Visual inspection of the funnel plot for PTSD outcomes suggested that one study (Karatzias et al., 2018) was an outlier. The removal of this study resulted in a symmetrical funnel plot (Figure S4). This finding was consistent with Egger’s test $(k=10,t=0.26,p=.80)$, and no studies needed to be imputed using the trim-and fill method (Table S1).

4. Discussion

It is widely accepted that individuals with CPTSD require differentiated treatment in which exposure to traumatic memories is delivered gradually and within a phase-based framework. However, this consensus has not yet been fully supported by empirical evidence. To address this gap, we conducted a systematic review and meta-analysis of RCTs on psychological interventions for patients with CPTSD. We examined whether phase-based, multi-phase, or exposure-based interventions yielded greater benefits for CPTSD symptoms than their counterparts. Through a systematic search, we identified 16 studies involving waitlist or TAU conditions, and 4 studies directly comparing phase-based or non-phased interventions. Effect size estimations indicated that most comparisons between phase- and non-phased-based interventions, multi-phase and single-phase interventions, and exposure and non-exposure interventions failed to yield significant differences in either the waitlist and TAU comparisons or head-to-head comparisons. Exceptions emerged for multi-phase interventions, which tended to have significantly larger effect sizes on PTSD as well as DSO-affect dysregulation. Similarly, phase-based or exposure interventions showed larger effect sizes on DSO–affect dysregulation than non-phase-based and non-exposure-based interventions. Nevertheless, non-phase-based interventions also had meaningful effects on PTSD (small-to-moderate), CPTSD (moderate-to-large), and DSO–disturbances in the relationships (moderate-to-large). Additionally, non-exposure interventions had beneficial effects on PTSD (moderate-to-large), CPTSD (moderate-to-large), DSO symptoms (small-to-moderate), and DSO–disturbances in relationships (small-to-moderate).

The subgroup analyses revealed no significant differences in the relationships between phase- and non-phase-based interventions for PTSD, CPTSD, and DSO–disturbances in relationships. Furthermore, head-to-head analysis yielded no significant superiority of phase-based interventions over non-phase-based interventions. One possible explanation is that the limited number of studies may have reduced statistical significance, thereby limiting our ability to detect differences. A more plausible explanation, however, is that non-phase-based interventions are likely to have effects comparable to those of phase-based interventions in reducing the psychological distress of CPTSD. Four studies (Oprel et al., 2021; Raabe et al., 2022; Sele et al., 2023; van Vliet et al., 2018) directly compared phase-based interventions (i.e. STAIR + exposure) with non-phase interventions (i.e. exposure only); none of them reported significant differences between the two interventions. Furthermore, a recent meta-analysis on the effectiveness of psychological interventions in a population with probable CPTSD (Hu et al., 2025) suggests that unimodal approaches (such as CPT, EMDR, and mindfulness) also have moderate-to-large effects on PTSD symptoms. Although further replications with larger samples is warranted, these findings contribute to the discourse by implying that effective treatment of complex PTSD symptoms can be achieved without necessarily prioritizing the stabilization phase before other components.

Subgroup analyses demonstrated no significant differences in the relationships between exposure and non-exposure interventions for PTSD, CPTSD, DSO symptoms, and DSO-disturbances in relationships. This may indicate that non-exposure interventions have effects comparable to those of exposure-based interventions in reducing CPTSD symptoms. These findings are in line with those of a prior review (Rubenstein et al., 2024), that found that non-trauma-focused and non-exposure interventions (e.g. present-centered therapy, interpersonal psychotherapy, non-trauma-focused CBT, ACT) are as effective as – and non-inferior to – exposure-based interventions. Moreover, one meta-analysis (McLean et al., 2022) showed that exposure therapy had a smaller effect than a non-trauma-focused comparator, and a negligible effect compared to other trauma-focused treatments. This indicates that effective treatment of PTSD can be achieved through multiple therapeutic pathways. Future research should therefore move beyond the question of whether exposure is necessary and instead focus on treatment diversity and patient-centered tailoring.

For PTSD outcomes, the subgroup analyses demonstrated that multi-modular approaches showed statistically larger effect sizes than single-phase interventions. This suggests that multi-modular approaches may be better options for alleviating PTSD symptoms. Consistent with this, Karatzias et al. (2019) asserted that flexible multi-modular approaches are likely to be more effective than unimodular ones in treating CPTSD, given their distinct symptom clusters. This view is further supported by Bohus et al. (2020), who studied women with childhood abuse-associated PTSD. This study found that a flexible modular approach (dialectical behaviour therapy–PTSD [DBT–PTSD]) had low early drop-out rates and symptomatic remission in PTSD compared to CPT, which focuses on cognitive distortions related to trauma. These results emphasize the potential value of interventions that integrate diverse phases (including stabilization, exposure, and reintegration) to alleviate PTSD symptoms.

Moreover, in DSO–affect dysregulation, multi-phase, phase-based, and exposure interventions demonstrated significantly larger effect sizes than their counterparts. In other words, as far as affect dysregulation is concerned, a multi-modular approach that sequences stabilization before exposure might be more effective at reducing the symptoms of DSO–affect regulation than others. This finding could be attributed to the distinctive features and change mechanisms of affect dysregulation compared with the other DSO symptom clusters.

First, affect dysregulation appears to function as a predictor of vulnerability and maintaining factor in CPTSD, making it a critical early target for treatment. Villalta et al. (2020) reported that baseline emotion dysregulation – but not negative self-concept or interpersonal problems – significantly predicted both functional impairment and re-traumatization four to five months after a sexual assault. Previous studies have consistently identified emotion dysregulation as a core vulnerability and maintenance factor of PTSD (Conti et al., 2023), a mediator linking childhood maltreatment to comorbidities such as substance use and eating disorders (Messman-Moore & Bhuptani, 2017), and a common underlying feature of BPD.

Second, several studies have supported the idea that affect dysregulation must be addressed through a gradual, multi-modular approach. Cloitre et al. (2010) showed that interventions beginning with STAIR before exposure (STAIR + exposure) led to greater improvements in emotion regulation at post- and follow-up assessments than interventions that began with supportive counseling before exposure. Likewise, Jaycox et al. (1998) found that individuals exhibiting emotional disengagement (e.g. avoidance, dissociation, numbing) benefited less from exposure therapy than those who were able to engage emotionally. Together, these findings highlight the need to conceptualize the treatment of affect dysregulation as a gradual and multi-modular process in which stabilization is followed by exposure.

4.1. Limitations

This study has some limitations. First, the lack of a sufficient number of treatment studies on CTPSD may have reduced the statistical power to detect subgroup differences. Second, for practical reasons, we required the presence of only two DSO symptom clusters (i.e. emotional dysregulation and disturbances in relationships). Thus, populations that meet the full diagnostic criteria for CPTSD may have yielded different results. Third, individual studies employed diverse measures and, in some cases, reported only total scores rather than the subfactors of PTSD or DSO, making it difficult to pool effect size estimates. Finally, follow-up analyses were not feasible because only a small number of studies reported data at heterogeneous time points, which prevented us from examining the duration of the intervention effects. Future research would thus benefit from a larger number of trials, assessing all three DSO symptom clusters, and using standardized, multi-dimensional measures to provide more robust evidence.

4.2. Conclusion and implications

This study provides a meta-analytic synthesis to inform the ongoing discourse regarding the comparative effectiveness of phase-based interventions for CPTSD. Across most outcome variables, no significant differences were observed between the phase- and non-phase-based interventions, between multi-phase and single-phase interventions, or between exposure and non-exposure interventions. However, we found notable exceptions: Multi-phase interventions demonstrated superior effects on PTSD, and phase-based, multi-phase, and exposure interventions showed greater improvements in DSO–affect regulation compared to their counterparts.

These findings contribute empirical evidence to the ongoing discourse, highlighting that non-phase-based and non-exposure interventions may be comparable to more structured approaches in many contexts, while certain outcomes may benefit from multi-modular or exposure-based designs. These results have several practical implications for clinicians working with patients who meet the diagnostic criteria for CPTSD. (1) For PTSD symptoms: a strict sequencing of stabilization before exposure may not be essential to successfully alleviate symptoms. Employing a multi-modular approach could effectively reduce PTSD symptoms. (2) For affect dysregulation (DSO–AD): multi-phase interventions that begin with stabilization and are followed by gradual exposure may be particularly effective. (3) Flexible tailoring: For other symptom domains, clinicians are encouraged to flexibly tailor their treatment strategies to the individual patient’s primary difficulties and readiness for change, rather than focusing exclusively on whether stabilization should precede exposure.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Data availability statement

The data employed in this study are available in the Figshare repository at .

Supplemental Material

Supplemental data for this article can be accessed online at https://doi.org/10.1080/20008066.2026.2644112.