New findings questioning the construct validity of complex posttraumatic stress disorder (cPTSD): let’s take a closer look

ABSTRACT

This commentary provides a broader context for interpreting evidence from Latent Class and Latent Profile analyses on complex posttraumatic stress disorder (CPTSD) that was provided in a recent contribution to the European Journal of Psychotraumatology. These data analytic strategies are not alone sufficient to test the construct validity of CPTSD. They base their conclusion on the empirical finding of substantial variation in latent models obtained with different analytic procedures and interpretations of the fit of different latent models, as well as interesting additional evidence of dispersion when individual patients’ symptom counts and symptom severity scores on PTSD and CPTSD are examined. However, the results of their analyses actually do provide support for one feature of construct validity, demonstrating discriminant validity by showing a consistent differentiation between PTSD and CPTSD (with expectable variation in both PTSD and CPTSD severity level by persons). Even in a sample of patients diagnosed with PTSD, there may be a Disorders of Self Organization (DSO) sub-group with low PTSD symptom severity. More detailed examination of which DSO symptoms and sub-domains characterize the DSO sub-group and the CPTSD sub-group is needed in order to clarify the nature of the DSO/CPTSD construct. Other analyses needed to fully test construct validity also are described.

Complex posttraumatic stress disorder (cPTSD) has been the subject of controversy since it was first proposed almost three decades ago (Herman, 1992). In this commentary, I will concur with the conclusion of a recent study by Achterhof et al. (2019) that demonstrating construct validity for CPTSD requires more evidence than can be provided by latent class (LCA) and latent profile (LPA) analyses alone. However, I will challenge their assertion that their LCA and LPA results cast doubt on the construct validity of CPTSD, and suggest that instead their findings actually provide consistent evidence of distinct PTSD, CPTSD, and also Disorders of Self Organization (DSO) sub-groups even in a sample of patients diagnosed with PTSD. I also will suggest that variability in individual symptom levels for PTSD and CPTSD is expectable and does not constitute evidence of a lack of discriminant validity for either PTSD or CPTSD, and that more detailed examinations of the specific symptoms characterizing putative PTSD or CPTSD/DSO sub-groups and syndromes is necessary in order to accurately characterize and test the validity of PTSD and CPTSD.

Although it has been argued that cPTSD is unnecessary and better conceptualized as PTSD plus comorbidities or a variant of borderline personality disorder (Resick et al., 2012), a growing body of research (Ford, 2015; Olff et al., 2019) indicates that cPTSD ‘is both conceptually and clinically useful [as a] distinct entity [that] is highly prevalent, across different cultures, in survivors of prolonged, repeated trauma’ (Herman, 2012, p. 256). Indeed, a version of cPTSD that combines PTSD with Disorders of Self Organization (i.e. affect dysregulation, interpersonal detachment, negative self-concept) has been accepted in the 11th revision of the International Classification of Diseases (ICD-11) based on the growing empirical evidence base (Brewin et al., 2017).

A critical foundational question that extends beyond the decision whether to codify cPTSD as a psychiatric diagnosis is whether cPTSD has validity as a conceptual construct. Evidence of structural distinctiveness of cPTSD and PTSD in confirmatory factor analysis studies (Gilbar, Hyland, Cloitre, & Dekel, 2018; Karatzias et al., 2017; Shevlin et al., 2017), and of distinct sub-groups of trauma-exposed persons with specific symptom profiles consistent with cPTSD versus PTSD (Ben-Ezra et al., 2018; Karatzias et al., 2017, 2017) support the validity of the cPTSD construct.

However, other research has been reported that cogently ‘raise[s] concerns about the distinctions between CPTSD and PTSD proposed for ICD-11’ (Wolf et al., 2015, p. 215). Using a factor mixture model with two latent dimensional variables representing distinct structures of interrelationships among symptoms, and four latent classes representing sub-groups with distinct symptom profiles, with civilian and military samples, Wolf et al. (2015) found evidence that levels of symptom severity but not a differentiation of PTSD and cPTSD symptoms best accounted for derived sub-sets of both symptoms and individuals. More recently, with a sample of resettled adult refugees, subgroups were found with symptom profiles consistent with PTSD and with CPTSD, but also a subgroup with prominent affect dysregulation symptoms, as well as evidence of cumulative trauma exposure only for the PTSD and cPTSD classes (Liddell et al., 2019). In a United States adult community sample, cumulative trauma exposure in adulthood and specific adverse childhood events were associated with both PTSD and cPTSD, but cumulative childhood trauma exposure had a clear dose–response relationship with CPTSD but was associated with PTSD only at the highest cumulative exposure level – and to a much lesser degree than for CPTSD (i.e. with odds ratios of 2.6 vs. 21.9) (Cloitre et al., 2019). Further, persons who met criteria for cPTSD had more severe and extensive psychiatric symptoms and poorer well-being than those who met criteria only for PTSD or who met neither diagnostic criteria. Thus, whether cPTSD is the best construct to represent post-traumatic affect dysregulation, as well as whether cPTSD represents a sub-type of PTSD associated with high levels of cumulative trauma exposure rather than a distinct construct, remain in question.

In the midst of this rapidly evolving controversy, Achterhof et al. (2019) have reported a timely empirical examination of the construct validity of cPTSD in a sample of adult patients diagnosed with PTSD. Using both latent class (LCA) and latent profile (LPA) analytic procedures, sub-groups were identified with symptom profiles consistent with PTSD and cPTSD. However, the two syndromes did not have consistent symptom profiles due to variation in how individuals were classified and how the LCA and LPA model fit statistics were interpreted. In a two-class LCA, only symptom frequency across both PTSD and cPTSD symptoms distinguished the classes, consistent with Wolf et al. (2015). When a three-class LCA solution was accepted, a distinct PTSD class emerged, along with two other classes that differed primarily based on high versus moderate frequency of symptom endorsement across the full range of PTSD and cPTSD symptoms. In a four-class LPA model, two classes represented, respectively, high and low symptom severity, again consistent with Wolf et al. (2015) but, as in the three-class LCA, a distinct PTSD class again emerged. A fourth class reflected high severity with the exception of low levels of PTSD intrusion symptoms. Overall, the findings are consistent with the ICD-11 formulation, with one class characterized by high levels or severity of PTSD symptoms and another class characterized by high levels or severity of PTSD and cPTSD symptoms (Brewin et al., 2017).

Achterhof et al. (2019) also provided an informative additional graphic analysis on a person-specific basis, via scatterplots of counts (from the LCA) and cumulative severity (from the LPA) of PTSD and cPTSD symptoms. The LCA scatterplot shows a relatively good separation of the PTSD, cPTSD, and moderate symptom classes, but what is striking is that the moderate symptom class has generally high levels of cPTSD symptoms (i.e. 3 or more of a possible 5) and low levels of (i.e. 1–3) PTSD symptoms. On the LPA scatterplot, similarly, PTSD and cPTSD classes were relatively separate, but a moderate severity symptom class had relatively low/moderate cPTSD as well as PTSD symptoms, and the unexpected fourth class overlaps almost entirely with the cPTSD class. Achterhof et al. (2019) conclude that these results are ‘inconsistent with the theory of separate PTSD vs. CPTSD cluster [and] clearly demonstrate that there are no natural well-separated clusters … at most show[ing] a gradual difference between the classes defined by the LPA.’

An alternative interpretation might be that the graphs depict a PTSD class and a cPTSD class as proposed, although each has a range of symptoms/severity from moderate to high and therefore it cannot be assumed that either PTSD or cPTSD involves uniformly high numbers or severity of class-specific symptoms. The fact that some members of those two classes were positioned close to the border of the other class does not call into question the validity of cPTSD as a construct any more than it calls into question the validity of PTSD. Both constructs are presumed to have variability on the level of individuals – not so much as the 636,120 variations that can occur in DSM-5 PTSD (Galatzer-Levy & Bryant, 2013), but certainly not so much uniformity that persons could not be classified as meeting the criteria for the construct/diagnosis while having absolute symptom levels at the low or high end. The dispersion in symptom levels thus does not invalidate cPTSD (or PTSD) as a construct.

On the other hand, the additional classes in the three-class LCA and four-class LPA raise interesting questions, particularly when the scatterplot findings are considered. Overall, most PTSD scores were in the moderate to high range (as expected in a sample comprised of patients diagnosed with PTSD), but the third LCA class and fourth LPA class had numerous members with relatively low PTSD symptom count/severity (Cloitre et al. 2013). Members of those classes had generally moderate to high cPTSD symptom counts (i.e. 3–5 of 5 possible) and moderate cPTSD symptom severity (i.e. approximately 15 on a 5–25 scale). Could this class represent a Disorders of Self-Organization (DSO) sub-group with low PTSD symptoms and cPTSD/DSO symptoms that are clinically significant but generally not as extreme as those attributed to a disorder of particularly severe self/relational dysregulation such as borderline personality disorder (BPD; see below)? Or could this class represent a sub-group with prominent affect dysregulation but no major problems in relationships of sense of self, similar to a class found in a study with re-settled adult refugees that was characterized by severe affect dysregulation and current life stressors (i.e. residential insecurity) but no other cPTSD or PTSD symptoms (Liddell et al., 2019)? Alternately, these classes could represent patients who are better characterized as experiencing impairing depression or anxiety, rather than DSO or cPTSD-related affect dysregulation (Gilbar, 2020).

A closer look at which DSO symptoms were present/severe for the members of those classes, as well as assessment of depression and anxiety symptoms that are not specifically trauma-related, would be necessary in order to determine how best to characterize them and their disorder(s). Indeed, they might best be characterized by a single DSO construct (with variability in symptom presence/severity), but alternatively, they might comprise several rather than only a single clinical syndrome(s) that could be trauma-related or -unrelated. This finer-grained examination is important in order to test the validity of the DSO component of cPTSD when it occurs in the absence of PTSD (Ford, 1999), as well as its potential variations and boundaries and discriminability from other psychiatric and behavioural disorders.

Then, there is the anomalous fourth LPA class, in which low levels of intrusive re-experiencing symptoms occurred in conjunction with high levels of all other PTSD and cPTSD symptoms. Does this class represent the DSM-5 dissociative sub-type of PTSD (Lanius, Brand, Vermetten, Frewen, & Spiegel, 2012), with low levels of intrusive re-experiencing due to dissociative hypoaraousal and attenuation of reactivity to trauma cues and conscious awareness of trauma memories? Or a hyperarousal syndrome consistent with agitated depression or extreme anxiety (e.g. panic) (Gilbar, 2020)? Or a severe variant of borderline personality disorder with multiple forms of extreme biopsychosocial dysregulation that are associated with childhood-onset cumulative adversity (Ford & Courtois, 2014)? Here too, examination of person-specific symptom profiles would be needed in order to determine whether this class represents one or several clinical constructs that may differ from or serve as variants of cPTSD – and thus to better define and determine the validity of cPTSD.

Fundamentally, construct validity requires evidence that cPTSD: (1) is associated with measures of integrally related constructs (e.g. affect dysregulation, relational problems, trauma-impacted self-concept (i.e. convergent validity); (2) is empirically distinguishable from not only PTSD but also other psychiatric syndromes (i.e. discriminant validity); and (3) has a replicable structure consistent with the proposed conceptual domains (i.e. internal consistency and structural validity). For example, a study replicating findings across the populations of adults in substance abuse treatment and incarcerated adults tested a progenitor of cPTSD, disorders of extreme stress not otherwise specified (DESNOS), and found replicated evidence of symptom factors that were internally consistent and showed evidence of convergent and discriminant validity (Scoboria, Ford, Lin, & Frisman, 2008). In that study, notably, some symptoms that had been proposed to belong in cPTSD were not included in the empirically derived factors (e.g. dissociation), and this began a winnowing and streamlining of cPTSD that has culminated in the ICD-11 version. More recently, in a study with children in paediatric and mental health care in the United States, a variant of cPTSD (‘developmental trauma disorder’) has shown evidence of convergent validity and discriminant validity in relation to PTSD and both internalizing and externalizing disorders, as well as based on a pattern of traumatic antecedents specifically involving victimization and relational and/or attachment adversity (Ford, Spinazzola, van der Kolk, & Grasso, 2018; Spinazzola, van der Kolk, & Ford, 2018; van der Kolk, Ford, & Spinazzola, 2019).

A theory-informed and evidence-based approach to determining the nature and validity of all of the posttraumatic disorders could begin with the widely held and well supported view that PTSD is an adaptation to existential threat in which a preoccupation with threat-related cues (in the form of intrusive memories, hypervigilance) and corresponding attempts at harm avoidance (Colic et al., 2018) that is most likely to occur when traumatic experiences are recent and clearly recalled. By contrast, the additional cPTSD, symptoms that have been characterized as disorders of self organization (DSO) could be a posttraumatic adaptation to traumatic experiences involving interpersonal betrayal (Freyd, 1994) and the resultant affect dysregulation due to a sense of vulnerability in close relationships (Van Dijke, Hopman, & Ford, 2018) and an associated sense of shame (Lopez-Castro, Saraiya, Zumberg-Smith, & Dambreville, 2019) and damage to one’s sense of self. This conceptualization might co-locate PTSD with disorders involving fear (e.g. panic, phobias), anxiety (e.g. generalized or social anxiety disorder), while placing cPTSD closer to disorders of dysregulation of emotion, consciousness, identity or self-control (e.g. dissociation disorders, depression, borderline personality disorder, and childhood externalizing or dysregulation disorders).

With this theoretical foundation, research could systematically test construct validity by examining whether PTSD and cPTSD are discriminable from closely aligned comorbid disorders. In a clinical adult sample, cPTSD was distinguished from PTSD by higher levels of dissociative, depression, self-harm, suicidality, and borderline personality disorder (BPD) symptoms (Hyland, Shevlin, Fyvie, & Karatzias, 2018). BPD has been hypothesized to involve enmeshment in labile relationships based on a fear of abandonment or rejection, in contrast to viewing cPTSD as involving fear of closeness and associated relational numbing and detachment (Ford & Courtois, 2014). Consistent with this view, a study with a trauma-exposed United Kingdom population sample found distinct PTSD, cPTSD, and BPD latent variables that were characterized, respectively, by (1) threat-related intrusive memories, avoidance, and hyperarousal (PTSD); (2) emotional numbing, a sense of self as damaged, and relational detachment (cPTSD); and, (3) emotional reactivity and disinhibition, terror of abandonment, and absence of a clear or sustained sense of self (BPD) (Hyland, Karatzias, Shevlin, & Cloitre, 2019). However, a structural equation modelling study with a Dutch clinical sample found that fear of both abandonment and closeness, and affect dysregulation, mediated the relationship between childhood trauma and both cPTSD and BPD – but what distinguished the disorders was mediation by different forms of dissociation: dissociative fragmentation and somatoform dissociation were mediators for cPTSD, while dissociative intrusions and rumination were mediators for BPD (Van Dijke et al., 2018).

Thus, complex interconnections appear to be involved both within and between the constructs and syndromes/diagnoses associated with posttraumatic adaptations. Network analysis or symptomics may provide important additional clarification of the nature and validity of posttraumatic symptom/adaptation constructs (Armour, Fried, & Olff, 2017). The PTSD symptoms of emotional and physiological reactivity to trauma cues and the cPTSD symptoms of emotional dysregulation (e.g. numbing, angry rumination) are consistently identified as having high centrality (i.e. many strong interconnections) in the symptom networks of a variety of trauma-exposed adult populations (Fried et al., 2018; Greene, Gelkopf, Epskamp, & Fried, 2018; Moshier et al., 2018; Sullivan, Smith, Lewis, & Jones, 2018; von Stockert, Fried, Armour, & Pietrzak, 2018). Two network analysis studies have extended these findings by including symptoms of depression and anxiety disorders as well as PTSD and cPTSD. In a sample of men who have sex with men, PTSD and depression symptoms clustered relatively separately, but they also had overlapping symptoms (e.g. sleep and concentration problems, and only PTSD avoidance symptoms were associated with risky sexual behaviour (Choi, Batchelder, Ehlinger, Safren, & O’Cleirigh, 2017). In a sample of men who perpetrated domestic violence, Gilbar (2020) similarly found evidence of clustering of symptoms consistent with conceptual models of PTSD (except avoidance and hypervigilance were strongly interconnected and did not constitute separate sub-syndromes), depression, and anxiety – but the proposed cPTSD symptoms tended to be dispersed rather than representing a distinct syndrome or set of sub-syndromes. Relational problems and one of the emotion dysregulation symptoms (emotional numbing) clustered together and with the depression symptom of worthlessness, and the worthlessness symptom also was highly connected with cPTSD negative self-concept symptoms. The other cPTSD emotion dysregulation symptom (difficulty recovering from distress) clustered with both the anxiety and depression symptoms but was not connected to any other cPTSD (or PTSD) symptom.

Although these two network analyses studied specialized sub-populations of high-risk men, their findings are consistent with those of Achterhof et al. (2019) – suggesting that the symptoms currently being used to define cPTSD in the ICD-11 may either require modification in order to constitute a distinct, cohesive, and valid construct – or may represent complex extensions of other disorders including not only PTSD but also depression, anxiety, and dysregulation disorders. Achterhof et al. (2019) have contributed to this quest by raising the question of cPTSD’s construct validity, and by reminding us that clarifying methodological assumptions and examining individual variability in symptom presentation are crucial in tests of construct validity, when evaluating the meaning of empirically derived cross-individual classes, profiles, or symptom factor structures. Each new set of findings should always be looked at still more closely and deeply, so that we can gain incrementally greater clarity about our focal constructs rather than accepting any of them as fixed and final.

Of course, one can only find what one’s instruments are able to make visible, so it also is important to recognize that the specific expressions of PTSD and cPTSD symptoms that we use in construct validity, diagnostic, epidemiologic, and treatment outcome studies may not yet represent the fullest or most accurate depiction of the construct we seek to measure – the adaptations, experience, and recovery of trauma survivors. The map is not the territory, and the maps our assessments provide may be psychometrically sound – yet we still must be careful to never substitute them for the existential truth of each trauma survivor.

Funding Statement

This work was supported by the Substance Abuse and Mental Health Services Administration [5U79SM080013].

Disclosure statement

Julian Ford is co-owner of Advanced Trauma Solutions, Inc., the sole licenced distributor of the TARGET model copyrighted by the University of Connecticut.