Lifetime Trauma History and Cognitive Functioning in Major Depression and Their Role for Cognitive-Behavioral Therapy Outcome

Participants and Procedures

Recruitment was conducted within the outpatient unit of the Institute of Clinical Psychology and Psychotherapy of the Technische Universität Dresden, as well as via flyers and local advertisements. Individuals were included in the study if they were aged between 18 and 65 years, not pregnant (women), and did not report any severe physical diseases (e.g., cancer, encephalopathy) over the past five years. Further exclusion criteria concerned hair-related and endocrine factors due to biomarker analyses reported elsewhere (e.g., glucocorticoid medication; Steudte et al., 2013; Steudte-Schmiedgen et al., 2014). The presence of MD and any other DSM-IV (American Psychiatric Association, 2007) mental disorders was assessed using the standardized Munich Composite International Diagnostic Interview (DIA-X/M-CIDI; Wittchen & Pfister, 1997) conducted by therapists of the outpatient unit or trained research team members and confirmed by an experienced clinical psychologist. Twenty-eight participants from the MD groups showed psychiatric comorbidities within the last 12 months (one: n = 15, two: n = 8, three or more: n = 5). Those encompassed specific (n = 12) or social phobia (n = 13), somatoform disorders (n = 6), panic disorder with or without agoraphobia (n = 8), generalized anxiety (n = 3), obsessive-compulsive (n = 2), adjustment (n = 2), or eating disorders (n = 1).

An assignment to the MD groups was based on a current primary 12-month MD diagnosis and no 12-month diagnosis of substance abuse or dependence (except for nicotine) or any lifetime diagnoses of psychosis, severe depressive disorder with psychotic symptoms, or bipolar disorder. Notably, individuals meeting the lifetime diagnostic criteria for PTSD were also excluded from the study, in order to allow insights into the role of lifetime trauma exposure per se for cognitive functioning in MD. Participants were included in the control group if they did not report any lifetime mental disorders according to the DIA-X/M-CIDI stem questions and the Mini International Neuropsychiatric Interview (M.I.N.I.; Sheehan et al., 1998). Participants were further classified as exposed or non-exposed to lifetime trauma based on the Posttraumatic Stress Diagnostic Scale (PDS; Ehlers, Steil, Winter, & Foa, 1996). For an allocation to the T+ groups, both the “objective” A1 (“actual or threatened death or serious injury or a threat to the physical integrity of oneself or others”) and the “subjective” A2 criterion (“intense fear, helplessness or horror”) had to be met, following the DSM-IV requirements that qualify life events as traumatic (American Psychiatric Association, 2007). The control groups are the same as in the parallel study on patients with PTSD (Schindler et al., 2020). For further participant characteristics, see Table 1 and Supplementary Materials (type of lifetime trauma history).

CBT for MD groups was conducted within the outpatient unit based on established manuals (Hautzinger, 1998, 2008) and supervised by experienced therapists. After 25 sessions, MD+T+ and MD+T- patients were re-invited for clinical and cognitive testing, with MD-T- participants being contacted in a parallel fashion (no difference regarding months between assessments: M = 13.5, SD = 3.86; M = 11.56, SD = 4.03; and M = 14.76, SD = 6.97, respectively; F(2, 68) = 1.78, p = .177, ${\mathrm{\eta }}_{\mathrm{p}}^2$ = .05). Among the 63 patients with MD examined at baseline, 6 (9.5%) were only interested in the cross-sectional study, 16 (25.4%) dropped out of CBT, and 41 (65.1%) completed therapy. Between those who dropped out of CBT and those who did not, no differences emerged regarding pre-treatment clinical variables (all ps ≥ .219). All participants had provided written informed consent before study inclusion. The study protocol was approved by the ethics committee of the Technische Universität Dresden (EK 65022010) and conducted in accordance with the Declaration of Helsinki.

Clinical and Psychological Measures

Self-developed questionnaires were applied for socio-demographic (age, sex, education status) and health-related variables (smoking, chronic physical diseases, regular medication intake). Depressive symptoms over the previous two weeks were assessed via the Beck Depression Inventory-II (BDI-II, Hautzinger et al., 2006). The PDS (Ehlers et al., 1996) provided insights into the presence or absence of lifetime trauma history and the severity of symptoms associated with posttraumatic stress according to DSM-IV criteria. The Trauma History Questionnaire (THQ, Maercker, 2002) provided an overview over number and frequency of potentially traumatic events fulfilling the DSM-IV A1, but not A2 criterion (Hooper et al., 2011). Furthermore, to obtain information on the severity of childhood maltreatment (irrespective of fulfilling DSM-IV A criteria), the Childhood Trauma Questionnaire (CTQ, Gast et al., 2001) was used. At follow-up, patients with MD additionally received the revised version of the Questionnaire of Changes in Experience and Behavior (Veränderungsfragebogen des Erlebens und Verhaltens VEV-R; Zielke & Kopf-Mehnert, 2001). This allowed a classification of patient-evaluated therapy effects via 42 items of opposite polarity (e.g., “Compared with the time prior to initiation of therapy, I feel more relaxed/no change/more tense.”) into three categories (i.e., symptom improvement, no change, and worsening).

Cognitive Tasks

WM was examined using the Wechsler Memory Scale digit span task (Wechsler, 1997). Participants repeated a series of numbers read out loud by the experimenter in a forward (information maintenance) or backward fashion (information manipulation). Interference susceptibility and conflict adaptation were assessed by a number version of the Simon task (Fischer et al., 2008). In brief, participants categorized the numbers 1 to 9, except 5, as smaller or larger than five by pressing a left (Alt) or right (Alt Gr) key on a QWERTZ keyboard with their left or right index finger, respectively. Although task-irrelevant, stimulus location automatically facilitates the pressing of the corresponding response button, either in accordance, or in conflict with the required action, resulting in compatible and incompatible trials, respectively. The resulting difference in RTs and PEs comprises the Simon, and the typical reduction of interference susceptibility after conflict trials the conflict adaptation effect (Botvinick et al., 2004; Simon, 1990). Participants completed a 16-trial practice, followed by three 64-trial test blocks, resulting in 192 test trials (for further details, see Schindler et al., 2020; Steudte-Schmiedgen et al., 2014). Indices for interference susceptibility (I – C) and conflict adaptation [(cI – cC) – (iI – iC)] (lowercase letters: compatibility of the previous, uppercase letters: compatibility of the current trial, larger values indicating more pronounced effects) were calculated (van Steenbergen et al., 2010).

Autobiographical memory specificity was assessed via the standardized Autobiographical Memory Test (Williams & Broadbent, 1986). Participants were instructed to read words out loud (practice phase: three neutral words, testing phase: five positive and five negative words in a pseudo-randomized order, starting with a positive word and alternating valence) and briefly describe a related specific autobiographical memory. The words were randomly chosen from a word pool from a previous study (Schönfeld & Ehlers, 2006) matched for word frequency, emotionality, imagery, and pleasantness (apart from positive words rated as more pleasant than negative ones; Hager & Hasselhorn, 1994), with different sets used at baseline and follow-up. Answers were tape-recorded, transcribed and coded by trained research assistants. As an outcome variable, the number of specific memories was used, defined as having happened at a particular place and time more than one week ago and having lasted for one day or less. If no answer was provided within 30 seconds, the trial was considered an omission. For assessing inter-rater-reliability, a second, independent rater re-assessed a random sample (10%) of the tape-recorded sequences, resulting in κ = .76 for the baseline and κ = .82 for the follow-up assessment.

Statistical Analyses

Analyses were conducted via SPSS for Windows, version 25 (IBM, Armonk, NY), R (R Core Team, 2017), and STATA 15.1 (StataCorp LLC, 2017). Cross-sectional group comparisons were carried out via univariate analyses of variance (ANOVAs; continuous variables) and Χ² contingency tables (dichotomous variables). For the Simon task, the first trial of each block (1.6%), posterror trials (3%), target repetitions (11.3%), and, for RT analyses, error trials (3%) were excluded. AMT data from one MD+T- and one MD-T- participant were missing.

For longitudinal analyses, as a first step, participants from the MD+T+, MD+T-, and MD-T- groups with available longitudinal data were re-examined regarding baseline demographic and clinical differences. Simon task data from one MD+T+ and three MD-T- participants, and AMT data from one MD+T- and two MD-T- participants were missing. Again, the first trial of each block (1.6%), posterror trials (baseline: 2.8%, follow-up: 2.7%), target repetitions (baseline: 11.5%, follow-up: 10.9%) and, for RT analyses, error trials (baseline: 2.8%, follow-up: 2.7%) were excluded. Repeated-measures ANOVAs with time [2, baseline vs. follow-up] as within-subject and group [3, MD+T+ vs. MD+T- vs. MD-T-] as between-subject factor were applied to assess clinical and cognitive changes over CBT.

Exploratory linear/logistic regression analyses were conducted for examining the predictive value of lifetime trauma history (PDS; yes/no) for changes of depressive symptom severity (BDI-II) and dropout from care as core outcome measures, respectively. Due to the small sample size for the longitudinal analyses, and the high correlations between depressiveness (BDI-II) and the subjectively evaluated therapy effects (VEV-R, r = -.66, p < .001) at the follow-up assessment, we decided to omit the VEV-R from the predictive analyses. For the BDI-II, a change score was computed by subtracting baseline from follow-up values, and baseline values were included as a covariate to the regression analyses. As a second step, baseline cognitive performance (centered around the mean to avoid multicollinearity issues), and, as a third, the interaction of lifetime trauma history (yes/no) and baseline cognitive performance were added to the model.

Whenever hypothesis testing referred to one major cognitive domain (i.e., EF and learning/memory) and were not exploratory in nature, Holm-Bonferroni correction (Holm, 1979) for family-wise error (FWER) per respective domain was applied. As the assumptions of conventional GLMs (ANOVA, linear regression) are frequently violated in psychological data possibly leading to poor power and inaccurate effect sizes (Field & Wilcox, 2017), we repeated hypothesis testing using robust regressions. These drop GLM assumptions by using a robust sandwich estimation of standard errors, down-weighting observations with large residuals, and omitting outlying residuals (Royall, 1986). Predictive analyses were repeated using mixed-effects regressions with random intercept parameter addressing regression to the mean, which can otherwise yield biased results (Oberg & Mahoney, 2007). However, due to the higher prevalence and familiarity of conventional GLMs in the field, whenever both analyses yielded the same results, conventional GLMs were reported.

Sample Characteristics, Clinical Symptomatology, and Baseline Cognitive Functioning

The groups were well-matched regarding age, sex, and physical diseases (all ps ≥ .134, see Table 1). However, group differences emerged for educational status ( ${\mathrm{{\rm X}}}_{\mathrm{12}}^2$ = 22.26, p = .035) and smoking ( ${\mathrm{{\rm X}}}_{\mathrm{3}}^2$ = 9.24, p = .026). Furthermore, both clinical groups reported higher medication intake than the non-clinical ones ( ${\mathrm{{\rm X}}}_{\mathrm{3}}^2$ = 21.02, p < .001), mainly driven by psychiatric medication. However, including these variables as covariates did not change the cross-sectional results. For depressive symptom severity (BDI-II), both MD+T- and MD+T+ individuals reported higher levels than the control groups (all ps ≤ .001), with post-hoc analyses indicating no difference between them. For number and frequency of DSM-IV A1 traumatic events, both MD+T+ and MD-T+ scored higher than MD-T- individuals, with MD+T- individuals in between (THQ, all ps ≤ .006). For the severity of childhood maltreatment, both MD+ groups as well as the MD-T+ participants scored higher than the MD-T- group (CTQ, all ps ≤ .005).

No group differences emerged for forward, backward, and overall digit span (all ps ≥ .283, see Table 2). For the Simon task, groups differed regarding conflict adaptation of median RTs with a medium effect size, F(3, 143) = 3.23, p = .024, ${\mathrm{\eta }}_{\mathrm{p}}^2$ = .063, 90% CI [0, .12], see Figure 1), with higher levels in MD+T+ compared to MD-T- individuals (p = .017) and no other differences (all ps ≥ .43). Neither for conflict adaptation of mean PEs, nor for interference susceptibility did group differences emerge (all ps ≥ .424). Regarding OGM, for positive and negative words and the overall score, MD+T- participants scored lower than both MD-T+ and MD-T- ones with, again, medium effect sizes (all ps ≤ .002), and no other differences (all ps ≥ .118).

While OGM results remained stable after Holm-Bonferroni correction for FWER, the group difference for conflict adaptation of median RTs lost statistical significance (p = .168). Applying robust regressions did not considerably change the results, except for the difference between MD-T+ and MD-T- participants regarding conflict adaptation of median RTs and the interference effect of median RTs emerging as non-significant trends, β = -15.2, 95% CI [-30.7, 0.2], p = .053 and β = -11.5, 95% CI [-24.2, 1.2], p = .076).

Clinical and Cognitive Treatment Outcome Under Consideration of Lifetime Trauma History

MD+T+ (n = 21), MD+T- (n = 16), and MD-T- participants (n = 34) available for longitudinal analyses did not differ regarding baseline demographic/health-related characteristics (all ps ≥ .136, see Supplementary Materials), except for higher medication intake in both MD groups ( ${\mathrm{{\rm X}}}_{\mathrm{2}}^2$ = 13.9, p = .001). However, including it as a covariate did not affect the longitudinal results. MD+T+ individuals reported a higher number of DSM-IV A1 traumatic events (THQ) than MD+T- ones, which, in turn, reported more than MD-T- individuals (all ps ≤ .036). With respect to their frequency (THQ), as well as for childhood maltreatment severity (CTQ), both MD+ groups scored higher than the MD-T- one (all ps ≤ .035 and all ps ≤ .002, respectively).

Notably, while CBT led to substantial clinical improvements, MD+T+ and MD+T- individuals did not differ regarding depressive symptom changes (BDI-II), subjectively evaluated therapy effects (VEV-R), and percentage of dropouts (all ps ≥ .605, see Table 3). Furthermore, no cognitive improvements over CBT in the clinical groups emerged (all ps ≥ .272, see Table 3). However, for digit span, medium-to-large time effects indicated better performance at follow-up over all groups (all ps ≤ .009). Robust regressions yielded similar results.

Regression analyses on the predictive value of lifetime trauma history (yes/no) for therapy outcome (BDI-II changes of depressive symptom severity and dropout status, respectively) yielded no associations (all ps ≥ .391). When, in a second step, adding respective facets of baseline cognitive functioning, more pronounced reductions of depressive symptom severity (BDI-II) emerged with smaller forward digit span, b = 1.48, 95% CI [0.07; 2.90], p = .041), while for all other measures of cognitive functioning, no predictive value emerged (all ps ≥ .059). Adding, in a third step, interaction terms of lifetime trauma history (yes/no) and baseline cognitive functioning did not predict CBT outcome regarding BDI-II and dropout status (all ps ≥ .058). Notably, robust regressions led to similar results.

Baseline Cognitive Functioning

On a cross-sectional level, the results support the role of lifetime trauma history for cognitive functioning in MD. While no differences emerged for interference susceptibility and WM, MD+T+ patients showed higher conflict adaptation of median RTs than MD-T- participants, with MD+T- and MD-T+ in between. This corresponds with previous findings from our group of more pronounced conflict adaptation in traumatized individuals with and possibly also without PTSD (Schindler et al., 2020; Steudte-Schmiedgen et al., 2014). However, as there also are suggested associations of conflict adaptation and depressive symptom severity (van Steenbergen et al., 2012), albeit without considering trauma history, further studies are desirable.

Interestingly, autobiographical memory yielded contrasting findings: MD+T- patients showed more pronounced OGM compared to the healthy control groups, corresponding with our previous findings of OGM in PTSD, but not trauma exposure per se (Schindler et al., 2020), and suggestions from reviews and meta-analyses (Moore & Zoellner, 2007; Ono et al., 2016; Sumner et al., 2010; Williams et al., 2007). Further, it supports the findings of Kuyken, Howell, and Dalgleish (2006) of OGM only in MD without (childhood) trauma history, but not those of Aglan et al. (2010) of OGM in MD with history of CSA. In sum, neither the results for conflict adaptation, nor those for OGM speak for a mere additive effect of trauma and MD on cognitive functioning, but rather for complex patterns with different impacts on different processes, and, potentially, different implications for clinical practice.

Clinical and Cognitive Treatment Outcome Under Consideration of Lifetime Trauma History

In contrast to several previous studies particularly on childhood trauma (reviewed in Nemeroff, 2016; Teicher & Samson, 2013), our data suggest CBT to be equally effective in individuals with MD with and without the history of at least one traumatic event according to the DSM-IV. Several aspects may contribute to this divergence. Firstly, it is plausible that lifetime trauma, as examined in this study, does exert different effects than childhood trauma. Importantly, in our study, MD+T+ and MD+T- groups reported equal CTQ childhood maltreatment severity, and it is conceivable that this may have contributed to lacking group differences with respect to CBT effectiveness. Notably, also with respect to the THQ, the MD+T+ and the MD+T- groups did only differ on a descriptive level. However, it is important to consider that this instrument refers to the number and frequency of potentially traumatic events, for which the presence of the complete DSM-IV criteria are not checked. In order to better understand the role of childhood and adulthood trauma for CBT effectiveness, studies explicitly contrasting individuals with MD (i) without lifetime trauma, (ii) with exclusively childhood, and (iii) with exclusively adulthood trauma as defined by the current diagnostic criteria are necessary. Furthermore, treatment differences might have played a role. Most prominently, the majority of studies reporting similar therapy outcome for MD with and without (particularly childhood) trauma history had applied combined psychotherapy and antidepressant medication (Lewis et al., 2010; Miniati et al., 2010; Nemeroff et al., 2003; but Asarnow et al., 2009), as was the case for approximately half of our sample. Further, we cannot rule out whether, in our study, trauma status had led to slight individual treatment adaptations by the responsible therapists. This might, for instance, have led to combined modifications of trauma-related and -unrelated automatic thought patterns, or the encouraging of restarting activities avoided after the trauma during behavioral interventions within the context of the utilized CBT manuals (Hautzinger, 1998, 2008). Thus, future studies applying more strictly manualized CBT and investigating larger MD groups with and without medication intake are required.

Additionally, the results corroborate previous findings of cognitive alterations in MD being highly stable over CBT (reviewed in Köhler et al., 2015; Moore & Zoellner, 2007; Snyder & Hankin, 2019), and of this to be irrespective of trauma history. While WM improved from baseline to follow-up, this is presumably attributable to practice/habituation effects, as it also concerned MD-T- individuals. As cognitive impairments are assumed to be associated with worse psychosocial functioning and increased relapse risk in MD (Rock et al., 2014), the continuous finding of this to not be adequately addressed by CBT shows the necessity to strive for “cognitive”, next to clinical remission in MD (Bernhardt et al., 2019; Bortolato et al., 2016). For example, this might be achieved by directly targeting cognitive functioning during MD-centered CBT. While research on EF training in MD is still in its infancy (for a meta-analysis, see, e.g., Motter et al., 2016), there are promising results that OGM, as well as MD symptomatology itself may be influenced by interventions directly focusing on autobiographical recall, albeit with long-term stability still questionable (for a meta-analysis, see Barry, Sze, & Raes, 2019).

The exploratory predictive analyses on lifetime trauma history and cognitive functioning for CBT do not provide clear results from which robust next steps could be derived. What can be clearly stated as of now is that there, again, was no evidence for a relevant role of lifetime trauma history. Further, only a singular association with cognitive parameters emerged, suggesting smaller WM to be associated with more pronounced depressiveness-related CBT effects. In sum, this pattern, albeit stemming from a very small sample size, supports the findings of Goodkind et al. (2016) on interference susceptibility, but stands at variance with those of Sumner et al. (2010) suggesting a predictive role of autobiographical memory specificity in MD. Future studies are needed to follow up on autobiographical memory in this context, or investigate whether other cognitive markers might be more suitable to predict clinical outcome after standardized psychotherapeutic/pharmacological treatment (e.g., Groves et al., 2018) with or without taking trauma history into account.

Strengths, Limitations, and Outlook

One central strength of the study is the naturalistic, highly ecologically valid study design. While the inclusion of a waiting control group of MD+T+/MD+T- patients not receiving CBT was impossible for ethical reasons, the fact that a healthy control group was studied longitudinally alongside the MD individuals is a further major strength, as it allowed the separation of CBT-associated and mere practice effects on cognitive functioning. However, limitations resulting from the naturalistic design are the heterogeneous manifestations of psychopathology and medication and the group differences in educational status and smoking. Further limitations include the lack of an objective, observer-rated outcome of depressiveness (e.g., the Hamilton Rating Scale for Depression; Hamilton, 1960), as well as the small sample sizes and the thus reduced statistical power for detecting especially interactive relationships. However, the fact that the vast majority of associations were confirmed in robust analyses corroborates the validity of the findings. Finally, behavioral tasks established in cognitive psychology, such as the ones used in our study, are characterized by task impurity, which describes the impossibility of assessing “pure” cognitive processes without simultaneously eliciting others (Miyake et al., 2000; e.g., Scott et al., 2015). In order to maximize transparency in data reporting, we chose to report subscale scores of the cognitive tasks for which different properties are discussed (Botvinick et al., 2004; Wechsler, 1997; Williams & Broadbent, 1986). In addition, we acknowledge that for any of the assessed tasks, additional cognitive processes such as processing speed, attention, and motivation – while not directly studied – are inevitably involved.

Conclusions

In conclusion, the study is the first to examine lifetime trauma history, cognitive functioning, and their interaction in the context of CBT in patients with MD. On a cross-sectional level, conflict adaptation and autobiographical memory specificity emerged to be differentially affected in MD with and without lifetime trauma history. Contrary to previous research on childhood trauma, we found no evidence for a differential treatment response in patients with MD with and without lifetime trauma history as defined by the DSM-IV. Further, the cognitive parameters were stable over CBT, and only a singular predictive association of forward digit span, but no other facets of baseline cognitive functioning, lifetime trauma history, or their interaction with treatment outcome emerged. These insights into the interaction between lifetime trauma history and cognitive functioning provide unique extensions for research on MD psychopathology and treatment and underline the relevance of “cognitive” remission (Bernhardt et al., 2019; Bortolato et al., 2016). For achieving this aim, further research is required to allow more profound, neuroscience-informed diagnostic processes and personalized, multi-modal treatment approaches depending on patients’ individual manifestation of cognitive functioning (De Raedt, 2020).