Childhood Maltreatment and Cognitive Performance in Bipolar Disorder: The Potential Role of Inflammation

Abstract

Background:

Cognitive deficits are common in individuals with bipolar disorder (BD), but there is considerable variability in cognitive functioning. Childhood maltreatment (CM), which is frequently reported in BD, has been linked to poorer cognitive performance, potentially through mechanisms such as inflammation. However, the relationship between CM and global cognition, and the mediating role of inflammation in BD warrant further investigation.

Methods:

The study sample consisted of 112 BD individuals and 83 healthy controls (HC). Participants completed the MATRICS Consensus Cognitive Battery (MCCB), the Wisconsin Card Sorting Test, and the Childhood Trauma Questionnaire (CTQ). A composite inflammation index was created using blood levels of C-reactive protein (CRP), interleukin (IL)-6, and tumor necrosis factor (TNF)-α, and was used in primary analyses.

Results:

The BD group, compared to HC, showed higher levels of inflammation and CM. Across the entire sample, higher total CM was associated with poorer global cognitive performance, with a medium effect size, even after accounting for diagnostic status. The associations were evident across all CM subscales. Specific cognitive domains affected included speed of processing, working memory, visual learning, and reasoning and problem solving. The association between CM and poorer global cognitive performance was partially mediated by inflammation (indirect effect: beta= −.048, CI= −.10, −.009). Within the BD group, higher total CM was similarly associated with worse global cognitive performance. The associations were evident across all CM subscales, except for physical neglect. Significant associations were observed between total CM and MCCB domains of speed of processing, attention and vigilance, working memory, visual learning, reasoning and problem solving, as well as cognitive flexibility. Within the HC group, only emotional neglect and physical neglect were associated with poorer global cognition.

Conclusions:

This study provides evidence that total CM and its subscales are associated with poorer global cognitive performance in a sample of individuals with BD and HC, with stronger associations found within the BD group. In addition, inflammation partially mediated the relationship between CM and global cognition. These findings highlight the importance of trauma-informed and cognition-focused interventions aimed at enhancing cognitive outcomes and slowing cognitive decline in individuals with BD who have a history of CM. Furthermore, the results suggest that while inflammation plays a role in the CM-cognition link, its effects are complex and likely interact with other biological and environmental factors.

1. Introduction

Cognitive deficits are common in individuals with bipolar disorder (BD). However, there is considerable heterogeneity in cognitive functioning among individuals with BD, with some experiencing significant deficits while others show minimal impairment.¹ Studies indicate that 12%−40% experience global impairments across multiple cognitive domains, including attention, psychomotor speed, verbal memory, and executive function. Additionally, 29%−40% demonstrate selective deficits in attention and psychomotor speed, while 32%−48% maintain relatively intact cognitive functioning compared to healthy, age-matched individuals.^2–6 Notably, deficits in attention, speed of processing, memory, and aspects of executive functioning often persist even during remission.^5–8 These cognitive deficits are strongly associated with poor quality of life, real-world functioning, and disability.^2,9–13 Therefore, understanding the mechanisms underlying this variability is critical for identifying potential targets for intervention.

Given this heterogeneity, identifying factors that contribute to cognitive variability in BD is essential. One such contributor is exposure to stressful environmental factors, such as childhood maltreatment (CM)—a broad term encompassing emotional, physical, and sexual abuse, as well as emotional and physical neglect.¹⁴ CM is widely recognized for its lasting impact on mental health¹⁵ and has been linked to adult cognitive impairment in healthy individuals.¹⁶ However, relatively few studies have examined its relationship with cognition in individuals with BD.^17–19 A recent meta-analysis of 20 studies examining the association between CM and cognitive performance in BD found that CM exposure was associated with worse cognitive outcomes.¹⁹ Of these, only two studies assessed global cognitive functioning,^20,21 and six assessed general IQ.^20,22–26 This underscores the need for further research employing comprehensive cognitive batteries to assess global cognitive performance in BD, which remains underrepresented in the existing literature. Given that CM is at least two times more common in BD than in the general population, and it is associated with a threefold increased risk of developing the disorder,²⁷ it may play a critical role in BD-related cognitive dysfunction.^17,18 Despite this potential relevance, the specific role of CM in BD-related cognitive dysfunction is not yet well characterized.

Emerging evidence suggests that CM is associated with poorer performance across multiple cognitive domains, including attention, memory, and executive functioning, in adults with BD.^{20,22,26,28,29} The few studies that evaluated the effect of CM on global cognitive functioning have reported a significant relationship,¹⁹ with effects observed in both individuals with BD and healthy controls.²⁰ This suggests that CM may have a more general negative impact on cognition, rather than a BD-specific one, underscoring the need for further research to clarify the role of CM as a contributor to cognitive impairment in BD.

One potential mechanism linking CM to cognitive dysfunction is inflammation. Research suggests that CM affects both brain development—particularly in the hippocampus and prefrontal cortex³⁰—and immune function³¹, with downstream consequences for cognition. Meta-analytic findings show that adults with a history of CM have elevated levels of inflammatory markers, including C-reactive protein (CRP), interleukin-6 (IL-6), and tumor necrosis factor-alpha (TNF-α).³² However, findings across studies are mixed, likely due to methodological inconsistencies. Kerr et al. (2021) conducted a systematic review and found more consistent associations between CM and elevated inflammation in prospective studies, whereas studies relying on retrospective self-report often found no association or associations that attenuated after adjusting for covariates.³³ Chronic inflammation, whether central or peripheral, can further impact the prefrontal cortex and hippocampus, leading to deficits in memory and executive function.¹⁷ Meta-analyses confirm that patients with BD exhibit elevated peripheral pro-inflammatory markers compared to healthy controls during acute phases of the illness.^34,35 Although pro-inflammatory cytokine levels tend to normalize between mood episodes;^34,35 repeated episodes result in a cumulative burden that leads to persistently low-grade inflammation, even during periods of remission.³⁴ This pattern mirrors the trajectory of cognitive impairment observed in some individuals with BD.³⁶ Notably, elevated levels of inflammatory markers such as CRP, IL-6, IL-1 family molecules, and TNF family molecules have been linked to poorer cognitive functioning in BD, particularly in domains such as speed of processing, memory, and executive function.^36–38 A recent study identified white matter subtypes in individuals with bipolar II depression. One subtype with altered white matter integrity had elevated inflammatory markers, greater childhood emotional maltreatment, and increased symptom severity—highlighting the potential interplay between early-life stress and the neuroimmune and neurodevelopmental pathways involved in the disorder.³⁹

Recent research indicates that systemic inflammation could act as a mechanism connecting CM to cognitive deficits later in life.⁴⁰ For example, inflammation (a latent measure of IL-6, TNF-α, and CRP) has been shown to mediate the association between childhood physical neglect and worse cognitive performance across diverse domains (i.e., logical memory, full scale IQ, emotion recognition) in a combined sample of individuals with schizophrenia and healthy controls.⁴¹ Similarly, data from the Midlife in the United States (MIDUS) study found that, in older adults, systemic inflammation partially mediated the association between childhood maltreatment and poorer episodic memory performance.⁴² While prior research has examined associations between CM, inflammation, and cognitive functioning in BD, further research is needed to determine whether inflammation serves as a key mediating pathway linking CM to cognitive dysfunction within this population. Importantly, our study addresses this gap by measuring cognitive performance using a comprehensive battery, allowing for a more detailed and nuanced assessment of cognitive domains affected by CM and inflammation in BD.

1.1. Aims of the Study

The primary objective of this study is to investigate the associations between CM and cognitive performance in patients with BD and healthy controls. Additionally, in exploratory analyses conducted on a subset of participants with available inflammatory data, we examined the relationship between CM and inflammatory markers, and assessed the mediating role of inflammation in the link between CM and cognitive performance. We hypothesized that a) more severe CM would be associated with poorer cognitive performance across all participants, with stronger associations expected in the BD group vs. HC; and b) higher levels of CM would be associated with increased inflammation, and inflammation would partially mediate the negative effect of CM on cognitive performance.

2. Methods

2.1. Participants

Participants for this study were recruited from the Boston, MA metropolitan area. The sample included individuals diagnosed with bipolar disorder (BD) as well as healthy controls (HC). Inclusion criteria for the BD group included age 18 to 68 years, English fluency, and a diagnosis of BD I or II according to the Structured Clinical Interview for DSM-5 (SCID-5)⁴³. We did not require strict euthymia, rather we included individuals who were outpatients who were affectively stable enough to participate in the study, as indicated by a Clinical Global Impression–Bipolar (CGI-BP) overall severity score of no more than moderately ill. Individuals were excluded if they had a history of central nervous system trauma, neurological disorders, attention deficit hyperactivity disorder diagnosed and treated during childhood, known learning disability, or electroconvulsive therapy within the past year. Other exclusionary criteria included current diagnoses of mild cognitive impairment, dementia, active substance use disorder within the past three months (based on SCID), or a medical condition affecting cognitive functioning that was active or unstable. Inflammatory-related medical conditions were not excluded unless they were deemed unstable based on patient-reported medical history (e.g., recent hospitalizations), and these conditions were controlled for in relevant statistical analyses. Eligibility criteria for the HC group were similar to those for the BD group but excluded individuals with any current or past Axis I psychiatric diagnosis per SCID.

2.2. Procedure

This investigation is part of an ongoing longitudinal research project examining cognitive and functional outcomes in BD. The current analysis used data exclusively from the baseline visit.

2.3. Measures

2.3.1. Neurocognitive Measures.

Cognitive performance was measured across six domains included in the MATRICS Consensus Cognitive Battery (MCCB): attention and vigilance, speed of processing, working memory, visual learning, reasoning and problem-solving, and social cognition.^44,45 For verbal learning, the California Verbal Learning Test (CVLT) was administered in place of the MCCB’s standard Hopkins Verbal Learning Test (HVLT) to provide a more challenging assessment.⁴⁶ Age- and sex-normative T-scores from the CVLT were substituted for the HVLT scores. Composite scores from the MCCB were used to represent global cognitive function, with adjustments made for age and sex based on standardized norms (T-score scale: mean=50, SD=10). A subset of participants completed the Wisconsin Card Sorting Test (WCST) (N=78HC+87BD) to measure cognitive flexibility.⁴⁷ The number of completed categories and perseverative errors were recorded, with perseverative errors reported as T-scores adjusted for sex, age, and education. Premorbid IQ was estimated using the Wide Range Achievement Test-3 Reading (WRAT-3), which evaluates single-word reading ability and is resistant to cognitive decline from neurological disorders.⁴⁸

2.3.2. Clinical Measures.

The SCID-5 was utilized to verify diagnostic eligibility. Depressive and manic symptoms were assessed using the 24-item Hamilton Depression Rating Scale⁴⁹ (HDRS-24) and the Young Mania Rating Scale⁵⁰ (YMRS), respectively, with mood assessments conducted within two weeks of neurocognitive testing.

2.3.3. Childhood Trauma Questionnaire.

The Childhood Trauma Questionnaire (CTQ) total score was used to assess experiences of childhood maltreatment (as a child and a teenager up until age 16), including physical abuse, emotional abuse, sexual abuse, physical neglect, and emotional neglect.⁵¹ Participants rated each item on a 5-point scale: “1 (never true), 2 (rarely true), 3 (sometimes true), 4 (often true), 5 (very often true).” Each subscale consisted of 5 items, totaling 25 items, with scores ranging from 5 (indicating no history of abuse or neglect) to 25 (indicating a very extreme history of abuse or neglect). A total childhood maltreatment score was calculated by summing the scores of the five subscales. We assessed whether a subject had a positive history of childhood maltreatment in a specific category using cut-off scores referenced in previous studies.^20,26 The thresholds were ≥13 for emotional abuse, ≥10 for physical abuse, ≥8 for sexual abuse, ≥10 for physical neglect, and ≥15 for emotional neglect. Participants were deemed to have a positive history of CM if any one of the subscale scores met or exceeded the specified cut-off. In this study, Cronbach’s alpha for total CM was .94 for the BD group and .92 for the HC group. Regarding CM subscales, Cronbach’s alpha for the BD group was .89 (emotional abuse), .85 (physical abuse), .96 (sexual abuse), .93 (emotional neglect), and .84 (physical neglect); for the HC group, the alphas were .90, .72, .89, .84, and .67, respectively. The CTQ includes a three-item Minimization/Denial subscale to detect potential underreporting of CM. In line with previous studies, we operationalized this subscale categorically, with scores ≥1 (i.e., endorsing “very often true” on at least one item) indicating possible minimization.^52,53 Sensitivity analyses including this variable as a covariate were conducted to account for response bias.

2.3.4. Inflammatory Markers.

Blood samples were collected at the Brigham Center for Clinical Investigation, processed, and stored at −80°C until analysis. Immune markers were batch-analyzed at the Brigham Research Assay Core. C-reactive protein (CRP) and tumor necrosis factor-alpha (TNF-α) levels were measured using Enzyme-linked immunoassay (ELISA, R&D Systems, Inc., Minneapolis, MN). IL-6 levels were quantified using the Access Chemiluminescent Immunoassay Systems (Beckman Coulter, Fullerton, CA). To control for outliers, values exceeding 3 SDs above the mean were winsorized to 3 SDs for each inflammatory marker, an approach that has been used in prior work.⁵⁴

2.4. Analytical Plan

Statistical analyses were performed using SAS (version 9.4; SAS Institute, Cary, NC) and IBM SPSS Statistics (version 22.0; Armonk, NY). Statistical significance was defined as p<.05. For between group comparisons, either nonparametric Mann-Whitney U tests (continuous measures) or Chi-square tests (categorical measures) were employed. A multivariate analysis of covariance (MANCOVA) examined group differences (BD vs. HC) in cognitive function and CM, accounting for relevant covariates. An exploratory MANCOVA examined whether cognitive functioning differed across CM categories (i.e., participants with and without a history of CM). Analyses of covariance (ANCOVA) were conducted to further explore group differences in the inflammation composite score. Additionally, multiple linear regression analyses assessed the relationship between CM, cognitive performance and inflammation in the full sample and within each group. Standardized beta estimates and 95% confidence intervals (CI) are reported for each association. In addition, partial eta-squared (ηp2) is reported as the estimate of effect size. The values of .01, .06 and .14 for partial eta-squared have been suggested to represent small, medium, and large effect sizes, respectively.⁵⁵

Dimensional Approach.

To increase statistical power and investigate broader patterns across the full sample, multiple linear regression and mediation analyses were applied to the combined sample of BD patients and controls. This approach aimed to determine whether the relationships between CM, inflammation, and cognitive performance reflect general trends or are specific to individuals with BD.

Inflammatory markers.

Log-10 transformation of inflammatory data was conducted prior to analyses to normalize the data. To reduce the risk of type I error from multiple comparisons, a composite inflammation index was created from CRP, IL-6, and TNF-α. Principal component analysis (PCA) with varimax rotation supported a single inflammatory factor. This approach has been used by prior work.^54,56 The Kaiser-Meyer-Olkin (KMO) measure of sampling adequacy was .63,⁵⁷ with factor loadings between .69 and .84. This composite index was included in regression and mediation models.

Mediation Analysis.

Mediation analysis was performed using the PROCESS macro for SAS (version 4.3) to evaluate whether inflammation mediates the relationship between CM and global cognitive performance. The inflammation composite was examined as the mediator, with age and sex included as covariates. We used the percentile bootstrap confidence interval, a widely recommended method for inference about the indirect effect in mediation analysis, which has been shown to balance validity and power considerations without relying on normality and homoscedasticity assumptions.^58,59 The PROCESS macro generated 95% bootstrap confidence intervals based on 5,000 resamples.

3. Results

The demographic and clinical characteristics of the sample are shown in Table 1. The study sample consisted of 112 individuals with BD (mean age: 42.2 ± 14.9) and 83 healthy controls (mean age: 46.9 ± 14.7). The BD group had a mean depression score of 7.1 ± 6.6 and a mean mania score of 2.1 ± 3.1. Among the BD patients, 72% were diagnosed with BD I, and the remainder had BD II. Approximately 64% of the BD group were euthymic and 46% had a lifetime history of psychosis (noncurrent).

Table 1.

Characteristics of the sample

Variables	BD (N = 112)	HC (N = 83)	P value
Sex: N (%)	F: 87 (78%)	F: 48 (58%)	.003
Age: Mean (SD)	42.2 (14.9)	46.9 (14.7)	.04
BMI: Mean (SD)	29.3 (8.6)	25.8 (4.4)	.02
Race: N (%)
White or Caucasian	86 (77%)	56 (67%)
Black or African American	14 (12.5%)	9 (11%)	.03
Asian	5 (4.5%)	8 (10%)
Other	7 (6%)	10 (12%)
Education; in years: Mean (SD)	16.0 (2.3)	16.6 (2.0)	.04
Premorbid IQ (WRAT-3): Mean (SD)	110.3 (10.1)	107.8 (8.9)	.02
MCCB Composite T Score: Mean (SD)	47.2 (7.5)	51.9 (6.0)	<.001
HDRS-24 Score: Mean (SD)	7.1 (6.6)	.25 (.84)	<.001
Range	0–27	0–5
YMRS Score: Mean (SD)	2.1 (3.1)	.02 (.15)	<.001
Range	0–15	0–1
Euthymiaa: N (%)	72 (64%)	NA
BD-I: N (%)	81 (72%)	NA
Medical comorbidity: N (%)	44 (39%)	11 (13%)	<.001
History of Psychosis: N (%)	52 (46%)	NA
Number of Prior Hospitalizations: Mean (SD)	5.7 (12.2)	NA
Total Number of Prior Mood Episodes: Mean (SD)	27.4 (57.0)	NA
Total Number of Prior Depressive Episodes: Mean (SD)	14.9 (46.3)	NA
Total Number of Prior Manic Episodes: Mean (SD)	6.1 (28.3)	NA
Number of psychotropic medications: Mean (SD)	2.4 (1.3)	NA
Duration of Illness; in years: Mean (SD)	22.8 (18.8)	NA
CTQ Toal Score: Mean (SD)	48.1 (17.6)	33.1 (9.7)	<.001
Emotional Abuse: Mean (SD)	12.3 (5.7)	7.3 (3.5)	<.001
Physical Abuse: Mean (SD)	7.7 (3.9)	6.0 (1.8)	.003
Sexual Abuse: Mean (SD)	7.8 (4.8)	5.4 (1.5)	<.001
Emotional Neglect: Mean (SD)	12.3 (5.5)	8.3 (3.6)	<.001
Physical Neglect: Mean (SD)	8.0 (3.7)	5.9 (1.9)	<.001
Inflammatory Markers: Mean (SD)	N=77	N=48
CRP (mg/L)	2.3 (2.6)	1.6 (1.7)	.18
IL-6 (pg/ml)	2.4 (1.9)	1.9 (2.9)	.026
TNF-α (pg/ml)	1.0 (.49)	.83 (.25)	.06

^aEuthymia was defined based on YMRS≤7 and HAMD-24≤9.

Abbreviations: BMI, Body Mass Index; BD, Bipolar Disorder; CTQ, Childhood Trauma Questionnaire; HDRS-24, Hamilton Depression Rating Scale ─ 24 items; HC, Healthy Control; WRAT-3, Wide Range Achievement Test; YMRS, Young Mania Rating Scale

3.1. BD versus HC

Significant differences were observed between BD and HC groups in age, BMI, education, premorbid IQ estimates, medical comorbidity, and sex and race composition (ps<.05; Table 1). The BD group performed significantly worse on the MCCB composite score (F=35.8, p<0.001) compared to the HCs, controlling for BMI, education, race (white/nonwhite), premorbid IQ, and medical comorbidity. Significant differences were evident with respect to MCCB speed of processing (F=22.5, p<.001), attention and vigilance (F= 6.5, p=.01), working memory (F=7.8, p=.006), visual learning (F=21.2, p<.001), reasoning and problem solving (F=13.8, p<.001), and verbal learning (F=24.9, p<.001), but not social cognition (F=1.2, p=.30). No significant group differences were observed on WCST measures (Fs<.1, ps>.2).

BD patients reported significantly higher total childhood maltreatment (i.e., total CTQ) scores, as well as higher scores across five subscales of emotional abuse, physical abuse, sexual abuse, emotional neglect and physical neglect, compared to HCs, controlling for age, sex, race, premorbid IQ, education, medical comorbidity, depressive symptoms and BMI (ps<.05; Table 1). A significantly larger proportion of the BD group (66%) had a history of CM, defined as having at least one subscale exceeding the predefined cutoffs, compared to 22% of the HC group (p<.001). Compared to HCs, the BD group had higher levels of inflammation composite (F=4.20, p=.04), controlling for age, sex, race, BMI, education and medical comorbidity.

3.2. Demographic and Clinical Correlates of CM

In the entire sample, higher total CTQ was associated with lower education (r=−.24, p<.001), lower premorbid IQ (r=−.15, p=.03) and higher BMI (r=.22, p=.002). Women reported higher total CTQ than men (43.7±17.9 vs. 37.3±12.0; t=2.92, p=.004). No significant differences in total CTQ scores were observed by race (p=.62). A trend-level difference in total CTQ scores was observed among individuals with medical comorbidities (p=.06). Based on these findings, education, premorbid IQ, BMI, and medical comorbidity were included as covariates in the relevant statistical analysis for the entire sample.

Within the BD group, no significant associations were observed between total CTQ and the following clinical variables: BMI, depressive and manic symptoms, age of onset, number of prior depressive or manic episodes, prior psychiatric hospitalizations, medication load, medical comorbidity, and the use of antidepressants, antipsychotics, and lithium (all ps>.1). However, negative associations were found between total CTQ and education (r=−.21, p=.03) as well as premorbid IQ (r=−.30, p=.001). Notably, higher total CTQ correlated with a greater number of prior suicide attempts (r=.30, p=.001). Post-hoc regression analysis revealed that the positive association between total CTQ and prior suicide attempts remained significant (beta=.25, p=.02, CI=.03, .47, partial eta-squared (ηp2)=.059) after controlling for BMI, education, premorbid IQ, depressive and manic symptoms, age of onset, number of mood episodes, prior psychiatric hospitalizations, medication numbers, medical comorbidity, and the use of antidepressants, antipsychotics, and lithium. Based on these findings, the number of suicide attempts was included as a covariate in the relevant regression models within the BD group, alongside antipsychotic use and BMI, which were associated with MCCB scores. No significant differences in total CTQ were observed across race or sex within the BD group. The prevalence of CTQ subscales and their clinical correlates are presented in the Supplemental Materials (Supplemental Tables 1 & 2). No significant differences were observed in total CTQ score, MCCB composite score, or inflammation composite score between euthymic and non-euthymic BD patients.

3.3. Childhood Maltreatment and Cognition

3.3.1. CM predicting cognitive functioning in the entire sample (n=195)

Higher total CTQ was associated with worse global cognitive performance, as measured by MCCB composite score, controlling for age, sex (as per corrected scores), diagnosis (BD vs HC), education, premorbid IQ, BMI and medical comorbidity (beta=−.25, p=.0004, CI=−.39, −.11, partial eta-squared (ηp2)=.066). This association was significant across five CTQ subscales (Table 2). Regarding MCCB domains, higher total CTQ showed significant associations with speed of processing, working memory, visual learning, reasoning and problem solving (ps<.05), a trend for attention and vigilance (p=.07) and non-significant associations with verbal learning and social cognition (Supplemental Table 3). No associations were observed between total CTQ and WCST measures after controlling for diagnosis (fully adjusted results are presented in Supplemental Table 3). Cognitive performance across CT categories is presented in supplemental materials. A non-significant interaction of total CTQ and diagnosis was observed on global cognitive performance (p=.18). Despite this, we examined the association between CT and cognitive performance within the BD and HC groups to explore potential effect sizes and better understand group-specific patterns.

Table 2.

Multiple regression using childhood maltreatment to predict MCCB composite score (i.e., global cognition) in the entire sample of BD patients and HCs, Beta, SE, confidence intervals, partial eta-square, and p values.

	b1 (SE)	95% Confidence Intervals		Partial Eta-Square	P value
CTQ Total Score → MCCB Composite Score
Model 1a	−.25 (.07)	−.39	−.11	.066	.0004**
Emotional Abuse → MCCB Composite Score
Model 1a	−.15 (.07)	−.29	−.01	.025	.03*
Physical Abuse → MCCB Composite Score
Model 1a	−.13 (.07)	−.27	−.003	.02	.04*
Sexual Abuse → MCCB Composite Score
Model 1a	−.22 (.06)	−.35	−.09	.055	.001**
Emotional Neglect → MCCB Composite Score
Model 1a	−.22 (.06)	−.35	−.09	.055	.001**
Physical Neglect → MCCB Composite Score
Model 1a	−.15 (.07)	−.28	−.02	.027	.02*

^**p < 0.01,

^*p < 0.05,

^†p < 0.1.

¹b denotes standardized beta.

^aModel 1 includes diagnosis, education, premorbid IQ, BMI and medical comorbidity as covariates.

3.3.2. CM predicting cognitive functioning within each diagnostic group

Within the BD group, higher total CTQ was associated with worse global cognitive performance controlling for age, sex (as per corrected scores), education, premorbid IQ, BMI, antipsychotic use and number of suicide attempts (beta=−.29, p=.001, CI=−.46, −.12, partial eta-squared (ηp2)=.098; Figure 1). This association was significant for emotional abuse, physical abuse, sexual abuse and emotional neglect, but not for physical neglect (Supplemental Table 4). Regarding MCCB domains, higher total CTQ was significantly associated with speed of processing, attention and vigilance, working memory, visual learning, reasoning and problem solving (ps<.05), but not for verbal learning or social cognition (Supplemental Table 4). In addition, higher CTQ was associated with worse performance on WCST measures. Figure 2 illustrates the effect sizes for the associations between CTQ total scores and cognitive performance within the BD and HC groups. Correlations between CTQ subscales and all MCCB domains in the BD sample are shown in Supplemental Table 5. Of note, the association between total CTQ and global cognition remained significant in the subgroup of euthymic BD patients (beta =−.30, p<.05).

Figure 1.

Association between childhood maltreatment (CTQ total score) and global cognition in BD patients and healthy controls.

Figure 2.

Partial-eta squared for the associations between total CTQ score and cognitive function in BD patients and healthy controls. *p<.05. The values of .01, .06 and .14 for partial eta-squared have been suggested to represent small, medium, and large effect sizes, respectively.

Covariates included in the BD sample: age & sex (as per corrected scores), education, premorbid IQ, BMI, antipsychotic use, number of suicide attempts. Covariates included in the HC sample: premorbid IQ, BMI, education.

Abbreviations: CTQ, Childhood Trauma Questionnaire; Composite, MCCB composite score; ProcSpeed, speed of processing; AttVig, attention and vigilance; WorkMem, working memory; VisLearn, visual learning; ReaProbSolv, reasoning and problem solving; VerbalLearn, Verbal Learning; SocCog, social cognition; WCSTCat, Wisconsin Card Sorting Test, number of completed categories; WCSTPersEr, WCST Perseverative Errors.

Within the HC group, higher levels of emotional neglect and physical neglect were associated with worse global cognitive performance controlling for education, premorbid IQ and BMI (Supplemental Tables 6). Higher total CTQ was associated with worse performance on MCCB verbal learning (Supplemental Tables 6). There were no other significant associations for total CTQ or any CTQ subscale with any other cognitive measures in the HC sample (Figure 1; Supplemental Tables 6–8).

As shown in Figure 1, a small number of outliers were detected in the CTQ data and MCCB composite scores. In a supplementary analysis, we removed the outliers from the dataset and re-ran the analysis. The results revealed that excluding these outliers did not significantly alter the findings. Also, including the denial subscale as a covariate in sensitivity analyses did not meaningfully alter the main findings.

3.4. Inflammation, Childhood Maltreatment, and Cognitive Performance

Inflammatory marker data were available for a subsample of participants (n=125; 77BD, 48HC). In the entire sample of BD patients and HC, age was positively associated with inflammation composite (r=.26, p=.004), and participants with medical comorbidity exhibited higher inflammation (t=3.74, p<.001). Consequently, age and medical comorbidity were included as covariates in subsequent regression models. Inflammation was not significantly associated with education or premorbid IQ (p>.05), and no significant sex differences in inflammation were observed (p>.10). In this subsample, BMI was modestly correlated with CM (r=.28, p=.001) and strongly correlated with the inflammation composite (r=.60, p<.001). BMI is likely part of the causal pathway linking CM, inflammation, and cognitive functioning. As such, BMI was not included as a covariate in the regression analysis examining the association between CM and inflammation to avoid overadjustment that could obscure these relationships. Additionally, the limited sample size for inflammatory data precluded our ability to test a more complex serial mediator model that could better account for BMI’s potential mediating role. This limitation and topic are further elaborated upon in the discussion section.

In the entire sample, regression analysis revealed that higher total CTQ was associated with higher inflammation composite scores (beta=.19, p=.02, partial eta-squared (ηp2)=.042), controlling for age and medical comorbidity (Table 3). The association between total CTQ and inflammation was no longer significant when diagnosis was added as a covariate, suggesting that controlling for the shared variance between diagnosis and CM diminishes the unique contribution of CM to inflammation. No significant interaction between total CTQ and diagnosis was observed on inflammation (p=.60).

Table 3.

Multiple regression using childhood maltreatment to predict inflammation in the entire sample of BD patients and HCs, Beta, SE, confidence intervals, partial eta-square, and p values.

	b1 (SE)	95% Confidence Intervals		Partial Eta-Square	P value
CTQ Total Score → Inflammation composite
Model 1a	.19 (.08)	.03	.36	.042	.02*
Model 2b	.11 (.10)	−.09	.30	.01	.28
Emotional Abuse → Inflammation composite
Model 1a	.21 (.08)	.04	.37	.049	.014*
Model 2b	.13 (.09)	−.05	.32	.016	.16
Physical Abuse → Inflammation composite
Model 1a	.22 (.08)	.06	.39	.056	.008**
Model 2b	.17 (.09)	.001	.35	.032	.048*
Sexual Abuse → Inflammation composite
Model 1a	.14 (.08)	−.02	.31	.023	.09
Model 2b	.07 (.09)	−.10	.25	.005	.42
Emotional Neglect → Inflammation composite
Model 1a	.11 (.08)	−.06	.27	.013	.21
Model 2b	−.002 (.09)	−.19	.18	.00	.98
Physical Neglect → Inflammation composite
Model 1a	.09 (.08)	−.07	.26	.01	.26
Model 2b	.004 (.09)	−.18	.18	.00	.96

^**p < 0.01,

^*p < 0.05,

^†p < 0.1.

¹b denotes standardized beta.

^aModel 1 includes age and medical comorbidity as covariates.

^bModel 2 includes age, medical comorbidity and diagnosis as covariates.

Among the CTQ subscales, higher scores for emotional abuse and physical abuse were individually associated with higher inflammation (see Table 3). When controlling for diagnosis, only the association between the physical abuse subscale and inflammation remained significant (Table 3). In addition, higher levels of inflammation composite were associated with worse cognitive performance as measured by MCCB composite score, controlling for age, sex (per corrected scores), and medical comorbidity (beta=−.28, p=.002, CI=−.46, −.10, partial eta-squared (ηp2)=.073). Results remained significant even after additional adjustment for diagnosis (beta=−.19, p=.02, CI=−.36, −.03, partial eta-squared (ηp2)=.043). Correlations between each inflammatory marker (i.e., CRP, IL-6, TNF-α) and the total CTQ score, as well as its subscales, are presented in Supplemental Table 9.

3.5. Inflammation Mediation of the CM-Cognition Association in the Entire Sample (n=125)

Results of the mediation analysis in the entire sample of BD patients and healthy controls are shown in Figure 3. Age and sex were included as covariates to account for their potential influence, given the observed higher CM in women than men and significant association between higher age and elevated inflammation.

Figure 3.

Total, direct, and indirect effects of childhood maltreatment total score on global cognition in a sample of patients with BD and healthy controls. Covariates in the mediation model: age, sex.

Abbreviations: c_cs: standardized effect for c; c′_cs: standardized effect for c′; SE: standard error; BootCI: 95% bootstrap confidence interval; BootSE: a bootstrap estimate of the standard error on the indirect effect; CM, Childhood Maltreatment

We found that the indirect effect of total CTQ on global cognition (measured by the MCCB composite score) through inflammation was statistically different from zero, as revealed by a 95% confidence interval that was entirely below zero (ab: beta= −.048, CI= −.10, −.009), holding age and sex constant. Findings revealed a significant direct effect of total CTQ on global cognition independent of its effect on inflammation (c′: beta= −.40, p<.001, CI=−.57, −.24), meaning that an individual with a higher CT but an equal inflammation level was estimated to be .40 units lower in the global cognitive score. The total effect of total CTQ on global cognition (c: c′ + ab: beta= −.40 + −.048 ≈ −.45) was significant (p<.001, CI=−.61, −.29), meaning that two individuals who differ one unit in CM were estimated to differ by .45 units in their global cognitive scores. Together, these findings suggest that the negative effect of CM on global cognitive performance was partially mediated by inflammation (approximately 11%: ab/c=[(−.048)/(−.45)]*100).⁶⁰ The indirect effect was no longer statistically different from zero when diagnosis was added to the model as a covariate (ab: beta=−.02, CI=−.07, .01).

4. Discussion

In the present study, we showed that higher childhood maltreatment (CM) was associated with poorer global cognitive performance in patients with bipolar disorder (BD) and in healthy controls (HC), with a particularly strong effect observed in BD individuals. Furthermore, peripheral inflammation emerged as a partial mediator of the CM-global cognition link.

In line with previous studies, we observed a threefold higher prevalence of severe CM in the BD group versus HC (66% vs. 22%), emphasizing a significant disparity in CM exposure between groups.^61,62 More severe CM was associated with a higher number of prior suicide attempts, with a medium effect size, aligning with existing literature.^63,64 These results reinforce the importance of considering CM as a risk factor for BD, and a key factor in understanding its clinical course, particularly in relation to suicidality. Furthermore, our findings suggest that CM may impact cognitive reserve, as indicated by its association with lower premorbid IQ and educational attainment. This highlights the potential neurodevelopmental impact of early life stress and its role in shaping illness trajectory, emphasizing the need for early intervention strategies to mitigate long-term cognitive and clinical consequences.

Childhood Maltreatment and Cognitive Performance

Consistent with our hypotheses, higher CM (i.e., total CTQ) was associated with poorer global cognitive functioning with a medium effect size (ηp2=.066) across the entire sample, even after accounting for the influence of diagnostic status. This suggests that the impact of CM on cognitive functioning extends beyond the presence of a psychiatric diagnosis. The associations were evident across all subscales of abuse and neglect. Specific cognitive domains affected included speed of processing, working memory, visual learning, and reasoning and problem solving, with the largest effect size observed for working memory (ηp2=.065; medium level), followed by speed of processing (ηp2=.043; small-to-medium). These findings align with prior research indicating that early-life stress negatively impacts cognitive function in adulthood in both patient populations and healthy controls.^17,18,65,66

Within the BD group, higher total CM was similarly associated with worse global cognitive functioning, with a medium-to-large effect size (ηp2=.098). The associations were evident across all CM subscales, except for physical neglect. Within the BD group, significant associations were observed between total CM and MCCB domains of speed of processing, attention and vigilance, working memory, visual learning, reasoning and problem solving as well as cognitive flexibility (measured by WCST). Notably, medium to large effect sizes were found for working memory (ηp2=.11), speed of processing (ηp2=.076) and measures of cognitive flexibility (ηp2=.059 and .067). Surprisingly, a strong association was found between total CM and MCCB verbal learning in HC, but not in BD. Several illness-related clinical factors (e.g., earlier age of onset, antipsychotic use) have been associated with poorer verbal learning performance in BD,^6,67,68 which may obscure or attenuate the observed association between CM and verbal learning. Given that, in our sample, the effect size of the CM–verbal learning association in BD was smaller compared to other cognitive domains, a larger sample size may be necessary to detect a significant relationship. Within the HC group, only emotional neglect and physical neglect were associated with poorer global cognition, with approximately medium effect sizes (ηp2=.059 and .088, respectively). There were non-significant relationships between all other CM subscales and global cognition in the HC sample. This is not unexpected, given the low reported prevalence of physical, emotional, and sexual abuse in the HC sample, which may have limited statistical power to detect associations between abuse and cognition within the HC group. However, the low prevalence in the HC group may also reflect underreporting, possibly due to reasons such as greater self-serving attributional bias.⁵³ Consistent with prior studies⁵³, we found that the HC group had higher minimization/denial scores than the BD group. Importantly, some individuals in the HC group may have experienced CM but demonstrated resilience or positive adaptation in the face of adversity,⁶⁹ which could buffer against detectable cognitive effects.

Our findings regarding the CM-global cognition link in BD are consistent with the existing literature in mood disorders.^17,18 A recent meta-analysis of 20 studies in individuals with BD found that total CM was negatively associated with cognitive performance.¹⁹ A limited number of studies have examined cognition using a comprehensive neuropsychological battery, with a focus on global cognitive performance (n=2)^20,21 and/or general IQ (n=6).^20,22–26 Of these, six reported negative associations with global cognition,^20,21 or general IQ.^22,24–26 Additionally, the meta-analysis reported that physical abuse, sexual abuse and physical neglect were also associated with global cognition and general IQ. This meta-analysis found that among cognitive domains, total CM was associated with attention and speed of processing, and verbal learning and memory, but not with working memory, executive function (EF) or social cognition. Their finding regarding total CM-working memory association contrasts with our results, where we found the strongest association between total CM and working memory in both the full sample and within the BD group. This discrepancy could be due to the small number of existing studies that have examined the CM-working memory association,^20,22,70,71 and the small sample sizes and limited number of tasks used to assess working memory in some of these studies. One study in individuals with BD (n=345) and HC (n=183) found that total CM was negatively associated with working memory among cognitive domains after accounting for relevant demographic and clinical covariates.²⁰ Extending these findings, neuroimaging research has shown that structural changes to the prefrontal cortex mediate the association between early childhood stress and working memory performance.⁷² Our results also showed an association between CM and cognitive flexibility (as measured by WCST) within the BD sample. Prior work on the CM-EF association in BD has been less consistent, with some studies reporting an association with EF more broadly,^26,28,29 and some finding no CM association with other components of EF.^70,73 Given the global nature of the associations between CM and cognition (e.g., significant effects on global cognition) and strong intercorrelations among cognitive measures, the patterns of associations across specific domains may reflect differential psychometric sensitivity of the tests themselves rather than a truly distinct pattern across cognitive domains. Taken together, these findings highlight inconsistencies in the literature regarding the cognitive domains most affected by CM in BD. Given the limited number of well-powered studies using comprehensive cognitive assessments, further research is needed to clarify the specific cognitive effects of total CM—particularly on working memory, speed of processing, and executive functioning.

Regarding CM subscales, our results showed that physical abuse and sexual abuse were negatively correlated with all cognitive domains, except for social cognition, within the BD group. Notably, all CM subscales correlated with working memory, which broadly aligns with meta-analytic findings indicating that physical and emotional abuse, sexual abuse and physical neglect were associated with poorer working memory performance. In contrast to our findings—where all CM subscales were correlated with speed of processing (except for a marginal trend for physical neglect) in BD—the meta-analysis did not report significant associations between CM subscales and attention and processing speed.¹⁹ In line with the meta-analytic results,¹⁹ we found negative correlations between executive function (as measured by WCST) and physical abuse, sexual abuse and physical neglect. Overall, given the limited number of well-powered studies in these domains, further research is warranted to determine the robustness and specificity of CM-related cognitive impairments in BD.

The association between CM and cognition is not surprising, as exposure to CM can significantly impact brain development. Evidence links CM to structural and functional changes in brain regions, including the hippocampus, prefrontal cortex (PFC), and amygdala—areas that are critical for various cognitive processes that may contribute to emotion dysregulation.^74,75 Lower total white matter volume was associated with poorer global cognition and cognitive subdomain performance across BD, MDD and HC participants.⁷⁶ Notably, higher CM was linked to lower total white matter volume, independent of diagnostic status. Similarly, CM was associated with reduced gray matter volume in the frontal lobe in BD and HC participants.⁷⁷ These neurobiological findings underscore the importance of further longitudinal studies to clarify the mechanisms linking CM to brain structure alterations and consequent cognitive impairments.

Inflammation as a Mediator

Immune dysregulation and epigenetic modifications have been identified as potential mechanistic pathways linking adverse childhood experiences to health outcomes across the lifespan.³¹ Research shows that CM is associated with increased peripheral inflammation in both population-based samples and individuals with mood disorders.^32,78,79 Although few studies have explored the relationship between CM and inflammation in BD, some evidence suggests a link. Moraes et al. (2017) reported an association between elevated CRP and a history of CM in BD patients⁸⁰ and Aas et al. (2017) reported an association between increased CRP and greater severity of childhood abuse in a combined sample of patients with BD and schizophrenia.⁸¹ We found higher levels of inflammation in BD patients than controls, and greater CM was associated with elevated inflammation in the entire sample of BD patients and controls, with the largest effect sizes observed for subscales that assess abuse rather than neglect. Of note, the association between total CM and inflammation was no longer significant after adjusting for diagnosis, suggesting that other factors common in people with BD contribute to heightened inflammation (e.g., manic episodes, disrupted sleep), attenuating the unique effect of CM on inflammation.

Increased peripheral levels of inflammatory markers have been observed in patients with BD during depression, mania and euthymia.^34,35 Inflammation has also been associated with poorer cognitive performance and cognitive decline in both population-based samples and BD patients.^38,82 The current body of literature suggests a role for C-Reactive protein (CRP), IL-6, and Tumor Necrosis Factor (TNF)-α in contributing to the development and/or progression of cognitive impairment in BD.³⁷ Thus, immune-inflammatory pathways are potential mechanisms linking early life stress to poorer cognitive function, but very few studies have investigated this in BD. Congio et al. (2022) found BD patients with high inflammation (CRP≥5 mg/L) had a history of CM and worse cognitive functioning.⁸³ While prior research has examined associations between CM, inflammation, and cognitive functioning in BD, the mediating role of inflammation in the relationship between CM and global cognitive functioning using a comprehensive cognitive battery in BD remains unclear and was explored in this study, yielding preliminary findings. We showed that inflammation partially mediates the association between CM and global cognitive performance in a sample of BD patients and healthy controls, accounting for approximately 11% of the effect that CM has on global cognitive performance. This is consistent with a neuroimaging study by Poletti et al. (2022) showing that the effect of adverse childhood experiences on white matter microstructure was mediated by inflammation in BD,⁸⁴ and it aligns with the growing body of literature suggesting that while inflammation plays a role in BD-related cognitive impairments, its impact is complex and likely intertwined with other neurobiological and environmental factors.

There are several mechanisms that may explain the association between CM, inflammation and cognitive impairment in mood disorders. One key mechanism is dysregulation of the hypothalamic-pituitary-adrenal (HPA) axis,⁸⁵ where chronic stress leads to prolonged HPA axis activation and sustained cortisol elevation. Persistently high glucocorticoid levels can exert detrimental effects on hippocampal neurons, reduce neurogenesis, and impair prefrontal cortex function, all of which contribute to cognitive impairment. Additionally, excessive cortisol disrupts immune system homeostasis, triggering a pro-inflammatory state. This convergence of HPA axis dysregulation and inflammation may create a feedforward loop that increases the vulnerability to cognitive decline in BD. Meta-analytic data provide consistent evidence of hyperactivity of the HPA axis, as shown by elevated cortisol (both basal and post-dexamethasone), adrenocorticotropic hormone (ACTH) levels, and an increased response to the dexamethasone/corticotropin-releasing hormone test in BD, an effect that is more pronounced in mania but still noted during euthymic periods.⁸⁶ Aas et al. (2011) reported an association between a blunted cortisol awakening response and poor performance on speed of processing and verbal memory in patients with first-episode psychosis,⁸⁷ further supporting the link between HPA axis dysregulation and cognitive performance in mood disorders.

Strengths and Limitations

Our study has several strengths. We included BD patients who were mostly affectively stable (64% euthymic)—a population that is often understudied. Additionally, the comprehensive assessment of cognitive function in our study allowed us to determine effect sizes for the associations between CM and global cognition, as well as more specific cognitive domains. Finally, we measured multiple inflammatory markers, allowing us to examine both acute phase protein (CRP) and cytokines (IL-6 and TNF-α) that are relevant to both early life stress and cognitive function.

Several limitations should be acknowledged. First, the cross-sectional design precludes causal inferences. Second, while we accounted for multiple covariates, unmeasured confounding factors such as stressful life events during adulthood may have influenced our results. Third, the limited sample size for the inflammation analyses constrained our ability to test a more complex serial mediator model, with both BMI and inflammation as potential mediators of the link between CM and cognitive functioning – which we would speculate as likely. Fourth, mood ratings were collected up to two weeks prior to cognitive testing and inflammatory marker assessment, which may introduce minor temporal variability in interpreting associations across these measures. Fifth, the CTQ does not capture important contextual factors such as the timing and duration of exposure—factors that appear to influence cognitive outcomes.^16,88,89 This constraint limits the interpretation of our findings and underscores the need for future studies to employ more detailed, longitudinal assessments that consider both the timing and chronicity of maltreatment.

Clinical Implications

Our findings underscore the importance of routinely assessing and addressing CM in individuals with BD and those at higher risk for BD, as it appears to contribute to cognitive impairment. Routine CM screening into standard psychiatric settings may enable more personalized care. These results support models positing that early life stress contributes to neurodevelopmental alterations that persist into adulthood, particularly in individuals with psychiatric disorders. The differential impact of CM on cognitive function in BD versus HC suggests that individuals with BD may have heightened susceptibility to the long-term consequences of early life stress, possibly due to underlying genetic or neurobiological vulnerabilities. Early interventions targeting at-risk groups could reduce this long-term burden. Given the impact of cognitive impairment on functional outcomes, interventions incorporating a trauma focus may be beneficial for improving quality of life in BD. Cognitive remediation therapy, mindfulness-based interventions, and anti-inflammatory strategies may also hold promise in reducing the negative effects of CM on cognitive function. Finally, preventative strategies that focus on enhancing cognitive reserve or mitigating cognitive decline may offer significant advantages in prospectively sparing patients from less-than-optimal trajectories. Future longitudinal research is needed to clarify causal pathways and identify optimal timing and targets for intervention.

5. Conclusion

This study provides evidence that (1) total CM and its subscales were associated with poorer global cognitive performance in a sample of individuals with BD and HC, (2) the CM-cognition association was stronger within the BD group, (3) total CM showed the strongest association with working memory, followed by speed of processing, in both the full sample and the BD subgroup, and (4) inflammation partially mediated the relationship between CM and global cognition in the full sample. Our findings suggest that while inflammation contributes to cognitive impairments associated with CM, its effects are complex and likely interact with other biological and environmental factors. These findings highlight the need for targeted interventions aimed at enhancing cognitive outcomes and slowing cognitive decline in individuals with BD who have a history of early life maltreatment.

Data Availability Statement

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