Depression, cardiometabolic disease, and their co-occurrence after childhood maltreatment: an individual participant data meta-analysis including over 200,000 participants

This IPD meta-analysis is an effort of the EarlyCause consortium [19], which investigates the association between early-life stress and comorbid depression and cardiometabolic outcomes. Studies within and outside the consortium were selected based on the consortium network. Study inclusion criteria were as follows: having retrospective reports on childhood maltreatment (at least physical and emotional abuse) before the age of 18 and having data on depression and/or cardiometabolic diseases in adulthood. Studies were excluded if participants were younger than 18 years at assessment time. In total, 13 studies from Germany, the Netherlands, the UK, and the USA were included.

Three studies were case–control studies with an overrepresentation of individuals with affective disorders: the Marburg-Münster Affective Disorders Cohort Study (MACS) [20], the Netherlands Study of Depression and Anxiety (NESDA) [21], and the Netherlands Study of Depression in Older persons (NESDO) [22]. Ten studies were population-based cohort studies: the mothers and partners of the Avon Longitudinal Study of Parents and Children (ALSPAC, see Additional file 1: Sect. 1) [23‐25], the mothers of Generation R Study (GenR, see Additional file 1: Sect. 2) [26], the Healthy Life in an Urban Setting (HELIUS) study [27], the first wave of the Midlife in the United States (MIDUS) study [28], the first and second Netherlands Mental Health Survey and Incidence Studies (NEMESIS-1 and NEMESIS-2) [29, 30], two independent cohorts of the Study of Health In Pomerania (SHIP-Trend baseline assessment and SHIP-Legend) [31], and the UK Biobank (UKBB) [32, 33]. Each cohort study was approved by local ethics committees and all participants provided informed consent. This research project was pre-registered on the international prospective register of systematic reviews PROSPERO in March 2021 (reference CRD42021239288). The PRISMA-IPD guidelines [34] were followed except for the systematic review-related steps which were not applicable in this research.

A two-step IPD meta-analysis was carried out [54]. In the first step, cohorts applied a standardized protocol for data harmonization to create the required variables and carry out statistical analyses estimating the associations between childhood maltreatment and the different outcomes. In the second step, we meta-analyzed cohorts’ aggregate effect sizes with random-effects models using inverse-variance weighting. We chose random-effects models to pool the aggregate effect sizes because these effect sizes are drawn from different populations. Cohorts with cell count < 5 across exposure and outcome categories were excluded from the meta-analyses. Heterogeneity parameters Q, I², and \({\tau }^{2}\) were calculated. Scripts of the two steps can be found on the Github EarlyCause repository (see Additional File 1: Sect. 3).

The odds of having depression increased almost three folds in those with a history of childhood maltreatment compared to those without (model 1, OR [95% CI] = 2.82 [2.40; 3.30], Fig. 1A). This positive association was seen in all cohorts, albeit with significant heterogeneity across studies (Q = 70.44, p < .001, I² = 91.8%, \({\tau }^{2}\) = 0.07). Adults with a history of childhood maltreatment, as compared to those without, were also more likely to have a cardiometabolic disease (model 2, OR [95% CI] = 1.34 [1.23; 1.46], Fig. 1B). Effect sizes were relatively homogeneous across studies (Q = 18.09, p = 0.113, I² = 27.1%, \({\tau }^{2}\) = 0.01). The results of all meta-analyzed models are shown in Table 2.

The analysis of the association of childhood maltreatment with comorbidity status (model 3) was restricted to nine cohorts (n = 146,167) due to crosstab cells with less than five cases for some outcomes in four cohorts. In these nine cohorts, adults with a history of childhood maltreatment had twice higher odds of depression only (OR [95% CI] = 2.68 [2.39; 3.00], Fig. 2) and also higher odds of cardiometabolic disease only (OR [95% CI] = 1.27 [1.18; 1.37], Fig. 2). The effect size for cardiometabolic disease was around half of the effect size for depression only. The strongest association was found for comorbid depression and cardiometabolic disease (OR [95% CI] = 3.04 [2.51; 3.68]), Fig. 2), and this positive association was statistically significant in most cohorts. Since the UKBB represented the largest dataset in the pooled analysis (n = 98,619, 67.5% of participants, see Additional File 1: Table S7 for weights in pooled estimate), we re-ran the meta-analysis excluding the UKBB and observed that results remained largely similar (OR [95% CI] depression only = 2.67 [2.33; 3.06], OR [95% CI] cardiometabolic disease only = 1.29 [1.15; 1.44], OR [95% CI] comorbidity = 2.82 [2.43; 3.27]). These results were mostly consistent with results of the UKBB only (OR [95% CI] depression only = 2.66 [2.54; 2.78], OR [95% CI] cardiometabolic disease only = 1.26 [1.13; 1.40], OR [95% CI] comorbidity = 4.11 [3.73; 4.53]), although the odds of comorbidity seemed to be slightly higher in the UKBB than in the other cohorts. Subgroup analyses showed that results were largely unaffected by depression assessment type (see Additional File 1: Table S8) and by current vs. lifetime depression (see Additional File 1: Sect. 5). Additionally, we exploratively ran a post hoc test to evaluate whether childhood maltreatment was more strongly associated with comorbidity than with the individual diseases. Since the UKBB was the largest sample for which we had direct access to the individual-level data, we used that sample to calculate the following two odds ratios after childhood maltreatment: depression only vs. comorbidity and cardiometabolic disease only vs. comorbidity. Instead of using the outcome level “absence of depression and cardiometabolic disease” as reference category as in the previous calculations of odds ratios, we used the outcome level “comorbidity” as new reference category to statistically test whether childhood maltreatment was more strongly associated with comorbidity than with the single diseases. We found that the association of childhood maltreatment with depression only (OR [95% CI] = 0.65 [0.59; 0.71]) and cardiometabolic disease only (OR [95% CI] = 0.31 [0.27; 0.35]) were significantly smaller than with comorbidity.

Additional adjustment for smoking status, alcohol consumption, and physical activity (model 4) did not substantially change the associations of childhood maltreatment with depression only (OR [95% CI] = 2.57 [2.28; 2.89]), cardiometabolic disease only (OR [95% CI] = 1.25 [1.15; 1.36]), and comorbidity (OR [95% CI] = 2.99 [2.46; 3.63]), highlighting the independence of these associations from lifestyle factors. Similar associations to those obtained in the main model 3 were observed for males (model 5a) and females (model 5b). However, the association with cardiometabolic disease only seemed stronger in females than in males (OR [95% CI] in females = 1.41 [1.26; 1.57], in males = 1.13 [1.01; 1.27], with OR difference z = 2.77, p = .006; Table 2, models 5a and 5b). All types of childhood maltreatment (physical, emotional, and sexual abuse) were associated with increased odds of developing depression only, cardiometabolic disease only, and comorbidity, although physical and emotional abuse were particularly strong predictors of the comorbidity (Table 2, model 6). Finally, although all numbers of maltreatment types were related to comorbidity status, there seemed to be a dose–response relationship: When two or more types of childhood maltreatment were experienced, the odds of (comorbid) depression and cardiometabolic disease exceeded the odds found after one type of maltreatment only (Table 2, model 7).

Results were similar in sensitivity analyses applied to comorbidity status model 3 when using different operational definitions of cardiovascular disease based on stricter or broader definition (models 8 and 9; Table 2) or different definitions of cardiometabolic diseases and depression additionally including information on medication (model 10; Table 2).

This study used data from 13 international cohorts involving 217,929 persons to systematically investigate the association of childhood maltreatment with (comorbid) depression and cardiometabolic disease in adulthood. In order to obtain a consistent set of aggregate data across cohorts, individual participant data were harmonized and cohort-level analyses were standardized. Main findings show that adults with a history of childhood maltreatment, as compared to those without, are 1.27 times more likely to have cardiometabolic disease only and 2.68 times more likely to have depression only. The largest difference between maltreated and non-maltreated individuals was found for the co-occurrence of both conditions: Maltreated individuals were 3.04 times more likely to suffer from comorbid depression and cardiometabolic disease in adulthood. Post hoc analyses showed that this association was larger than the ones for either disease alone. Results remain similar in sensitivity analyses using different outcome ascertainment definitions, suggesting findings are robust.

Our results are in line with findings from existing meta-analyses. Two relatively recent meta-analyses [2, 3] evaluated the association of childhood maltreatment history with depression and found that childhood maltreatment was associated with 2.03 (95% CI = [1.37; 3.01]) and 2.81 (95% CI = [2.35; 3.36]) increased odds of depression in adulthood, using pooled samples of 4579 [2] and 26,536 [3] participants, respectively. The current study found similar heightened odds of depression (OR [95% CI] = 2.82 [2.40; 3.30]) after childhood maltreatment, based on by far the largest sample size (n = 155,869). Additionally, although age was demonstrated to moderate this association [3], pooled effect sizes from the previous meta-analyses were either based on study-level effect sizes with inconsistent handling of covariates or on raw data excluding covariate adjustments. In contrast, the current research facilitated cohort-level analyses in a systematic manner, enabling the estimation of pooled effect sizes adjusting for important sociodemographic and lifestyle covariates. Although the effect size reported in Li et al. [2] is slightly smaller than the one in the current study, the difference may be explained by the definition of Li et al.’s exposure variable: Childhood maltreatment was based on official records, which are more likely to suffer from underreporting than retrospective self-reports. Despite this difference in assessment, the consistent direction of findings increase confidence in the validity of maltreatment self-reports. Further, a previous meta-analysis [57] of 29 studies (N = 247,393) showed that cumulative childhood adversity was moderately related to cardiometabolic disease in adulthood (OR [95% CI] = 1.36 [1.27; 1.46]). Although the exposure (an index including at least two adverse childhood experiences) and outcome (cardiometabolic disease including metabolic syndrome) definitions slightly differ from the ones of the current meta-analysis, results align closely (our findings: OR [95% CI] = 1.34 [1.23; 1.46]). Finally, because the current meta-analysis consistently adjusted associations for the same covariates, it provides the unique possibility to directly compare the increased odds of each disease after maltreatment. The results show that, compared to non-maltreated individuals, maltreated adults are 2.68 times more likely to suffer from depression and “only” 1.27 times more from cardiometabolic disease. Although linked to both, childhood maltreatment is therefore more strongly related to depression than to cardiometabolic disease in adulthood.

A striking result is that the odds of comorbid depression and cardiometabolic disease after childhood maltreatment (OR [95% CI] = 3.04 [2.51; 3.68]) are higher than for each disease alone. Although previous studies report that depression and cardiometabolic disease tend to co-occur, the current meta-analysis is the first study to investigate and support the relationship between childhood maltreatment and the co-occurrence of depression and cardiometabolic disease in adulthood. This association is possibly explained by the early-life stress triggering mechanisms common to both depression and cardiometabolic disease. Previous research suggests that childhood maltreatment activates interrelated biological and behavioral pathways [17] potentially leading to adverse health outcomes. Because childhood maltreatment occurs during a critical period for brain neuroplasticity, it may dysregulate stress-related neural circuits [58, 59]. Longitudinal studies show that childhood maltreatment is associated with structural and functional neural changes [60]. Among others, these changes may subsequently dysregulate neuroendocrine and immune systems. The HPA axis may be hyper- or hypo-activated due to impaired glucocorticoid receptor function and inflammation levels may be elevated [1, 61]. Although behavioral pathways are also hypothesized to contribute to poor health outcomes in people with childhood maltreatment [17], our results show that the associations of childhood maltreatment with comorbidity status survive adjustment for smoking, alcohol consumption and physical activity; suggesting that the increased likelihood of (comorbid) depression and cardiometabolic disease after maltreatment does not exclusively depend on one’s lifestyle. Additionally, other non-biological factors (i.e., disease severity, age at diagnosis, working conditions) may also explain the strong association observed between childhood maltreatment and comorbidity and should be investigated. Lastly, the higher odds of comorbidity than single diseases after childhood maltreatment may be explained by the fact that depression and cardiometabolic disease have a bidirectional feedforward loop [10]. Both diseases likely magnify each other in a reinforcing vicious cycle, which is further stimulated by childhood maltreatment and its related biological, psychosocial and behavioral consequences.

We carried out additional analyses to explore how the associations between childhood maltreatment and (comorbid) depression and cardiometabolic disease varied across sex and maltreatment types. Both in males and females, childhood maltreatment was associated with more (comorbid) depression and cardiometabolic disease. Associations between childhood maltreatment and comorbidity status were mostly similar across males and females. However, females showed a slightly stronger association than males for cardiometabolic disease only (males: OR [95% CI] = 1.13 [1.01; 1.27], females: OR [95% CI] = 1.41 [1.26; 1.57]). Evidence from the literature on that matter is inconsistent [9, 57], and conclusions should therefore be drawn carefully. Further analyses were carried out to test the relationship between maltreatment types and (comorbid) depression and cardiometabolic disease. Because multi-type maltreatment is common [62], the different maltreatment types were entered as multiple predictors within the same model to obtain average estimates of the association between each maltreatment type with comorbidity status while accounting for the co-occurring experience of other types of maltreatment. Findings show that all maltreatment types were independently associated with (comorbid) depression and cardiometabolic disease. Zooming in on specific outcomes, depression only was particularly strongly associated with emotional abuse. Cardiometabolic disease only and comorbidity were particularly strongly associated with physical and emotional abuse. Previous research findings support our results: Physical and emotional abuse are stronger predictors of depression and cardiovascular disease than sexual abuse [3, 9]. Alternatively, the estimated associations of sexual abuse with the disease outcomes may be harder to detect because of the relatively low prevalence of sexual abuse compared to the other types of abuse [63] or because milder forms of sexual abuse are picked up, for instance from the population-based studies. Lastly, findings endorse a dose–response relationship of childhood maltreatment severity—here operationalized as the number of maltreatment types—with (comorbid) depression and cardiometabolic disease. This converges with previous evidence on various health outcomes [64, 65].

This study has several strengths. First, the meta-analysis gathered 13 international cohorts including 217,929 individuals from European countries and the USA. Second, the systematic methodology used with the two-step individual participant data design has essential advantages [54]. It enables the standardization of analyses across studies (i.e., harmonization of variables and consistent covariate adjustment of estimates) and increases the quality of aggregate data entering the meta-analysis. Third, the variety of cohorts involved (e.g., case–control and population-based studies; cohort oversampling persons with migration background) and comprehensiveness of the analyses carried out (e.g., sensitivity analyses with different outcome definitions, effects of different maltreatment types, stratified analyses across sex) strengthens the robustness of the findings across settings.

This study also has limitations. In some cohorts, especially those with younger samples, the prevalence of comorbid depression and cardiometabolic disease was low (weighted mean 2.1%), leading to some studies being excluded of the multinomial regression analyses due to small cell count. This is likely because cardiometabolic events usually happen in later life [66]. The prevalence of comorbidity increases with age as seen in the oldest cohort NESDO with the highest rate of comorbidity (22.6%). The relatively high average age across cohorts (52.4 years old) may have thereby facilitated finding existing associations. Despite the difference in outcome prevalence across cohorts of different ages, the associations found for depression and cardiometabolic disease are consistent across younger (e.g., ALSPAC mothers and partners) and older (e.g., NESDO and SHIP-Legend) cohorts. Therefore, there is no obvious indication of a differential effect of age. An alternative explanation for the low prevalence of comorbidity may be survival bias where patients with severe depression and cardiometabolic disease have died or are too ill to participate in the studies. Nevertheless, even after excluding studies with too few comorbidity cases from the multinomial regression analyses, the total sample used to investigate the association with comorbidity status still amounted to 146,167 individuals. Another limitation is that meta-analyzed associations of childhood maltreatment with depression and comorbidity showed non-negligible heterogeneity. However, we used random-effect models which, by definition, assume the included studies have different true effect sizes, and thereby account for heterogeneity in calculating pooled estimates. The heterogeneity could not be explained by differences in depression assessment but other factors could have possibly caused this divergence (e.g., study design, age, cultural differences in stigma reporting childhood maltreatment) and should be further investigated. Additionally, as with every assessment type, the reliance on self-reports has its set of limitations. Cardiometabolic diseases were assessed with self-reported diagnoses, which may be perceived as biased. However, previous research show that cardiometabolic disease assessment (self-reports vs. medical records) does not influence the association found between childhood maltreatment and cardiometabolic disease [9]. Childhood maltreatment was also assessed with self-reports. It has been suggested that depression may negatively bias someone’s recall of their childhood experiences [67] in which case, self-reports may spuriously inflate the association found between childhood maltreatment and depression as well as comorbidity. Recent evidence from published and unpublished research [67, 68] highlights the marginal susceptibility of maltreatment self-reports to negative recall bias as well as their temporal stability irrespective of depression diagnosis. In order to test this in the current study, we compared the associations found in population-based cohorts using lifetime vs. current depression assessments (see Additional file 1: Sect. 5) and found no evidence of negative recall bias. In addition, analyses were exclusively carried out on individuals with available data on childhood maltreatment. This may have introduced some bias as maltreatment non-response might be associated with the disease outcomes [8]. Moreover, the current study’s assessment of maltreatment was limited to experiences of abuse because neglect was assessed so differently across cohorts that we could not harmonize. Yet, childhood neglect is an important early-life stressor potentially affecting depression and cardiometabolic outcomes in adulthood and should be investigated in future studies. Furthermore, although the current study focusses on the comorbidity of depression with cardiovascular disease and diabetes, other comorbid psychiatric and somatic diseases may also be activated by early-life stress pathways and warrant further investigation. Another limitation concerns the fact that the current study did not test the role of maltreatment timing. Although a recent meta-analysis shows no evidence of consistent sensitive periods of childhood maltreatment linked to various health outcomes [69], future studies with detailed timing information are needed to determine with more certainty whether timing of childhood maltreatment exposure matters for (comorbid) depression and cardiometabolic disease. Finally, a last limitation concerns the unknown timeline of events. Although depression and cardiometabolic disease likely have their onset after—and we believe are caused by mechanisms stemming from—childhood maltreatment, the current study only articulates associations and does not inform about causality.

The current findings have clinical implications. First, our results raise awareness on the association between early-life stress and distal psychiatric and somatic health. Second, this study may be a first step in the process of preserving the health of individuals with a history of childhood maltreatment. If future evidence supports that childhood maltreatment triggers a cascade of mechanisms leading to (comorbid) depression and cardiometabolic disease, early intervention could prevent the dysregulation of biological stress systems and preserve the health of individuals with a history of childhood maltreatment. For instance, standard psychotherapy has been shown to effectively reduce depression severity in individuals with a history of childhood maltreatment [70]. One could therefore consider providing trauma-focused psychotherapy to help victims of maltreatment process the stress evoked by the trauma, or pharmacotherapy aiming to regulate biological stress systems, subsequently promoting somatic and mental health. In addition to individual interventions, societal action is an opportunity to prevent these comorbid diseases. Recent influential work emphasizes that promoting fair distribution of income, protecting work conditions, fostering gender equity, decreasing discrimination, and improving social cohesion/support have a great potential to prevent early-life stress, and in turn (comorbid) depression and cardiometabolic disease [71, 72].