A genetic exploration of the relationship between Posttraumatic Stress Disorder and cardiovascular diseases

PTSD, Cardiovascular disease, Genetic correlation, Linkage disequilibrium score regression, Genomic structural equation modeling, Mendelian randomization

Background and Aims Experiencing a traumatic event may lead to Posttraumatic Stress Disorder (PTSD), including symptoms such as flashbacks and hyperarousal. Individuals suffering from PTSD are at increased risk of cardiovascu- lar disease (CVD), but it is unclear why. This study assesses shared genetic liability and potential causal pathways between PTSD and CVD.

We leveraged summary-level data of genome-wide association studies (PTSD: N= 1,222,882; atrial fibril- lation (AF): N=482,409; coronary artery disease (CAD): N=1,165,690; hypertension: N=458,554; heart failure (HF): N=977,323). First, we estimated genetic correlations and utilized genomic structural equation modeling to identify a common genetic factor for PTSD and CVD. Next, we assessed biological, behavioural, and psychosocial factors as potential mediators. Finally, we employed multivariable Mendelian randomiza- tion to examine causal pathways between PTSD and CVD, incorporating the same potential mediators.

Significant genetic correlations were found between PTSD and CAD, HT, and HF ( rg =0.21-0.32, p≤ 3.08 · 10−16), but not between PTSD and AF. Insomnia, smoking, alcohol dependence, waist-to-hip ratio, and inflammation (IL6, C-reactive protein) partly mediated these associations. Mendelian randomization indicated that PTSD causally increases CAD (IVW OR=1.53, 95% CIs=1.19-1.96, p=0.001), HF (OR=1.44, CIs=1.08-1.92, p=0.012), and to a lesser degree hypertension (OR=1.25, CIs=1.05-1.49, p=0.012). While insomnia, smoking, alcohol, and inflammation were important mediators, independent causal effects also remained.

In addition to shared genetic liability between PTSD and CVD, we present strong evidence for causal effects of PTSD on CVD. Crucially, we implicate specific lifestyle and biological mediators (insomnia, substance 2 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted March 22, 2024. ; https://doi.org/10.1101/2024.03.20.24304533doi: medRxiv preprint use, inflammation) which has important implications for interventions to prevent CVD in PTSD patients. 3 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted March 22, 2024. ; https://doi.org/10.1101/2024.03.20.24304533doi: medRxiv preprint Translational perspective The significant mental and physical strain experienced by patients suffering from Post- traumatic Stress Disorder (PTSD) remains a domain necessitating further insight for the development of effective intervention strategies. Our study elucidates the complex genetic architecture that underlies the relationship between PTSD and cardiovascular disease. We present evidence supporting a causal link from PTSD to coronary artery disease and heart failure. Further, we identify various mediators of this causality, including inflammatory markers, substance use, waist-to-hip ratio and sleep deprivation. Our work calls for tar- geted preventive and therapeutic approaches to reduce the dual burden of mental and physical disease in PTSD patients. 4 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted March 22, 2024. ; https://doi.org/10.1101/2024.03.20.24304533doi: medRxiv preprint

Posttraumatic stress disorder (PTSD) is a disruptive mental illness triggered by trau- matic experiences and manifests as recurrent, adverse mental states related to the event. Symptoms may include flashback nightmares, dissociation and/or physiological reactions to external or internal event-related triggers. Traumas that may cause PTSD include acts or threats of (sexual) violence and witnessing the man- or disaster- caused passing of others [1]. While PTSD can only develop if a traumatic event has been experienced, the risk of developing it is partly determined by genetic liability. Twin-based studies estimated PTSD heritability between 40-60% and genome-wide association studies (GWAS) demonstrated that PTSD is characterized by high polygenicity, signifying that the risk of developing the disorder involves the influence of numerous genes [2, 3]. On top of debilitating psychological symptoms, PTSD has been associated with an increased risk of cardiovascular disease (CVD), as reported in prior observational research [4, 5]. Multiple studies and reviews have underscored an increased prevalence of diverse cardiovascular conditions among individuals with PTSD, including ischemic heart disease, heart failure, and atrial fibrillation [6, 7]. Why PTSD is associated with CVD is not fully understood. One possibility is that the development of PTSD and the development of CVD are (partly) due to the same genetic risk factors [8]. It could also be the case that there are causal associations, with PTSD acting as a risk factor for CVD, potentially mediated by biological (e.g. hypothalamic-pituitary-adrenal dysregulation, lower cortisol levels in the period following trauma, higher heart rates, elevated blood pressure, increased inflam- mation), behavioral (e.g. substance abuse, obesity, sleep disturbance) and/or psychosocial (e.g. social isolation) factors (extensively reviewed in [9] and [10]). Conversely, it may also 5 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted March 22, 2024. ; https://doi.org/10.1101/2024.03.20.24304533doi: medRxiv preprint be possible that a diagnosis or experience of life-threatening CVD could provoke PTSD [4]. It should be noted that while the DSM-4 considered life-threatening illnesses as PTSD triggers, the DSM-5 has limited these to unexpected medical catastrophes, like waking during surgery [1, 11]. We used advanced genetic methods to investigate the nature of the association between PTSD and CVD, focusing on shared genetics and causal explanations. In this study, we had two primary objectives: Firstly, we explored the shared genetics between the two conditions. Secondly, we established the directionality of a potential causal link and subse- quently elucidated underlying mechanisms, mediating causal relationships. To investigate this, we leveraged data from existing large-scale GWAS. These studies assess the associa- tion of millions of single-base DNA changes called single-nucleotide polymorphisms (SNPs) with a certain trait [12]. By integrating GWAS data from multiple traits, one can unveil genetic associations and potentially establish causality between different phenotypes [13].

The analysis plan was preregistered (for updates and alterations of the preregistered datasets please see Supplementary notes). We exploited summary-level data of large GWASs of European ancestry and applied advanced genetic methods to unravel the com- plex relationship between PTSD and CVD. PTSD genetic information was derived from the GWAS meta-analyses from the Psychiatric Genomics Consortium [3]. For the CVD traits, we selected atrial fibrillation (AF) [14], coronary artery disease (CAD) [15], hypertension (HT) [16] and heart failure (HF) [17] (Table 1). 6 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted March 22, 2024. ; https://doi.org/10.1101/2024.03.20.24304533doi: medRxiv preprint Genetic correlation and latent factor analysis We applied Linkage-Disequilibrium Score Regression (LDSC) [18] in a genomic structural equation modeling (Genomic SEM) framework [19] to assess genetic correlation and com- mon variance between PTSD and CVD. In essence, LDSC quantifies the extent to which genetic variants are linked to phenotypic traits. LD scores represent the correlation be- tween a genetic variant and nearby variants. LDSC calculates shared genetic influences between traits estimating genetic correlation to be negative (opposing effects), 0 (no over- lap) or positive (overlap). The slope in LDSC regression corrects for sample overlap that can inflate genetic correlations, improving the accuracy of genetic correlation estimates [19]. Genomic SEM relies on summary-level GWAS data as its primary input along with ancestry-matched LD scores (we used HapMap3 SNPs, [19]). Through the application of LDSC, Genomic SEM identifies genetic covariance patterns within and between all traits. Building on the idea of structural equation modeling, which allows us to systematically an- alyze relationships among (measured) variables, Genomic SEM translates genetic variance of various traits into a common genetic factor (as depicted in Figure 1). The contributions of the included traits are reflected by their factor loadings. The residual variance captures the genetic signatures unique to specific traits, representing genetic influences that are not shared across all traits. For a detailed explanation of Genomic SEM please see [19]. We extended factor models to include various mediators, categorized as biological (serotonin (SER) [20], cortisol (COR) [21], growth hormone receptor (GHR) [22], C-reactive protein (CRP) [23], interleukin (IL)-6 [24], IL-8 [ 25], Tumor necrosis factor α (TNFA) [26], BDNF [27]), lifestyle and behavioural (alcoholic beverages per week (AI) [28], alcohol dependence (AD) [29], smoking initiation (SI) [28], BMI [30], waist-to-hip ratio (WTH) [31], insomnia (IS) [32]) and psychosocial factors (loneliness (LN) [33], educational attainment (EA) [34]), 7 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted March 22, 2024. ; https://doi.org/10.1101/2024.03.20.24304533doi: medRxiv preprint to characterize underlying correlations (Table 1). The choice of mediators investigated in the current study rests upon prior research e.g. the possibility of physiological changes fol- lowing insomnia [35], immunosuppressive consequences of stress, as extensively researched in literature (e.g. [36] and [37]), and upon data availability. In a follow-up analysis, we aimed to elaborate on direct effects of PTSD on cardiovas- cular traits within our Genomic SEM framework. To do so, we used the common factor of CVD traits and PTSD and included genome-wide significant (p < 5 · 10−8), independent (r2 = 0.01) SNPs of the PTSD GWAS through the PTSD indicator itself on the separate cardiovascular traits (as introduced by [19]). This allowed to estimate the effect of SNPs that are robustly correlated to PTSD, later used as instrument variables for MR, on in- dividual cardiovascular traits. This provides some evidence on the potential causality of this relationship [19]. To address the potential influence of horizontal pleiotropy, meaning a direct effect of the SNP in question independent of the proposed exposure PTSD, we included direct loadings of the specific SNP on all three cardiovascular traits for each SNP one by one. For final model estimation, we retained all SNP effects that showed significance (at nominal level p <0.05) and removed any that had lost significance in the full model. The resulting final model thus incorporated significant paths from individual models that retained their significance even when considering all SNPs with associations acting directly on cardiovascular traits rather than through PTSD only. Causal associations between PTSD and CVD Finally, we applied two-sample MR in R (R version 4.2.2 (2022-10-31), ‘TwoSampleMR’ version 0.5.7, [39]), aiming to investigate causality underlying the relationship between 8 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted March 22, 2024. ; https://doi.org/10.1101/2024.03.20.24304533doi: medRxiv preprint Phenot ype Measure Study Sample Size (Thousands) SNPs PTSD clinician administered, self-report, DSM-4 Posttraumatic Stress Disorder Checklist, electronic health records Nievergelt et al., (2023) [3] 1,222.9 (1,085.7) 73 Atrial Fibrillation Paroxysmal or permanent AF, atrial flutter, Roselli et al., (2018)[14] 537.4 (482.3) 84 Coronary Artery Disease See original Supplementary Table 32 Aragam et al., (2022) [15] 1,165.7 (984.2) 241 Hypertension UKB Zhu et al., (2019) [16] 458.6 (313.8) 22 Heart Failure Clinical diagnosis, any etiology Shah et al., (2020) [17] 977.3 (930.0) 12 Biological factors Serotonin Plasma levels Hysi et al., (2022) [20] 8.8 Cortisol Morning plasma cortisol levels Crawford et al., (2021) [21] 25.3 4 Growth Hormone Receptor Plasma protein level Sun et al., (2018) [22] 3.3 1 BDNF Serum and plasma levels Li et al., (2020) [27] 11.8 7 C-reactive Protein UKB Said et al., (2022) [23] 575.5 266 IL-6 Blood samples Ahluwalia et al., (2021) [24] 52.7 9 IL-8 Serum or plasma levels Folkersen et al., (2020) [25] 21.8 1 TNF-α Sliz et al., (2019) [26] 13.6 1 Behavioural factors Alcohol Intake Drinks per week Saunders et al., (2022) [28] 2,428.9 496 Alcohol Dependence DSM-4 diagnosis Walters et al., (2018) [29] 46.6 (35.0) 2 Smoking Initiation Ever smoked regularly Saunders et al., (2022) [28] 2,669.0 1,346 BMI UKB Yengo et al., (2018) [30] 700 941 waist-hip-ratio UKB Pulit et al., (2019) [31] 694.6 346 Insomnia UKB Watanabe et al., (2019) [32] 386.1 31 Psychosocial factors Loneliness UKB Abdellaoui et al., (2019) [33] 511.3 (430.3) 16 Educational Attainment years-of-education, UKB Okbay et al., (2022)[34] 3,000.0 131 Educational Attainment years-of-education Okbay et al., (2016)[38] 293.7 74 T able 1: Summary-level GWAS data used for analysis. Effective sample size noted in brackets where available. UK-Biobank abbreviated as UKB. 9 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted March 22, 2024. ; https://doi.org/10.1101/2024.03.20.24304533doi: medRxiv preprint PTSD and CVD with more certainty. This approach leverages genetic variants that act as proxies for specific risk factors, mimicking the relationship between these risk factors and the outcome of interest. Because genetic variants are randomly transferred from parents to offspring, MR effectively reduces the impact of confounding and bias, offering a powerful tool for inferring causality in epidemiological research and providing valuable insights into the effects of various risk factors on health outcomes. Causal estimates rely on the use of instrumental variables, meaning SNPs that are strongly linked to the exposure (known as the relevance assumption) [40]. We identify SNPs based on their statistical significance (p < 5 · 10−8) and clump them using a European LD panel (r 2 = 0.01). MR further assumes exchangeability, prompting that these genetic variants should be interchangeable when assessing their impact on the outcome. Furthermore, it assumes that the genetic variant should only affect the outcome through its influence on the exposure, excluding the possibility of horizontal pleiotropy (extensively reviewed e.g. in [40]). We calculated causal links using inverse-variance weighted (IVW) estimates. To en- sure the robustness and credibility of these estimates, we conducted a series of sensitivity analyses, including weighted median [41] and weighted mode [42], which down-weigh or eliminate outliers and pleiotropic effects, thereby providing alternative estimates to the standard IVW method. To address the potential issue of horizontal pleiotropy, we employed the MR-Egger test. The MR-Egger intercept reflects the extent of horizontal pleiotropy, providing a corrected causal estimate with its slope [43]. Additionally, we utilized Steiger analysis [44] to gauge the influence of instrumental variables on the exposure in comparison to the outcome, thus helping to mitigate concerns regarding reverse causality. Finally, MR- PRESSO [45] was implemented to identify and remove outliers from our analysis, further enhancing the integrity of our results [40]. To avoid sample overlap and consequent infla- 10 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted March 22, 2024. ; https://doi.org/10.1101/2024.03.20.24304533doi: medRxiv preprint tion, we used the PTSD GWAS summary data excluding UKB for all mentioned methods. However, to increase power, we also used the GWAS summary data including UKB while correcting for the overlap by using the method MRlap. MRlap enhances our analysis by adjusting the IVW estimate, compensating for biases due to sample overlap through the deployment of cross-trait LDSC [46]. We define a reliable causal effect if direction and magnitude of the effect are compara- ble among all (sensitivity) methods. To rule out reverse causal effects of CVD on PTSD we conducted bidirectional MR. We investigated mediating roles of aforementioned factors using multivariable MR (MVMR) [47]. MVMR, an extension of MR, allows for estimating the causal effects of multiple exposures on an outcome by assessing the direct causal effects of each exposure. This allows to assess whether the total effect can partly be explained by certain (risk) factors. To perform MVMR, a set of SNPs associated with the expo- sure variables but not directly impacting the outcome are utilized to predict the exposures and estimate their effects on the outcome through multivariable regression analysis [47]. We found 46 independent significant SNPs for the PTSD GWAS (excluding UKB) and computed instrument strengths (F) and Cochran’s heterogeneity (Q). We instrumented 37, 36, 29 and 42 variables for the analysis of the effect of PTSD on AF, HF, HT, and CAD respectively. Initially, we could only identify 18 instrumental variables for HT, so we looked for proxies of remaining SNPs, finding an additional 11 instruments (LD > 0.9). To examine the possibility of a reverse causal effect we investigated the effect of AF, HF, HT, and CAD on PTSD respectively with 112, 10, 245, and 224 instrumental variables respectively. To ensure robustness and exclude instruments strongly correlated with the outcome, we performed Steiger filtering and repeated parameter estimation. 34, 27, 26 and 33 instrumental variables remained after Steiger filtering for the effect of PTSD on AF, 11 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted March 22, 2024. ; https://doi.org/10.1101/2024.03.20.24304533doi: medRxiv preprint HF, HT, and CAD respectively. Similarly, 112, 10, 245 and 224 instruments remained for the analysis of the effect of CVD on PTSD.

Genetic correlation and latent factor analysis /gid00002 /gid00003 CVD HT CAD HF PTSD 1 0.12 0.2 0.5 0.03 1 0.46 0.68 0.26 1 0.55 0.21 1 0.32 1 AF CAD HT HF PTSD AF CAD HT HF PTSD 0.63 (0.05) 0.43 (0.06) 0.19 (0.07) 0.68 (0.05) 0.84 (0.05) 0.71 (0.06) 0.32 (0.03) 1.00 (0.04) 1 (/f_ixed) Figure 1: Genomic SEM analyses of cardiovascular traits and PTSD. A Genetic correlation of cardio- vascular traits, that is, atrial fibrillation (AF), coronary artery disease (CAD), hypertension (HT), and heartfailure (HF), with post-traumatic stress disorder (PTSD). Factor analysis of cardiovascular traits and PTSD (B). We found modest genetic correlation of PTSD with HT, CAD, and HF ( rg = 0.21 (p = 3.21 · 10−23), rg = 0.26 (p = 2.29 · 10−37) and rg = 0.32 (p = 3.08 · 10−16) respectively, see Figure 1A). Between PTSD and AF, on the other hand, there was no clear evidence for genetic correlation (r g = 0.03 (p = 0.16), see Figure 1A), which is why we disregarded AF in the following Genomic SEM models. Using Genomic SEM, we fitted a model in which all remaining cardiovascular traits are represented by a shared latent factor. The genetic variance of the individual traits to the common genetic factor were consistently high (factor loadings > 0.4).Estimating correlation of the PTSD GWAS on the common 12 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted March 22, 2024. ; https://doi.org/10.1101/2024.03.20.24304533doi: medRxiv preprint cardiovascular factor revealed that 10.2% of the variance in the cardiovascular latent factor are shared SNP effects with PTSD (Figure 1B). To assess mediating mechanisms linking PTSD to the cardiovascular traits and their common factor, we first examined genetic correlations of PTSD and CVD traits with all mediators. We found low genetic correlation for Serotonine, Cortisol, Growth-hormone receptor, Brain-derived neurotrophic factor, TNF- α, alcohol intake and BMI. C-reactive protein, IL6, IL8, alcohol dependence, smoking initiation and insomnia showed moderate genetic correlation and loneliness high genetic correlation with PTSD (with high < −0.6< medium< −0.2< low< 0.2< medium< 0.6< high, see Figure S1). We report changes in shared genetic correlation in percentage when accounting for variance covered by the me- diator. First, we regarded cardiovascular traits individually (sketched in Figure S2A). We found that shared genetic variation of CAD and PTSD is mainly mediated by behavioural traits, with insomnia (∆ = 30.29%) and educational attainment (∆ = 27.19%) showing the largest effect. The association of PTSD with HF appeared to be mediated mostly by waist-to-hip ratio (∆ = 36.49%), followed by C-reactive protein (∆ = 25.94%), smoking initiation (∆ = 25.83%) and IL6 (∆ = 25.70%). The latter also played a leading mediating role in the correlation with HT (∆ = 39.37%), accompanied by insomnia (∆ = 39.19%) and again waist-to-hip ratio (∆ = 35.89%). Regarding mediators affecting the association of PTSD and the common genetic factor of CVD allows to draw conclusions of PTSD on general CVD rather than on individual traits (depicted in Figure S2B). We found, that insomnia (∆ = 30.37%) and waist-to-hip ratio (∆ = 29.81%) showed most mediation, fol- lowed by educational attainment (∆ = 26.43%), IL6 (∆ = 25.55%), and C-reactive protein (∆ = 24.87%). Smoking (∆ = 20.21%) and alcohol dependence (∆ = 17.52%) also play a significant role (Figure 2). 13 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted March 22, 2024. ; https://doi.org/10.1101/2024.03.20.24304533doi: medRxiv preprint IS EA SIWTH CRP IL6 AD IL8 GHR LN BMI SER BDNF AICOR TNFA −40 −30 −20 −10 0 Δ variance explained (%) CAD WTH CRP SI IL6 EA AD ISGHR IL8 SER COR AI BMI BDNF TNFA LN −40 −30 −20 −10 0 Δ variance explained (%) HF IL6 ISWTH EA AD CRP SIGHR COR BMI SER AITNFA BDNF LN IL8 −40 −30 −20 −10 0 10 Δ variance explained (%) HT ISWTH EA IL6 CRP SI AD GHR SER BMI AICOR IL8 LNBDNF TNFA −40 −30 −20 −10 0 Δ variance explained (%) CF behavioural biological psychosocial Figure 2: Extension of (factor) models to include various mediators, categorized as biological (serotonin (SER) [20], cortisol (COR) [21], growth hormone receptor (GHR) [22], C-reactive protein (CRP) [23], interleukin (IL)-6 [24], IL-8 [25], Tumor necrosis factorα (TNFA) [26], BDNF [27]), lifestyle and behavioural (alcoholic beverages per week (AI) [28], alcohol dependence (AD) [29], smoking initiation (SI) [28], BMI [30], waist-to-hip ratio (WTH) [31], insomnia (IS) [32]) and psychosocial factors (loneliness (LN) [33], educational attainment (EA) [34]), to characterize underlying correlations. We show deviations in the variance, reported as percentage change, shared between PTSD and individual CVD traits when correcting for these mediators (coronary artery disease (CAD), heart failure (HF), hypertension (HT), depicted in Figure S2A). Further, we report mediation of the relationship of PTSD and a common genetic factor of cardiovascular disease (CF) (depicted in Figure S2B)). We define a change in explained variance exceeding 10% as substantial concluding the specific factor to play a mediating role in the regarded association. Causal associations between PTSD and CVD We found strong evidence for a causal effect of PTSD on CAD (IVW OR: 1.53, 95% CI: [1.19, 1.96], p = 0.001) consistent across all (sensitivity) methods (Figure 3, Table S1). 14 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted March 22, 2024. ; https://doi.org/10.1101/2024.03.20.24304533doi: medRxiv preprint The MR-Egger intercept shows no evidence of horizontal pleiotropy (p = 0.90, Table S2). However, theI 2 statistic indicates an unreliable Egger estimate (I 2 = 0.24). Results of the analysis of Steiger filtered instruments confirmed PTSD as a causal risk factor for CAD (IVW OR: 1.37, 95% CI: [1.17, 1.61], p = 0.0001, Figure S3, Table S3). To assess the role of sample overlap in this results, we performed MRlap. The resulting IVW estimate corrected for sample-overlap also indicates a significant causal effect of PTSD on CVD, slightly smaller in magnitude (corrected OR: 1.28, 95% CI: [1.17, 1.40], p = 6 .68 · 10−8). Our MR analysis further indicated evidence for causality of PTSD on HT (IVW OR: 1.25, 95% CI: [1.05, 1.49], p = 0.012) (Figure 3) consistent across weighted median, weighted mode, and MR-PRESSO. MRlap supports this finding (corrected OR: 1.27, 95% CI: [1.09, A) PTSD as exposure B) Reverse effect Consistent statistically signi/f_icant effect of PTSD on CAD Figure 3: Analysis of Instrumental Variables of MR for the effect of A PTSD on various CVD traits and B of various CVD traits on PTSD. Number of facilitated instrumental variables in brackets. ∗p<.05, ∗∗p<.01, ∗∗∗p<.001. 15 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted March 22, 2024. ; https://doi.org/10.1101/2024.03.20.24304533doi: medRxiv preprint 1.47], p = 0.002). While MR-Egger indicated a point estimate OR < 1, this analysis was unreliable with an I 2 statistic of < 0.6 (I 2 = 0.25). Analysis of Steiger-filtered data did not produce significant estimates (Figure S3, Table S3). We also found evidence for a causal effect of PTSD on HF (IVW OR: 1.44, 95% CI: [1.08, 1.92], p = 0.012), consistent across IVW, weighted mode, weighted median and MR-PRESSO. While the point estimate of MR-Egger was contrasting in direction we again note unreliability of MR-Egger ( I 2 = 0.29) as well as wide confidence intervals. The MR-Egger intercept showed weak evidence of horizontal pleiotropy (p = 0.06, Table S2). Analysis of Steiger-filtered data indicated no evidence of significant effects of PTSD on HF (Figure S3). For both relations, PTSD HT CAD HF PTSD 0. HT CAD HF PTSD 0. HT CAD HF PTSD 0. HT CAD HF 0. rs34425 rs10487459 rs7408312 rs2107448 rs748832 rs7519259 rs10842260 rs143133717 rs11130221 rs10132977 rs10496632 rs10992779 rs1124372 rs11529859 rs11628299 rs11957630 rs12026766 rs12128161 rs12521971 rs13161130 rs13168358 rs13237518 rs13387644 rs1373273 rs1541903 rs175086 rs17514846 rs180963 rs1861188 rs1866560 rs197261 rs2186710 rs2292996 rs2899991 rs29242 rs295017 rs3132388 rs34811474 rs35791987 rs4129585 rs4489042 rs4632195 rs4652676 rs559566 rs58043442 rs6032660 rs61702433 rs61946067 rs6430728 rs6802567 rs7141058 rs7170398 rs7200432 rs7333625 rs7806900 rs78201023 rs816363 rs9479138 rs9651063 PTSD rs73338706 rs1476535 rs488769 rs34809719 rs10104247 rs2135029 rs6800637 0.04 (0.01) 0.18 (0.02) 0.21 (0.03) 0.56 (0.05) 0.48 (0.06) 0.23 (0.07) 0.66 (0.04) 0.66 (0.03) 0.81 (0.04)0.19 (0.03) 0.78 (0.04) Figure 4: Extension of the common factor model of cardiovascular traits and PTSD by including signif- icant independent SNPs from the PTSD GWAS. Significant associations were found between PTSD and cardiovascular traits, with the most pronounced effects observed on HF and CAD. Furthermore, 16 SNPs were identified that directly influence cardiovascular traits, independently of their effects on PTSD. Green factor loadings ultimately represent directional effects of PTSD-specific SNP effects on individual cardio- vascular traits. All indicated estimations are significant (p < 0.05) underscoring the robustness of our results. Variance in HT explained by PTSD is found to be negligibly small. Our findings indicate a causal association between PTSD and both CAD and HF. 16 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted March 22, 2024. ; https://doi.org/10.1101/2024.03.20.24304533doi: medRxiv preprint on CAD and on HF, Cochren’s Q-statistics shows significant heterogeneity for instruments (Table S2). All instrumental variables were of sufficient strength (F-statistics > 10; Table S2). We extent our exploration of causality in the relationship of PTSD and CVD by imple- menting individual SNP effects into our previously defined common genetic factor model from our Genomic SEM analysis. We included genome-wide significant independent SNPs of the PTSD GWAS on the PTSD indicator in the model. We found significant (p< 0.05) loadings on all regarded CVD traits with the largest effect seen on HF followed by CAD. Variance in HT explained by PTSD-specific SNPs is negligibly small. As reverse causal- ity is unlikely when facilitating genetic markers like SNPs, we conclude that these results indicate that a causal cascade from PTSD on CAD and HF may exist. Among the SNPs robustly associated with PTSD, 16 demonstrated significant direct effects on cardiovascu- lar health, thus acting on disease traits through paths distinct from PTSD. Most of these horizontal pleiotropies identified effects on CAD (Figure 4). In other words, these SNPs affect both PTSD and CAD, but the effects on CAD are not simply a consequence of their influence on PTSD. This finding implies that there are distinct biological pathways through which these SNPs exert their effects on CAD, separate from any pathways involving PTSD. MVMR analysis revealed that none of the mediators removed the entire effect of PTSD on CVD. Results indicated that insomnia (∆ = 19.86%) and smoking initiation (∆ = 23.59%) mediate part of the causal effect of PTSD on CAD. For the effect of PTSD on HF we found that smoking initiation (∆ = 42.35%) showed a change in effect, again agreeing with the Genomic SEM mediation models. Multivariable analysis with HT indicated alco- hol intake (∆ = 41.09%), insomnia (∆ = 35.79%) and BDNF (∆ = 20.83%) as possible 17 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted March 22, 2024. ; https://doi.org/10.1101/2024.03.20.24304533doi: medRxiv preprint drivers of causality. Mind that changes in effects reported are point estimates, confidence intervals of both the uni- and multivariable effects overlap for each mediator, allowing only soft indications of possible mediation. Mind that sufficient instrument strength was reached for a subset of the analyzed mediators only (F > 10 for IL6, AI, IS, SI for CAD, HF and HT and BDNF for HT, Figure S4). To examine the possibility of reverse causality, we performed MR analysis with CVD as the exposure and PTSD as outcome Table S4. All instruments were of sufficient strength (F-statistic > 10; Table S5). We found evidence that HF causally decreases the risk of PTSD (IVW OR: 0.96, 95% CI: [0.94, 0.99], p = 0.006, Figure 3, Table S4), a direction of effect that is unexpected and not in line with previous research. There was no strong evidence for horizontal pleiotropy from the MR Egger. Cochran’s test showed no evidence of heterogeneity (Table S5). I 2 statistics> 0.6 implied the need to apply Simex correction, which showed significant p-values associated with the effect on HF. MR on Steiger-filtered instruments, supported the role of HF on PTSD (IVW OR: 0.96, 95% CI: [0.94, 0.99], p = 0.006, Figure S5, Table S6).

This study was undertaken with the overarching goal of establishing the nature of the link between PTSD and CVD. We found notable genetic correlations of PTSD with HT, HF, and CAD, but not between PTSD and AF. Insomnia, waist to-hip ratio, educational attainment, smoking initiation and alcohol dependence along with inflammation indica- tors like IL6 and C-reactive protein, partially mediated these genetic associations. Using Mendelian Randomization, which allows causal inference, we found strong and consistent evidence for causal effects of PTSD on CAD and HF risk, and somewhat weaker for causal 18 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted March 22, 2024. ; https://doi.org/10.1101/2024.03.20.24304533doi: medRxiv preprint effects of PTSD on HT. With multivariable Mendelian randomization these causal effects too were partly mediated by lifestyle and biological factors, but importantly, independent causal effects of PTSD on CVD also remained. While AF is more common among people with PTSD (e.g. reviewed in [48]), in our study AF did not exhibit a significant genetic correlation with PTSD. This could be be- cause there is truly no genetic overlap between the two, or because the genome-wide genetic correlation was not able to capture the presence of localized opposing genetic correlations or divergent biological pathways between AF and PTSD. Although AF shares risk factors and biological mechanisms with the other CVD traits [49], its aetiology is predominantly electrophysical, distinguishing it from hypertension, coronary artery disease, and heart failure, which have more multifactorial causes. This distinction could contribute to the observed difference in genetic correlation with AF. As expected, AF did show a substantial genetic correlation with HF, which are known to co-occur. In future studies, advanced analytical techniques such as factor analysis or GWAS-by-subtraction [50] could provide valuable insights into these observed disparities in future investigations. We found particularly strong evidence that PTSD is a causal risk factor for developing CAD and HF (which can be the result of CAD). Whereas observational associations be- tween PTSD and CVD have been published by others (e.g. [51] and [52]), we now present convincing evidence, using the sophisticated method of MR, that PTSD causally increases CVD risk. Our evidence is strengthened by the fact that we did not find reverse, increasing effects of CVD on PTSD and is further corroborated by consistent results obtained using a novel extension of the innovative Genomic SEM framework. The consistency of findings across varied methodologies reinforces the validity of our conclusion. The knowledge on 19 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted March 22, 2024. ; https://doi.org/10.1101/2024.03.20.24304533doi: medRxiv preprint causality that we put forth is important because it can lead to targeted interventions that address mental health as a significant risk factor in CVD prevention and management. Currently, despite their heightened risk of CVD, mentally ill patients are notably under- served by clinical procedures, receiving substantially poorer screening and treatment [53]. Our study serves as important evidence that clinical intervention can aid to prevent CVD onset in PTSD patients. Our MR results also showed evidence for a casual effect of HF on PTSD. As previously noted, it is important to consider potential biases in the results due to the fact that the PTSD measure we used includes both DSM-4 and DSM-5 diagnoses. While the DSM-4 included CVD as a potential factor contributing to PTSD, the DSM-5 subsequently revised this classification, specifically excluding PTSD cases due to CVD [ 1, 11]. The fact that PTSD caused by CVD may have been excluded in clinicians assessment and electronic health records may have induced a negative association from HF to PTSD. All in all, our

at least indicate that the comorbidity between PTSD and CVD is not due to CVD increasing PTSD risk. In our study, we identified several critical mediators that explain a significant proportion of the (genetic) relationship between PTSD and CVD, offering deeper insights into their interconnectedness. Previous research has consistently shown increased inflammation to be associated with a higher risk of developing CVD [54]. Combined, our Genomic SEM and Mendelian randomization results now indicate that IL6 and CRP play a mediating, role in the causal relationship from PTSD on CVD. Our results further emphasize the role of insomnia as a critical mediator, confirming the known long-term effects of poor sleep on CVD risk. Prior evidence suggested that poor sleep is associated both with PTSD 20 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted March 22, 2024. ; https://doi.org/10.1101/2024.03.20.24304533doi: medRxiv preprint and CVD, possibly manifesting as autonomic nervous system dysregulation [35]. With our evidence of PTSD causally increasing the risk of CAD, we here emphasize an adverse effect of poor sleep quality and accompanying physiological alterations on cardiovascular health. Another important mediator that we highlighted is fat distribution. Prior research had noted that fat distribution, rather than BMI alone, may play an essential role in the predisposition for various diseases including cardiometabolic disease [31]. We here con- firm this finding as we found that waist-to-hip ratio played a significant mediating role in the relationship between PTSD and the overarching cardiovascular factor, whereas BMI did not. The finding that insomnia and waist-to-hip ratio significantly mediate the causal pathway from PTSD to CVD is highly valuable, as it suggest that sleep quality and fat distribution should be focused on in the prevention of CVD in PTSD patients. Finally, our research revealed mediating, causal effects of both smoking initiation and alcohol de- pendence on the association between PTSD and CVD. Further, we found that educational attainment partially explained the association of PTSD and CVD. This indicates that a higher socioeconomic status may play a protective role against CVD among PTSD patients. This study shed light on the directionality and mechanisms underlying the association between PTSD and CVD. LDSC, used for studying genetic correlations, effectively iden- tifies genome-wide associations between PTSD and cardiovascular traits, independent of sample overlap. Unfortunately, LDSC requires data stemming from similar ancestry to operate limiting its applicability in trans-ancestry samples [18]. Here, we implemented LDSC on European ancestry GWAS summary statistics only. Extending this, Genomic SEM allows the linking of multiple traits but inherits LDSC’s limitations and demands large sample sizes and intensive computation citepgrotzinger2019genomic. Neither method accounts for epigenetic or environmental factors in disease correlation. We further applied 21 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted March 22, 2024. ; https://doi.org/10.1101/2024.03.20.24304533doi: medRxiv preprint MR, which provides stronger evidence for causal effects than observational studies, making it invaluable for policy and clinical guidelines [40]. In summary, our findings showed shared genetic liability between PTSD and CVD, as well as strong evidence for causal effects of PTSD on CVD, for CAD and HF in particular. We identified biological and behavioural mediators of the causal cascade of PTSD on CVD which holds significant value for future clinical implications. We highlighted that moni- toring waist-to-hip ratio, inflammatory markers, smoking, alcohol use, and sleep quality is essential for early intervention and prevention of CVD in these patients. Ultimately, this study serves as a proof of an increased vulnerability of PTSD patients for CVD and calls to act accordingly in the clinical setting.

The authors thank the PGC-PTSD investigators for making their data available. Major financial support for the PGC-PTSD workgroup was provided by the Cohen Veterans Bio- science, Stanley Center for Psychiatric Research at the Broad Institute, and the National Institute of Mental Health (NIMH; R01MH106595, R01MH124847, R01MH124851). The authors further thank Michel G Nivard (Department of Biological Psychology and EMGP+ Institute for Health and Care Research, Vrije Universiteit Amsterdam, Amsterdam, the Netherlands) for valuable feedback. Funding EL, RRV, and JLT are supported by a Senior Scientist Dekker Grant from the Dutch Heart Foundation (project number 03-004-2022-0055). JLT is supported by a European Research 22 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted March 22, 2024. ; https://doi.org/10.1101/2024.03.20.24304533doi: medRxiv preprint Council (ERC) Starting grant (UNRAVEL-CAUSALITY, grant number 101076686) and by Foundation Volksbond Rotterdam. RP is supported by grants from the National Insti- tutes of Health (RF1 MH132337 and R33DA047527) and One Mind. G.A.P. acknowledges support from the Yale Biological Sciences Training Program (T32 MH014276), Alzheimer’s Association (AARF-22-967171), NIH National Institute of Aging (K99AG078503), and the Yale Franke Fellowship in Science & Humanities. Disclosure of interest RP is paid for his editorial work on the journal Complex Psychiatry and received a re- searchgrant outside the scope of this study from Alkermes. The other authors report no conflict of interest. Data availability statement Where available publicly available GWAS data was used. In absence of downloadable sum- mary statistics, we contacted corresponding authors. PTSD GWAS summary statistics will be available on the website of the Psychiatric Genomics Consortium (https://pgc.unc.edu/for- researchers/download-results/).

[1] American Psychiatric Association. Diagnostic and statistical manual of mental disor- ders. 5th ed.; 2013. [2] Smoller JW. The genetics of stress-related disorders: PTSD, depression, and anxiety disorders. Neuropsychopharmacology. 2016;41(1):297-319. 23 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted March 22, 2024. ; https://doi.org/10.1101/2024.03.20.24304533doi: medRxiv preprint [3] Nievergelt CM, Maihofer AX, Atkinson EG, Chen CY, Choi KW, Coleman JR, et al. Discovery of 95 PTSD loci provides insight into genetic architecture and neurobiology of trauma and stress-related disorders. medRxiv. 2023. Available from: https:// www.medrxiv.org/content/early/2023/09/02/2023.08.31.23294915. [4] Edmondson D, Cohen BE. Posttraumatic stress disorder and cardiovascular disease. Progress in cardiovascular diseases. 2013;55(6):548-56. [5] De Hert M, Detraux J, Vancampfort D. The intriguing relationship between coronary heart disease and mental disorders. Dialogues in clinical neuroscience. 2018;20(1):31- 40. [6] Hargrave AS, Sumner JA, Ebrahimi R, Cohen BE. Posttraumatic Stress Disorder (PTSD) as a Risk Factor for Cardiovascular Disease: Implications for Future Research and Clinical Care. Current Cardiology Reports. 2022:1-13. [7] Krantz DS, Shank LM, Goodie JL. Post-traumatic stress disorder (PTSD) as a systemic disorder: Pathways to cardiovascular disease. Health Psychology. 2022;41(10):651. [8] Pollard HB, Shivakumar C, Starr J, Eidelman O, Jacobowitz DM, Dalgard CL, et al. “Soldier’s Heart”: a genetic basis for elevated cardiovascular disease risk associated with post-traumatic stress disorder. Frontiers in molecular neuroscience. 2016;9:87. [9] Morris MC, Hellman N, Abelson JL, Rao U. Cortisol, heart rate, and blood pressure as early markers of PTSD risk: A systematic review and meta-analysis. Clinical psychology review. 2016;49:79-91. [10] Sumner JA, Cleveland S, Chen T, Gradus JL. Psychological and biological mech- 24 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted March 22, 2024. ; https://doi.org/10.1101/2024.03.20.24304533doi: medRxiv preprint anisms linking trauma with cardiovascular disease risk. Translational Psychiatry. 2023;13(1):25. [11] American Psychiatric Association. Diagnostic and statistical manual of mental disor- ders. 4th ed.; 1995. [12] Uffelmann E, Huang QQ, Munung NS, De Vries J, Okada Y, Martin AR, et al. Genome-wide association studies. Nature Reviews Methods Primers. 2021;1(1):59. [13] Sanderson E, Glymour MM, Holmes MV, Kang H, Morrison J, Munaf` o MR, et al. Mendelian randomization. Nature Reviews Methods Primers. 2022;2(1):6. [14] Roselli C, Chaffin MD, Weng LC, Aeschbacher S, Ahlberg G, Albert CM, et al. Multi-ethnic genome-wide association study for atrial fibrillation. Nature genetics. 2018;50(9):1225-33. [15] Aragam KG, Jiang T, Goel A, Kanoni S, Wolford BN, Atri DS, et al. Discovery and systematic characterization of risk variants and genes for coronary artery disease in over a million participants. Nature Genetics. 2022:1-13. [16] Zhu Z, Wang X, Li X, Lin Y, Liu C, Hobbs B, et al. Genetic Overlap of Chronic Obstructive Pulmonary Disease and Cardiovascular Diseases: A Large-Scale Genome- Wide Cross-Trait Analysis. In: C32. COPD: TRANSLATIONAL AND MECHANIS- TIC STUDIES. American Thoracic Society; 2019. p. A4549-9. [17] Shah S, Henry A, Roselli C, Lin H, Sveinbj¨ ornsson G, Fatemifar G, et al. Genome-wide association and Mendelian randomisation analysis provide insights into the pathogen- esis of heart failure. Nature communications. 2020;11(1):163. [18] Bulik-Sullivan BK, Loh PR, Finucane HK, Ripke S, Yang J, of the Psychiatric Ge- 25 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted March 22, 2024. ; https://doi.org/10.1101/2024.03.20.24304533doi: medRxiv preprint nomics Consortium SWG, et al. LD Score regression distinguishes confounding from polygenicity in genome-wide association studies. Nature genetics. 2015;47(3):291-5. [19] Grotzinger AD, Rhemtulla M, de Vlaming R, Ritchie SJ, Mallard TT, Hill WD, et al. Genomic structural equation modelling provides insights into the multivariate genetic architecture of complex traits. Nature human behaviour. 2019;3(5):513-25. [20] Hysi PG, Mangino M, Christofidou P, Falchi M, Karoly ED, Investigators NB, et al. Metabolome genome-wide association study identifies 74 novel genomic regions influ- encing plasma metabolites levels. Metabolites. 2022;12(1):61. [21] Crawford AA, Bankier S, Altmaier E, Barnes CL, Clark DW, Ermel R, et al. Variation in the SERPINA6/SERPINA1 locus alters morning plasma cortisol, hepatic corticos- teroid binding globulin expression, gene expression in peripheral tissues, and risk of cardiovascular disease. Journal of human genetics. 2021;66(6):625-36. [22] Sun BB, Maranville JC, Peters JE, Stacey D, Staley JR, Blackshaw J, et al. Genomic atlas of the human plasma proteome. Nature. 2018;558(7708):73-9. [23] Said S, Pazoki R, Karhunen V, V˜ osa U, Ligthart S, Bodinier B, et al. Genetic analysis of over half a million people characterises C-reactive protein loci. Nature communica- tions. 2022;13(1):2198. [24] Ahluwalia TS, Prins BP, Abdollahi M, Armstrong NJ, Aslibekyan S, Bain L, et al. Genome-wide association study of circulating interleukin 6 levels identifies novel loci. Human molecular genetics. 2021;30(5):393-409. [25] Folkersen L, Gustafsson S, Wang Q, Hansen DH, Hedman ˚AK, Schork A, et al. Ge- nomic and drug target evaluation of 90 cardiovascular proteins in 30,931 individuals. Nature metabolism. 2020;2(10):1135-48. 26 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted March 22, 2024. ; https://doi.org/10.1101/2024.03.20.24304533doi: medRxiv preprint [26] Sliz E, Kalaoja M, Ahola-Olli A, Raitakari O, Perola M, Salomaa V, et al. Genome- wide association study identifies seven novel loci associating with circulating cytokines and cell adhesion molecules in Finns. Journal of medical genetics. 2019;56(9):607-16. [27] Li S, Weinstein G, Zare H, Teumer A, V¨ olker U, Friedrich N, et al. The genetics of circulating BDNF: towards understanding the role of BDNF in brain structure and function in middle and old ages. Brain Communications. 2020;2(2):fcaa176. [28] Saunders GR, Wang X, Chen F, Jang SK, Liu M, Wang C, et al. Genetic diversity fuels gene discovery for tobacco and alcohol use. Nature. 2022;612(7941):720-4. [29] Walters RK, Polimanti R, Johnson EC, McClintick JN, Adams MJ, Adkins AE, et al. Transancestral GWAS of alcohol dependence reveals common genetic underpinnings with psychiatric disorders. Nature neuroscience. 2018;21(12):1656-69. [30] Yengo L, Sidorenko J, Kemper KE, Zheng Z, Wood AR, Weedon MN, et al. Meta- analysis of genome-wide association studies for height and body mass index in 700000 individuals of European ancestry. Human molecular genetics. 2018;27(20):3641-9. [31] Pulit SL, Stoneman C, Morris AP, Wood AR, Glastonbury CA, Tyrrell J, et al. Meta- analysis of genome-wide association studies for body fat distribution in 694 649 indi- viduals of European ancestry. Human molecular genetics. 2019;28(1):166-74. [32] Watanabe K, Stringer S, Frei O, Umi´ cevi´ c Mirkov M, de Leeuw C, Polderman TJ, et al. A global overview of pleiotropy and genetic architecture in complex traits. Nature genetics. 2019;51(9):1339-48. [33] Abdellaoui A, Sanchez-Roige S, Sealock J, Treur JL, Dennis J, Fontanillas P, et al. Phenome-wide investigation of health outcomes associated with genetic predisposition to loneliness. Human molecular genetics. 2019;28(22):3853-65. 27 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted March 22, 2024. ; https://doi.org/10.1101/2024.03.20.24304533doi: medRxiv preprint [34] Okbay A, Wu Y, Wang N, Jayashankar H, Bennett M, Nehzati SM, et al. Polygenic prediction of educational attainment within and between families from genome-wide association analyses in 3 million individuals. Nature genetics. 2022;54(4):437-49. [35] Wilson MA, Liberzon I, Lindsey ML, Lokshina Y, Risbrough VB, Sah R, et al. Com- mon pathways and communication between the brain and heart: connecting post- traumatic stress disorder and heart failure. Stress. 2019;22(5):530-47. [36] Black PH. Central nervous system-immune system interactions: psychoneuroen- docrinology of stress and its immune consequences. Antimicrobial agents and chemotherapy. 1994;38(1):1-6. [37] Reiche EMV, Nunes SOV, Morimoto HK. Stress, depression, the immune system, and cancer. The lancet oncology. 2004;5(10):617-25. [38] Okbay A, Beauchamp JP, Fontana MA, Lee JJ, Pers TH, Rietveld CA, et al. Genome- wide association study identifies 74 loci associated with educational attainment. Na- ture. 2016;533(7604):539-42. [39] Hemani G, Zheng J, Elsworth B, Wade K, Baird D, Haberland V, et al. The MR- Base platform supports systematic causal inference across the human phenome. eLife. 2018;7:e34408. Available from: https://elifesciences.org/articles/34408. [40] Richmond RC, Smith GD. Mendelian randomization: concepts and scope. Cold Spring Harbor perspectives in medicine. 2022;12(1):a040501. [41] Bowden J, Davey Smith G, Haycock PC, Burgess S. Consistent estimation in Mendelian randomization with some invalid instruments using a weighted median estimator. Genetic epidemiology. 2016;40(4):304-14. 28 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted March 22, 2024. ; https://doi.org/10.1101/2024.03.20.24304533doi: medRxiv preprint [42] Hartwig FP, Davey Smith G, Bowden J. Robust inference in summary data Mendelian randomization via the zero modal pleiotropy assumption. International journal of epidemiology. 2017;46(6):1985-98. [43] Bowden J, Davey Smith G, Burgess S. Mendelian randomization with invalid instru- ments: effect estimation and bias detection through Egger regression. International journal of epidemiology. 2015;44(2):512-25. [44] Hemani G, Tilling K, Davey Smith G. Orienting the causal relationship be- tween imprecisely measured traits using GWAS summary data. PLoS genetics. 2017;13(11):e1007081. [45] Verbanck M, Chen CY, Neale B, Do R. Detection of widespread horizontal pleiotropy in causal relationships inferred from Mendelian randomization between complex traits and diseases. Nature genetics. 2018;50(5):693-8. [46] Mounier N. MRlap: R-package to perform two-sample Mendelian Randomisation anal- yses using potentially overlapping samples; 2023. GitHub repository. Available from: https://github.com/n-mounier/MRlap. [47] Sanderson E, Spiller W, Bowden J. Testing and correcting for weak and pleiotropic instruments in two-sample multivariable Mendelian randomization. Statistics in medicine. 2021;40(25):5434-52. [48] Habbal AB, White CT, Shamim H, Al Shouli R, Mohammed L. Posttraumatic Stress Disorder (PTSD) and Instigation of Cardiovascular Events: Ischemic Heart Disease (IHD) and Atrial Fibrillation (AF). Cureus. 2022;14(10). [49] Bizhanov KA, Abzaliyev KB, Baimbetov AK, Sarsenbayeva AB, Lyan E. Atrial fibril- 29 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted March 22, 2024. ; https://doi.org/10.1101/2024.03.20.24304533doi: medRxiv preprint lation: Epidemiology, pathophysiology, and clinical complications (literature review). Journal of Cardiovascular Electrophysiology. 2023;34(1):153-65. [50] Demange PA, Malanchini M, Mallard TT, Biroli P, Cox SR, Grotzinger AD, et al. In- vestigating the genetic architecture of noncognitive skills using GWAS-by-subtraction. Nature Genetics. 2021;53(1):35-44. [51] Kim K, Tsai AC, Sumner JA, Jung SJ. Posttraumatic stress disorder, cardiovascular disease outcomes and the modifying role of socioeconomic status. Journal of Affective Disorders. 2022;319:555-61. [52] Boscarino JA, Chang J. Electrocardiogram abnormalities among men with stress- related psychiatric disorders: implications for coronary heart disease and clinical re- search. Annals of Behavioral Medicine. 1999;21(3):227-34. [53] Solmi M, Fiedorowicz J, Poddighe L, Delogu M, Miola A, Høye A, et al. Disparities in screening and treatment of cardiovascular diseases in patients with mental disorders across the world: systematic review and meta-analysis of 47 observational studies. American Journal of Psychiatry. 2021;178(9):793-803. [54] Ridker PM, Rane M. Interleukin-6 signaling and anti-interleukin-6 therapeutics in cardiovascular disease. Circulation research. 2021;128(11):1728-46. [55] Nath AP, Ritchie SC, Grinberg NF, Tang HHF, Huang QQ, Teo SM, et al. Multi- variate genome-wide association analysis of a cytokine network reveals variants with widespread immune, haematological, and cardiometabolic pleiotropy. The American Journal of Human Genetics. 2019;105(6):1076-90. [56] Backman JD, Li AH, Marcketta A, Sun D, Mbatchou J, Kessler MD, et al. 30 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted March 22, 2024. ; https://doi.org/10.1101/2024.03.20.24304533doi: medRxiv preprint Exome sequencing and analysis of 454,787 UK Biobank participants. Nature. 2021;599(7886):628-34. 31 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted March 22, 2024. ; https://doi.org/10.1101/2024.03.20.24304533doi: medRxiv preprint S1 Supplementary notes Data preparation and analysis Included traits and their respective GWAS are summarized in Table 1. Slight alterations to the preregistered data utilization were necessary as no effect size was published for cy- tokine meausurements [55] and one dataset contained mainly rare variants for mediacation (non-) adherence [56]. These datasets were excluded. A small deviation from the pub- lished preregistration included the usage of a larger dataset for alcohol intake [28]. For MR analysis we facilitate both alcohol intake and smoking initiation from the same authors excluding UK Biobank entries to avoid sample overlap with UKB-containing cardio traits in multivariate MR. As the largest and newest GWAS available of educational attainment contains UKB [34], we facilitated their earlier disocvery sample for MVMR to avoid sample overlap [38]. Data preparation for MR included filtering for p <5 · 10−8 and subsequent clumping (r 2 = 0.01) using the R-package TwoSampleMR [39]. 32 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted March 22, 2024. ; https://doi.org/10.1101/2024.03.20.24304533doi: medRxiv preprint 1 0.46 0.68 0.26 −0.07 0.08 0.14 −0.04 0.29 0.22 0.12 −0.07 0.01 0.18 0.24 0.3 −0.13 0.25 0.2 −0.28 1 0.55 0.21 −0.04 −0.08 0.2 0.09 0.3 0.32 0 −0.03 0 0.2 0.12 0.36 −0.07 0.23 0.14 −0.24 1 0.32 −0.25 −0.19 0.28 0.02 0.41 0.35 0.16 0 0.08 0.25 0.3 0.55 0.01 0.24 0.22 −0.27 1 −0.07 −0.09 0.09 −0.11 0.23 0.29 0.33 0.15 0.04 0.58 0.41 0.23 −0.09 0.49 0.73 −0.34 1 0.45 0.46 0.53 0.03 0.25 0.24 −0.26 −0.03 0.17 −0.14 0.02 −0.29 0.02 0.1 0.03 1 0.02 −0.33 −0.13 −0.44 −0.26 0.75 0.08 0.02 −0.1 −0.31 −0.07 −0.06 0.05 −0.04 1 −0.23 0.18 0.46 0.02 0.02 −0.15 0.23 −0.01 0.25 −0.03 0.03 0.22 −0.14 1 0.14 −0.03 0.57 −0.06 0.03 −0.27 0 0.09 −0.06 0.03 0.04 −0.18 1 0.59 0 0.06 −0.04 0.14 0.26 0.55 −0.1 0.17 0.2 −0.31 1 −0.22 −0.18 −0.03 0.45 0.32 0.51 0.03 0.11 0.16 −0.31 1 −0.26 0.21 −0.49 0.1 0.21 −0.04 0.09 0.22 −0.04 1 −0.02 −0.21 0.06 −0.04 −0.05 −0.03 −0.07 −0.01 1 0.6 0.41 −0.03 0 0.11 −0.01 0.04 1 0.67 0.12 −0.04 0.28 0.33 −0.41 1 0.25 −0.02 0.25 0.23 −0.38 1 −0.11 0.19 0.23 −0.32 1 −0.07 −0.07 0.14 1 0.42 −0.33 1 −0.3 1 CAD HT HF PTSD SER COR GHR BDNF CRP IL6 IL8 TNFA AI AD SI WTH BMI IS LN EA CAD HT HF PTSD SER COR GHRBDNF CRP IL6 IL8 TNFA AI AD SI WTH BMI IS LN EA Figure S1: Genetic correlation performed as LDSC for all relevant traits sed for Genomic SEM analysis. 33 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted March 22, 2024. ; https://doi.org/10.1101/2024.03.20.24304533doi: medRxiv preprint CVD HT CAD HF PTSD Mediator Mediator P/C C/P A B Figure S2: Correlation (green) and mediation (blue) models for A individual cardiovascular traits and PTSD, for B a common cardiovasular factor and PTSD and for C a common cardiovascular and PTSD factor. 34 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted March 22, 2024. ; https://doi.org/10.1101/2024.03.20.24304533doi: medRxiv preprint * *** *** ** AF (34) CAD (33) HF (27) HT (26) 1 2 3 4 5 6 7 8 9 10 Odds ratio Inverse variance weighted MR Egger Weighted median Weighted mode Figure S3: Analysis of Steiger filtered instrumental variables for the effect of PTSD on various CVD traits (number of facilitated SNPs). ∗p<.05, ∗∗p<.01, ∗∗∗p<.001 35 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted March 22, 2024. ; https://doi.org/10.1101/2024.03.20.24304533doi: medRxiv preprint AD EA SI IS TNFA GHR IL8 BDNF AI uni COR IL6 1.0 1.5 2.0 CAD SI AD TNFA EA IS COR IL8 uni GHR BDNF AI IL6 1.2 1.6 2.0 HF AD AI IS BDNF TNFA IL8 COR uni EA GHR SI IL6 1.001.251.501.75 HT Figure S4: Multivariate MR analysis of the effect of PTSD on CVD traits that previously showed a significant IVW. Dashed lines indicate insufficient instrument strengths measured by F-statistics falling below 10. 36 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted March 22, 2024. ; https://doi.org/10.1101/2024.03.20.24304533doi: medRxiv preprint * * ** ** AF (112) CAD (224) HF (10) HT (245) 0.85 0.90 0.95 1.00 Odds ratio Inverse variance weighted MR Egger Weighted median Weighted mode Figure S5: Analysis of Steiger filtered instrumental variables for the effect of various CVD traits (number of facilitated SNPs) on PTSD. ∗p<.05, ∗∗p<.01, ∗∗∗p<.001 37 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted March 22, 2024. ; https://doi.org/10.1101/2024.03.20.24304533doi: medRxiv preprint