Risk of cardiovascular disease after a diagnosis of common psychiatric disorders: a matched cohort study of disease susceptibility and progression trajectory in the UK Biobank

Background: Whether associations between psychiatric disorders and cardiovascular diseases (CVDs) can be modified by disease susceptibility and the temporal pattern of these associated CVDs remain unknown.

Methods: We conducted a matched cohort study of UK Biobank including 35,227 patients with common psychiatry disorders (anxiety, depression, and stress-related disorders) between 1997 and 2019, together with 176,135 sex- and birth year- individually matched unexposed individuals.

Results: The mean age at the index date was 51.76 years, and 66.0% of participants were females. During a mean follow-up of 11.94 years, we observed an elevated risk of CVD among patients with studied psychiatry disorders, compared with matched unexposed individuals (hazard ratios [HRs]=1.16, 95% confidence interval [CI]: 1.14-1.19), especially during the first six months of follow-up (HR=1.59 [1.42-1.79]). To assess the modification role of disease susceptibility, we stratified analyses by family history of CVD and by CVD PRS, which obtained similar estimates between subgroups with different susceptibilities to CVD. We conducted trajectory analysis to visualize the temporal pattern of CVDs after common psychiatry disorders, identifying primary hypertension, acute myocardial infarction, and stroke as three main intermediate steps leading to further increased risk of other CVDs.

Conclusions: The association between common psychiatry disorders and subsequent CVD is not modified by predisposition to CVD. Hypertension, acute myocardial infarction, and stroke are three initial CVDs linking psychiatric disorders to other CVD squeals, highlighting a need of timely intervention on these targets to prevent further CVD squeals among all individuals with common psychiatric disorders.

Funding: This work is supported by the National Natural Science Foundation of China (No. 81971262 to HS), 1.3.5 project for disciplines of excellence, West China Hospital, Sichuan University (No. ZYYC21005 to HS), EU Horizon2020 Research and Innovation Action Grant (847776 to UV and FF).


Disease susceptibility of CVD 151
In the present study, we assessed the disease susceptibility to CVD in two ways. First, we 152 defined family history of CVD by a self-reported history of heart diseases and stroke among 153 first-degree relatives (i.e., father, mother and siblings), obtained from baseline questionnaires 154 (Field ID: 20107, 20110, 20111). Second, a polygenic risk score (PRS) for CVD was 155 calculated for 376,019 individuals with eligible genotyping data (6,099,107 SNPs available 156 after a standard quality control (Marees et al., 2018)) using LDPred2 (Prive,Arbel,& 157 Vilhjalmsson, 2020), a method for PRS computation based on combination of summary 158 statistics and a matrix of correlations between genetic variants. The independent SNPs 159 associated with CVD were derived from meta-analyzed GWAS summary data from 160 CARDIoGRAMplusC4D Consortium (Nikpay et al., 2015). In a validation step, the generated 161 CVD PRS was shown to be significantly associated with an increased risk of the CVD 162 . CC-BY 4.0 International license It is made available under a perpetuity.

Covariates 165
We retrieved information on educational levels, ethnicity, smoking status, Townsend 166 deprivation index (TDI), and body mass index (BMI) through the baseline questionnaires. 167 Based on the UK Biobank inpatient hospital data, we calculated the Charlson comorbidity 168 index (CCI) on the index date for each individual, as a proxy of their baseline comorbidity 169 level (Quan et al., 2005). 170

Statistical analysis 171
We used Cox model to estimate hazard ratios (HRs) with 95% CIs of any CVD, in relation to 172 a previous diagnosis of common psychiatry disorders, using time since the index date as the 173 underlying time scale. The models were stratified by matched factors (sex and birth year), and 174 partially or fully adjusted for educational levels, ethnicity, smoking status, TDI, BMI, history 175 of other psychiatry disorders, family history of CVD, and CCI. As a high-risk time window 176 for CVD has been indicated after diagnosis of psychiatric disorders (Song et al., 2019), we 177 first visualized the changes of HRs over time by stratifying the Cox models to different 178 follow-up periods (≤3, 3-6, 6-12, 12-18, 18-24, 24-60, 60-120, 120-240, and >240 months 179 follow-up). As the HR during the first six months of follow-up period was greater than those 180 of the later times (eFigure 1 in the supplement), we assessed the risk of CVD within and 181 beyond the first six months (i.e., ≤6 or >6 months of follow-up) separately in later analyses. 182 In addition, we did separate analyses for six main types of CVDs as well as for anxiety, 183 depression and stress-related disorders. 184 . CC-BY 4.0 International license It is made available under a perpetuity.
is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted September 29, 2021. ;https://doi.org/10.1101https://doi.org/10. /2021 To determine the impact of disease susceptibility to CVD on the association between 185 common psychiatric disorders and CVD, we conducted stratified analyses by family history 186 of CVD (yes or no) and by level of CVD PRS (high, moderate, or low by tertile distribution). 187 The temporal progression of CVD subsequent to a diagnosis of psychiatry disorders was 188 explored using trajectory network analysis as described in our previous study (Han et al., 189 2021). In brief, a phenome-wide association analysis (PheWAS) was performed to identify 190 specific CVDs associated with a prior diagnosis of psychiatry disorders. In this step, the 191 p-value for statistical significance was set to 0.05/number of analyses performed (Bonferroni 192 corrections), to account for multiple testing. We then used binomial tests to identify pairs of 193 CVD events with a temporal order, and assessed the magnitude of the associations between 194 CVD pairs through a nested case-control study design and conditional logistic regression 195 (eFigure 2 in the supplement). 196 To test the robustness of our results to the definition of CVD, we reassessed these 197 associations by identifying CVD solely based on the primary diagnosis from UK Biobank 198 inpatient hospital data. All statistical analyses were conducted by R (version 4.0.2), PLINK 199 (version 1.9), Python (version 3.8) and Cytoscape (version 3.8.0). Codes script used in the 200 primary analyses are available (Source code 1). 201 202

203
In total, we identified 35,227 patients with studied common psychiatry disorders, together 204 with 176,135 sex-and birth year-matched unexposed individuals (Figure 1) is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted September 29, 2021. ;https://doi.org/10.1101https://doi.org/10. /2021 with different levels of CVD PRS, within six months of follow-up (2.29 [1.58-3.32

Trajectory network analysis 234
15 CVDs with more than 150 cases among patients with common psychiatry disorders were 235 involved in the PheWAS analysis (eFigure 2 in the supplement). Among these, eight CVDs 236 showed significant associations with a prior diagnosis of common psychiatry disorders 237 (eTable 5 in the supplement), forming 56 CVD pairs among which 21 passed the threshold 238 of prevalence (i.e., experienced by at least 75 individuals). Finally, 12 pairs contributed to the 239 visualized trajectory networks (Figure 2 and eTable 5 in the supplement), presenting as a 240 CVD disease tree originated from primary hypertension, acute myocardial infarction and 241 stroke, which further connected to a wide range of other CVDs. 242 In the sensitivity analysis, defining the CVD cases solely by primary diagnoses, instead 243 of all diagnoses, in UK Biobank inpatient data led to largely similar estimates (Tables 6 and 7 244 in the supplement). However, the number of CVDs involved in the trajectory network was 245 substantially reduced, due to insufficient data power, where we had only 2 CVD pairs in the 246 disease tree with an initial node of acute myocardial infarction (eFigures 4 and 5 in the 247 supplement). is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted September 29, 2021. ;https://doi.org/10.1101https://doi.org/10. /2021 In this matched cohort study of UK Biobank, we found that individuals with common 251 psychiatry disorders (i.e., anxiety, depression, and stress-related disorders) were at elevated 252 risk of multiple CVDs, especially within first six months of follow-up. With mutually verified 253 results obtained from stratified analyses by family history of CVD and by CVD PRS, the 254 observed associations seemed to be constant across individuals with different predisposition 255 to CVD. Furthermore, trajectory network analysis indicated that primary hypertension, acute 256 myocardial infarction, and stroke were the CVD types firstly affected by a diagnosis of 257 common psychiatric disorders which then further led to other CVDs downstream. This result 258 highlights the need of timely intervention on these targets for preventing or interrupting 259 further CVD squeals among individuals with common psychiatric disorders. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted September 29, 2021. ; https://doi.org/10.1101/2021.09.28.21264283 doi: medRxiv preprint additional adjustment for family history of CVD (Kubzansky, Koenen, Spiro, Vokonas, & 273 Sparrow, 2007;Sumner et al., 2015). Our analysis on CVD PRS expanded this 274 knowledgebase further and, for the first time, showed similar associations across groups with 275 different levels of disease susceptibility to CVD. Taken together, these results underscore the 276 importance of surveillance for CVD events among individuals with psychiatry disorders, 277 regardless of disease liability. 278 Based on the trajectory network analysis, we identified three specific CVDs first affected 279 by a prior diagnosis of common psychiatric disorders (i.e., primary hypertension, followed by is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted September 29, 2021. ; https://doi.org/10. 1101/2021 suggest that these key CVDs may be potential targets for preventing or interrupting further 295 CVD sequels among individuals with psychiatric disorders. 296 The major strength of our study includes the multidimensional data from UK Biobank, i.e., 297 rich phenotypic information on family history of CVD, health-related outcomes, and 298 individual-level genotype information, which enabled a comprehensive assessment on the 299 modification role of CVD susceptibility on the studied associations and the temporal patterns 300 of CVD development following a diagnosis of common psychiatry disorders. This rich data 301 source also enabled vigorous control of a wide range of important confounders, including 302 socio-demographic and behavioral factors, as well as baseline health status. Other strengths 303 include the long and complete follow-up (mean follow-up time of 11.94 years), providing 304 sufficient surveillance period for CVD outcomes, and the independent collections of 305 diagnoses for psychiatry disorders and CVD, which minimizes the risk of information bias. 306 Our study has several limitations. First, as both primary and secondary diagnoses from UK 307 Biobank inpatient hospital data were used for CVD identification, the diagnosis date for some 308 mild forms of CVDs (e.g., primary hypertension) may not be accurate, leading to a risk of 309 reverse causality. Consequently, the appearance of primary hypertension as the initial node in 310 the CVD trajectory tree of the main analysis may rather imply the importance of primary 311 hypertension as a comorbidity of psychiatric disorders on other subsequent CVDs, instead of 312 a key progression pathway. Nevertheless, our analyses on acute and severe CVDs (e.g., acute 313 myocardial infarction) should be less affected by such concern. Second, the SNPs used in the 314 calculation for CVD PRS was obtained from GWAS studies which mainly focused on 315 coronary arterial diseases, not all CVDs. However, a common genetic basis has been 316 . CC-BY 4.0 International license It is made available under a perpetuity.
is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint Dadgarmoghaddam, & Sadr-Nabavi, 2019), and the generated PRS has been confirmed to be 318 associated with the CVD phenotype of our dataset in our validation step. Furthermore, we 319 obtained largely similar result patterns in the analyses of six major categories of CVDs. Third, 320 as the trajectory network analysis is a data-driven approach which requires a large sample size, 321 our analysis might have missed some rare CVDs in the trajectory networks after psychiatric 322 disorders. Further, because some important confounders, such as BMI and smoking status, 323 were measured at baseline, they might not accurately reflect the real status of these factors at 324 the time of psychiatric disorder diagnosis. Last, generalization of our findings to the entire 325 UK population, or other populations, should be done with caution because the UK Biobank 326 participants were overall healthier than the general UK polulation (Sudlow et al., 2015). 327 In conclusion, individuals with common psychiatry disorders have an elevated risk of 328 CVD regardless of their predisposition to CVD. Hypertension, acute myocardial infarction, 329 and stroke represent the three initial CVDs that link psychiatric disorders to further CVDs, 330 highlighting a need of a timely intervention to prevent further CVD sequels among 331 individuals with common psychiatric disorders. 332 333

Acknowledgments 334
This research has been conducted using the UK Biobank Resource under Application 54803. 335

We thank the team members and colleagues involved in West China Biomedical Big Data 336
Center-UK Biobank project for their support. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted September 29, 2021. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint  is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted September 29, 2021. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted September 29, 2021. ; https://doi.org/10. 1101/2021 Outcomes. J Am Heart Assoc, 10 (1) is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted September 29, 2021. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted September 29, 2021. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted September 29, 2021. ; https://doi.org/10.1101/2021.09.28.21264283 doi: medRxiv preprint and ischemic heart disease: a twin study. Acta Psychiatr Scand, 140 (3) is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted September 29, 2021. ;https://doi.org/10.1101https://doi.org/10. /2021  is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted September 29, 2021. ;https://doi.org/10.1101https://doi.org/10. /2021  is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint   is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted September 29, 2021. ;https://doi.org/10.1101https://doi.org/10. /2021  b Hazard ratios (with 95% confidence intervals) of cardiovascular disease in patients with common psychiatry disorders compared to sex and year of birth matched unexposed 546 individuals, derived from Cox regression models and adjusted for covariates listed in model information column. Time since index date was used as time scale. 547 c Hazard ratios (with 95% confidence intervals) of subtypes of cardiovascular diseases in individual with common psychiatry disorders compared to sex and year of birth 548 matched unexposed individuals, derived from Cox regression models and adjusted for sex, birth year, educational level, ethnicity, smoking status, TDI, BMI, history of other 549 psychiatry disorders, family history of CVD, and CCI score. For subtypes of cardiovascular diseases with less than 20 cases in patients with common psychiatry disorders, 550 hazard rations were derived from Cox regression models and adjusted for sex, birth year. Definition of subtypes cardiovascular diseases by ICD-10 codes can be found in the 551 eTable 1 in the supplement. Time since index date was used as time scale. 552 553 32

Supplementary tables and figures eTable 1 International Classification of Diseases (ICD) codes and primary care codes for diseases identifications eTable 2 Mapping between ICD-10 codes and Combined ICD codes for
PheWAS analysis eTable 3 Hazard ratios (HRs) with 95% confidence intervals (CIs) of cardiovascular disease among patients with common psychiatry disorders compared to matched unexposed individuals, by different psychiatry disorders eTable 4 PheWAS results of hazard ratios (HRs) with 95% confidence intervals (CIs) of different individual cardiovascular diseases among patients with common psychiatry disorders compared to matched unexposed individuals eTable 5 Odds ratios (ORs) with 95% confidence intervals (CIs) for the significant pairs of cardiovascular diseases following a diagnosis of common psychiatry disorders eTable 6 Hazard ratios (HRs) with 95% confidence intervals (CIs) of primary diagnosis of cardiovascular disease among patients with common psychiatry disorders compared to matched unexposed individuals eTable 7 Hazard ratios (HRs) with 95% confidence intervals (CIs) of primary diagnosis of cardiovascular disease among patients with common psychiatry disorders compared to matched unexposed individuals, by different disease susceptibility eTable 8 Odds ratios (ORs) with 95% confidence intervals (CIs) for the significant pairs of primary diagnosis of cardiovascular diseases following a diagnosis of common psychiatry disorders eFigure 1 Hazard ratios (95% CIs) of cardiovascular disease among patients with common psychiatry disorders compared with their matched unexposed individuals, stratified by time of follow-up eFigure 2 Flow chart of identifying trajectory progression of cardiovascular disease following a diagnosis of common psychiatry disorder eFigure 3 Relative risks of six subtypes of cardiovascular diseases among patients with common psychiatry disorders compared with their matched unexposed individuals, stratified by disease susceptibility, by time of follow-up (<= 6 or > 6 months) eFigure 4 Flow chart of identifying trajectory progression of primary diagnosis of cardiovascular disease following a diagnosis of common psychiatry disorders eFigure 5 Trajectory progression of primary diagnosis of cardiovascular disease following a diagnosis of common psychiatry disorders . CC-BY 4.0 International license It is made available under a perpetuity.
is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted September 29, 2021. ; is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

eTable 1 International Classification of Diseases (ICD) codes and primary care codes for diseases identifications
The copyright holder for this this version posted September 29, 2021. ; eFigure 1 Hazard ratios (95% CIs) of cardiovascular disease among patients with common psychiatry disorders compared with their matched unexposed individuals, stratified by time of follow-up Abbreviation: CI: confidence interval. The X axis shows the different follow-up period: ≤3, 3-6, 6-12, 12-18, 18-24, 24-60, 60-120, 120-240, and >240 months follow-up. The Y axis shows the significant hazard ratios of cardiovascular disease among patients with common psychiatry disorders compared with their matched unexposed individuals, derived from Cox models adjusted for sex, birth year and ethnicity.
. CC-BY 4.0 International license It is made available under a perpetuity.
is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted September 29, 2021. ; https://doi.org/10.1101/2021.09.28.21264283 doi: medRxiv preprint eFigure 2 Flow chart of identifying trajectory progression of cardiovascular disease following a diagnosis of common psychiatry disorders . CC-BY 4.0 International license It is made available under a perpetuity.
is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted September 29, 2021. ; https://doi.org/10.1101/2021.09.28.21264283 doi: medRxiv preprint eFigure 3 Relative risks of six subtypes of cardiovascular diseases among patients with common psychiatry disorders compared with their matched unexposed individuals, stratified by disease susceptibility, by time of follow-up (<= 6 or > 6 months) Abbreviation: N: number of cases of specific cardiovascular disease; I: incidence.
Cox models were stratified by family history of cardiovascular disease (yes or no) and level of cardiovascular disease PRS (low, middle, and high) and adjusted for sex, birth year, educational level, ethnicity, smoke, TDI, BMI, history of other psychiatry disorders, and CCI score. For subtypes of cardiovascular diseases with less than 20 cases in patients with common psychiatry disorders, hazard rations were derived from Cox regression models and adjusted for sex, birth year. eFigure 4 Flow chart of identifying trajectory progression of primary diagnosis of cardiovascular disease following a diagnosis of common psychiatry disorders . CC-BY 4.0 International license It is made available under a perpetuity.
is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted September 29, 2021. ; https://doi.org/10.1101/2021.09.28.21264283 doi: medRxiv preprint eFigure 5 Trajectory progression of primary diagnosis of cardiovascular disease following a diagnosis of common psychiatry disorders This figure illustrates following trajectory progression of primary diagnosis of cardiovascular disease identified in our analysis. The combined cardiovascular diseases are shown within the circle. The color of the circle represents the hazard ratios of this cardiovascular disease when comparing patients with common psychiatry disorders to matched unexposed individuals. The number above the arrow connecting two circles corresponds to the number of pairs of two cardiovascular disease events among patients with common psychiatry disorders. The color of the arrows indicates the odds ratio of the sequential association between the two cardiovascular disease events among patients with common psychiatry disorders.
. CC-BY 4.0 International license It is made available under a perpetuity.
is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted September 29, 2021. ;https://doi.org/10.1101https://doi.org/10. /2021