Venous Thromboembolism Polygenic Risk Score Associates with Pulmonary Hypertension in the UK Biobank

Chronic thromboembolic pulmonary hypertension (CTEPH) is a form of pulmonary hypertension (PH) driven by incomplete resolution of pulmonary embolism (PE) and adverse vascular remodeling¹. A recent review of epidemiologic data suggests that CTEPH is relatively rare, with an incidence in the USA and Europe of 3–5 cases/100,000 individuals per year, and a calculated 4–7 cases per million². However, CTEPH is likely underdiagnosed due to the subtlety of symptoms early in the disease, as well its development in individuals with no known history of PE; a study utilizing estimated incidence of PE and resultant CTEPH, and data from pulmonary endarterectomies in the USA and Europe, indicates only 14% of cases are diagnosed². The familial enrichment of CTEPH cases and heightened venous thromboembolism (VTE) risk among relatives of individuals with CTEPH implies a heritable component³. Large-scale biobanks have enabled interrogation of common genetic variants that increase disease risk, including those which strongly associate with VTE⁴. Here, we hypothesized that genetic determinants of VTE might also promote downstream sequelae, such as PH, in the general population.

Leveraging a VTE GWAS previously conducted in the Million Veteran Program (MVP) biobank, we created a new polygenic risk score (PRS) for VTE⁴. The PRS was generated utilizing PRS-CS, a prediction algorithm based on a high dimensional Bayesian regression framework that utilizes a continuous shrinkage prior on SNP effect sizes⁵. European linkage disequilibrium was used as a reference, and HapMap SNPs with INFO>0.3 were included in the analysis. We validated the association of this new PRS comprising 1,067,348 variants with incident and prevalent VTE in the UK Biobank (UKB) (16,736 cases, 381,113 controls) using multivariable logistic regression adjusting for baseline age, sex, array, diabetes mellitus type 2, smoking history, body-mass index (BMI), and the first 10 principal components⁴. UKB access was granted under application 7089 and is accessible by application (https://bbams.ndph.ox.ac.uk/ams/). Secondary use was permitted by the Massachusetts General Brigham IRB. Prevalent and incident disease were defined by ICD code at enrollment or in the follow-up period, respectively; the median follow-up was 11.2 years.

The overall study population (N=397,836) was 54% female, with a mean age of 56.6 (standard deviation (SD) 8.0) years, and a mean BMI of 27.4 (SD 4.8) kg/m². A one SD increase in polygenic susceptibility to VTE was associated with 1.4 odds of VTE (95% CI 1.40–1.42, p< 2×10⁻¹⁶) (Figure 1A). We next examined whether the VTE PRS was associated with prevalent and incident PH (ICD-9 4160 and ICD-10 I27.0, I27.2). The VTE PRS was associated with PH with an OR of 1.07 (1,442 cases, 396,407 controls, 95% CI 1.01–1.12, p =0.005) (Figure 1A). Increased prevalence of PH was observed with higher VTE PRS scores (Figure 1B). Individuals in the top 10% of the normally distributed VTE PRS were at the highest risk for PH compared to remaining 90%, with an OR of 1.25 (95% CI 1.07–1.45). The prevalence of PH among those in the top 10th percentile of VTE PRS was 460 cases per million, far exceeding the estimated prevalence of CTEPH of 4–7 cases per million². When accounting for the diagnosis of VTE in the multivariate logistic regression model, the association of the VTE PRS with PH was no longer significant (OR 1.03, 95% CI 0.98–1.09, p=0.18) (Figure 1A). The VTE PRS was not significantly associated with the subset of PH cases with primary pulmonary hypertension (523 cases, 397,313 controls; OR 1.04, 95% CI 0.98–1.16, p=0.08), but was significantly associated with secondary PH (926 cases, 396,354 controls; OR 1.08, 95% CI 1.03 −1.16, p=0.005) (Figure 1A); primary and secondary PH are currently known as WHO Group I PH, and II-V PH, respectively. However, effect estimates between these groups did not differ statistically (p(heterogeneity)=0.55). CTEPH-specific coding was not collected by the UKB. There were no significant differences in age, sex, and BMI between groups.

Figure 1:

(A) Results of multivariable logistic regression testing association of phenotypes with VTE PRS, expressed in odds ratios per standard deviation increase in VTE PRS and 95% confidence intervals. VTE = venous thromboembolism, PH = pulmonary hypertension, PRS = polygenic risk score. * = p<0.05. (B) Percent prevalence of PH by VTE PRS percentile.

Our results show that there are shared genetic determinants between VTE and PH. A potential explanation for this finding is that a larger fraction of PH has contributions from a CTEPH spectrum than previously appreciated. The association is not present when accounting for VTE status, suggesting that a high VTE PRS promotes PH via clinically manifest VTE. Individuals with a VTE PRS score in the top 10th percentile appear to be at markedly higher risk of PH. The VTE PRS may thus be a useful tool to stratify risk for CTEPH, including during evaluation of PH etiology. Further parsing of PH phenotype would be optimal but is limited by low disease prevalence in the UKB. Because PH is a heterogeneous disease, reliance on an electronic health record based phenotypic definition biases the association between the VTE PRS and PH toward the null. Our findings reveal shared genetic susceptibility to VTE and PH, and further examination of these findings could allow finer targeting of the disease process.

Nonstandard Abbreviations and Acronyms:

CTEPH

chronic thromboembolic pulmonary hypertension

PH

pulmonary hypertension

PE

pulmonary embolism

VTE

venous thromboembolism

MVP

Million Veteran Program

GWAS

genome wide association study

PRS

polygenic risk score

SNP

single nucleotide polymorphism

UKB

UK Biobank

BMI

body mass index

IRB

institutional review board

ICD

international classification of diseases

SD

standard deviation

CI

confidence interval

OR

odds ratio

WHO

world health organization