Microbiome-based risk prediction in incident heart failure: a community challenge

Heart failure (HF) is a major public health problem. Early identification of at-risk individuals could allow for interventions that reduce morbidity or mortality. The community-based FINRISK Microbiome DREAM challenge (synapse.org/finrisk) evaluated the use of machine learning approaches on shotgun metagenomics data obtained from fecal samples to predict incident HF risk over 15 years in a population cohort of 7231 Finnish adults (FINRISK 2002, n=559 incident HF cases). Challenge participants used synthetic data for model training and testing. Final models submitted by seven teams were evaluated in the real data. The two highest-scoring models were both based on Cox regression but used different feature selection approaches. We aggregated their predictions to create an ensemble model. Additionally, we refined the models after the DREAM challenge by eliminating phylum information. Models were also evaluated at intermediate timepoints and they predicted 10-year incident HF more accurately than models for 5- or 15-year incidence. We found that bacterial species, especially those linked to inflammation, are predictive of incident HF. This highlights the role of the gut microbiome as a potential driver of inflammation in HF pathophysiology. Our results provide insights into potential modeling strategies of microbiome data in prospective cohort studies. Overall, this study provides evidence that incorporating microbiome information into incident risk models can provide important biological insights into the pathogenesis of HF.

A network medicine approach to study comorbidities in heart failure with preserved ejection fraction

BACKGROUND

Comorbidities are expected to impact the pathophysiology of heart failure (HF) with preserved ejection fraction (HFpEF). However, comorbidity profiles are usually reduced to a few comorbid disorders. Systems medicine approaches can model phenome-wide comorbidity profiles to improve our understanding of HFpEF and infer associated genetic profiles.

METHODS

We retrospectively explored 569 comorbidities in 29,047 HF patients, including 8062 HFpEF and 6585 HF with reduced ejection fraction (HFrEF) patients from a German university hospital. We assessed differences in comorbidity profiles between HF subtypes via multiple correspondence analysis. Then, we used machine learning classifiers to identify distinctive comorbidity profiles of HFpEF and HFrEF patients. Moreover, we built a comorbidity network (HFnet) to identify the main disease clusters that summarized the phenome-wide comorbidity. Lastly, we predicted novel gene candidates for HFpEF by linking the HFnet to a multilayer gene network, integrating multiple databases. To corroborate HFpEF candidate genes, we collected transcriptomic data in a murine HFpEF model. We compared predicted genes with the murine disease signature as well as with the literature.

RESULTS

We found a high degree of variance between the comorbidity profiles of HFpEF and HFrEF, while each was more similar to HFmrEF. The comorbidities present in HFpEF patients were more diverse than those in HFrEF and included neoplastic, osteologic and rheumatoid disorders. Disease communities in the HFnet captured important comorbidity concepts of HF patients which could be assigned to HF subtypes, age groups, and sex. Based on the HFpEF comorbidity profile, we predicted and recovered gene candidates, including genes involved in fibrosis (COL3A1, LOX, SMAD9, PTHL), hypertrophy (GATA5, MYH7), oxidative stress (NOS1, GSST1, XDH), and endoplasmic reticulum stress (ATF6). Finally, predicted genes were significantly overrepresented in the murine transcriptomic disease signature providing additional plausibility for their relevance.

CONCLUSIONS

We applied systems medicine concepts to analyze comorbidity profiles in a HF patient cohort. We were able to identify disease clusters that helped to characterize HF patients. We derived a distinct comorbidity profile for HFpEF, which was leveraged to suggest novel candidate genes via network propagation. The identification of distinctive comorbidity profiles and candidate genes from routine clinical data provides insights that may be leveraged to improve diagnosis and identify treatment targets for HFpEF patients.

A novel scatterplot-based method to detect copy number variation (CNV)

Objective: Most methods to detect copy number variation (CNV) have high false positive rates, especially for small CNVs and in real-life samples from clinical studies. In this study, we explored a novel scatterplot-based method to detect CNVs in microarray samples. Methods: Illumina SNP microarray data from 13,254 individuals were analyzed with scatterplots and by PennCNV. The data were analyzed without the prior exclusion of low-quality samples. For CNV scatterplot visualization, the median signal intensity of all SNPs located within a CNV region was plotted against the median signal intensity of the flanking genomic region. Since CNV causes loss or gain of signal intensities, carriers of different CNV alleles pop up in clusters. Moreover, SNPs within a deletion are not heterozygous, whereas heterozygous SNPs within a duplication show typical 1:2 signal distribution between the alleles. Scatterplot-based CNV calls were compared with standard results of PennCNV analysis. All discordant calls as well as a random selection of 100 concordant calls were individually analyzed by visual inspection after noise-reduction. Results: An algorithm for the automated scatterplot visualization of CNVs was developed and used to analyze six known CNV regions. Use of scatterplots and PennCNV yielded 1019 concordant and 108 discordant CNV calls. All concordant calls were evaluated as true CNV-findings. Among the 108 discordant calls, 7 were false positive findings by the scatterplot method, 80 were PennCNV false positives, and 21 were true CNVs detected by the scatterplot method, but missed by PennCNV (i.e., false negative findings). Conclusion: CNV visualization by scatterplots allows for a reliable and rapid detection of CNVs in large studies. This novel method may thus be used both to confirm the results of genome-wide CNV detection software and to identify known CNVs in hitherto untyped samples.

Dyslipidemia, inflammation, calcification, and adiposity in aortic stenosis: a genome-wide study

AIMS

Although highly heritable, the genetic etiology of calcific aortic stenosis (AS) remains incompletely understood. The aim of this study was to discover novel genetic contributors to AS and to integrate functional, expression, and cross-phenotype data to identify mechanisms of AS.

METHODS AND RESULTS

A genome-wide meta-analysis of 11.6 million variants in 10 cohorts involving 653 867 European ancestry participants (13 765 cases) was performed. Seventeen loci were associated with AS at P ≤ 5 × 10-8, of which 15 replicated in an independent cohort of 90 828 participants (7111 cases), including CELSR2-SORT1, NLRP6, and SMC2. A genetic risk score comprised of the index variants was associated with AS [odds ratio (OR) per standard deviation, 1.31; 95% confidence interval (CI), 1.26-1.35; P = 2.7 × 10-51] and aortic valve calcium (OR per standard deviation, 1.22; 95% CI, 1.08-1.37; P = 1.4 × 10-3), after adjustment for known risk factors. A phenome-wide association study indicated multiple associations with coronary artery disease, apolipoprotein B, and triglycerides. Mendelian randomization supported a causal role for apolipoprotein B-containing lipoprotein particles in AS (OR per g/L of apolipoprotein B, 3.85; 95% CI, 2.90-5.12; P = 2.1 × 10-20) and replicated previous findings of causality for lipoprotein(a) (OR per natural logarithm, 1.20; 95% CI, 1.17-1.23; P = 4.8 × 10-73) and body mass index (OR per kg/m2, 1.07; 95% CI, 1.05-1.9; P = 1.9 × 10-12). Colocalization analyses using the GTEx database identified a role for differential expression of the genes LPA, SORT1, ACTR2, NOTCH4, IL6R, and FADS.

CONCLUSION

Dyslipidemia, inflammation, calcification, and adiposity play important roles in the etiology of AS, implicating novel treatments and prevention strategies.

Exploration of an alternative to body mass index to characterize the relationship between height and weight for prediction of metabolic phenotypes and cardiovascular outcomes

OBJECTIVE

Body mass index (BMI) is the most commonly used predictor of weight-related comorbidities and outcomes. However, the presumed relationship between height and weight intrinsic to BMI may introduce bias with respect to prediction of clinical outcomes. A series of analyses comparing the performance of models representing weight and height as separate interacting variables to models using BMI were performed using Vanderbilt University Medical Center's deidentified electronic health records and landmark methodology.

METHODS

Use of BMI or height-weight interaction in prediction models for established weight-related cardiometabolic traits and metabolic syndrome was evaluated. Specifically, prediction models for hypertension, diabetes mellitus, low high-density lipoprotein, and elevated triglycerides, atrial fibrillation, coronary artery disease, heart failure, and peripheral artery disease were developed. Model performance was evaluated using likelihood ratio, R 2, and Somers' Dxy rank correlation. Differences in model predictions were visualized using heat maps.

RESULTS

Compared to BMI, the maximally flexible height-weight interaction model demonstrated improved prediction, higher likelihood ratio, R 2, and Somers' Dxy rank correlation, for event-free probability for all outcomes. The degree of improvement to the prediction model differed based on the outcome and across the height and weight range.

CONCLUSIONS

Because alternative measures of body composition such as waist-to-hip ratio are not routinely collected in the clinic clinical risk models quantifying risk based on height and weight measurements alone are essential to improve practice. Compared to BMI, modeling height and weight as independent, interacting variables results in less bias and improved predictive accuracy for all tested traits. Considering an individual's height and weight opposed to BMI is a better method for quantifying individual disease risk.

Consensus Transcriptional Landscape of Human End-Stage Heart Failure

Background Transcriptomic studies have contributed to fundamental knowledge of myocardial remodeling in human heart failure (HF). However, the key HF genes reported are often inconsistent between studies, and systematic efforts to integrate evidence from multiple patient cohorts are lacking. Here, we aimed to provide a framework for comprehensive comparison and analysis of publicly available data sets resulting in an unbiased consensus transcriptional signature of human end-stage HF. Methods and Results We curated and uniformly processed 16 public transcriptomic studies of left ventricular samples from 263 healthy and 653 failing human hearts. First, we evaluated the degree of consistency between studies by using linear classifiers and overrepresentation analysis. Then, we meta-analyzed the deregulation of 14 041 genes to extract a consensus signature of HF. Finally, to functionally characterize this signature, we estimated the activities of 343 transcription factors, 14 signaling pathways, and 182 micro RNAs, as well as the enrichment of 5998 biological processes. Machine learning approaches revealed conserved disease patterns across all studies independent of technical differences. These consistent molecular changes were prioritized with a meta-analysis, functionally characterized and validated on external data. We provide all results in a free public resource (https://saezlab.shinyapps.io/reheat/) and exemplified usage by deciphering fetal gene reprogramming and tracing the potential myocardial origin of the plasma proteome markers in patients with HF. Conclusions Even though technical and sampling variability confound the identification of differentially expressed genes in individual studies, we demonstrated that coordinated molecular responses during end-stage HF are conserved. The presented resource is crucial to complement findings in independent studies and decipher fundamental changes in failing myocardium.

Association of FADS1/2 Locus Variants and Polyunsaturated Fatty Acids With Aortic Stenosis

Question

Can genetic analysis identify additional causes of aortic stenosis?

Findings

In this genome-wide association study of 44 703 participants, each copy of a FADS1/2 (fatty acid desaturase) genetic variant was associated with a 13% decrease in the odds of aortic stenosis. Results of a meta-analysis with 7 replication cohorts showed genome-wide significance, with biomarker and mendelian randomization analyses implicating elevated ω-6 fatty acid levels as having a potentially causal association with aortic valve calcium and aortic stenosis.

Meaning

These findings demonstrate that the FADS1/2 locus and fatty acid biosynthesis are associated with aortic stenosis and should be examined further for their potential as therapeutic targets.

Importance

Aortic stenosis (AS) has no approved medical treatment. Identifying etiological pathways for AS could identify pharmacological targets.

Objective

To identify novel genetic loci and pathways associated with AS.

Design, Setting, and Participants

This genome-wide association study used a case-control design to evaluate 44 703 participants (3469 cases of AS) of self-reported European ancestry from the Genetic Epidemiology Research on Adult Health and Aging (GERA) cohort (from January 1, 1996, to December 31, 2015). Replication was performed in 7 other cohorts totaling 256 926 participants (5926 cases of AS), with additional analyses performed in 6942 participants from the Cohorts for Heart and Aging Research in Genomic Epidemiology (CHARGE) Consortium. Follow-up biomarker analyses with aortic valve calcium (AVC) were also performed. Data were analyzed from May 1, 2017, to December 5, 2019.

Exposures

Genetic variants (615 643 variants) and polyunsaturated fatty acids (ω-6 and ω-3) measured in blood samples.

Main Outcomes and Measures

Aortic stenosis and aortic valve replacement defined by electronic health records, surgical records, or echocardiography and the presence of AVC measured by computed tomography.

Results

The mean (SD) age of the 44 703 GERA participants was 69.7 (8.4) years, and 22 019 (49.3%) were men. The rs174547 variant at the FADS1/2 locus was associated with AS (odds ratio [OR] per C allele, 0.88; 95% CI, 0.83-0.93; P = 3.0 × 10⁻⁶), with genome-wide significance after meta-analysis with 7 replication cohorts totaling 312 118 individuals (9395 cases of AS) (OR, 0.91; 95% CI, 0.88-0.94; P = 2.5 × 10⁻⁸). A consistent association with AVC was also observed (OR, 0.91; 95% CI, 0.83-0.99; P = .03). A higher ratio of arachidonic acid to linoleic acid was associated with AVC (OR per SD of the natural logarithm, 1.19; 95% CI, 1.09-1.30; P = 6.6 × 10⁻⁵). In mendelian randomization, increased FADS1 liver expression and arachidonic acid were associated with AS (OR per unit of normalized expression, 1.31 [95% CI, 1.17-1.48; P = 7.4 × 10⁻⁶]; OR per 5–percentage point increase in arachidonic acid for AVC, 1.23 [95% CI, 1.01-1.49; P = .04]; OR per 5–percentage point increase in arachidonic acid for AS, 1.08 [95% CI, 1.04-1.13; P = 4.1 × 10⁻⁴]).

Conclusions and Relevance

Variation at the FADS1/2 locus was associated with AS and AVC. Findings from biomarker measurements and mendelian randomization appear to link ω-6 fatty acid biosynthesis to AS, which may represent a therapeutic target.

Big Data Approaches in Heart Failure Research

PURPOSE OF REVIEW

The goal of this review is to summarize the state of big data analyses in the study of heart failure (HF). We discuss the use of big data in the HF space, focusing on "omics" and clinical data. We address some limitations of this data, as well as their future potential.

RECENT FINDINGS

Omics are providing insight into plasmal and myocardial molecular profiles in HF patients. The introduction of single cell and spatial technologies is a major advance that will reshape our understanding of cell heterogeneity and function as well as tissue architecture. Clinical data analysis focuses on HF phenotyping and prognostic modeling. Big data approaches are increasingly common in HF research. The use of methods designed for big data, such as machine learning, may help elucidate the biology underlying HF. However, important challenges remain in the translation of this knowledge into improvements in clinical care.

BMI Is Causally Associated With Pulmonary Artery Pressure But Not Hemodynamic Evidence of Pulmonary Vascular Remodeling

BACKGROUND

There is an unclear relationship of obesity to the pathogenesis and severity of pulmonary arterial hypertension (PAH) and pulmonary venous hypertension (PVH).

RESEARCH QUESTION

Is BMI casually associated with pulmonary artery pressure (PAP) and/or markers of pulmonary vascular remodeling?

STUDY DESIGN AND METHODS

The study design was a two-sample inverse-variance weighted Mendelian randomization. We constructed two BMI genetic risk scores from genome-wide association study summary data and deployed them in nonoverlapping cohorts of subjects referred for right heart catheterization (RHC) or echocardiography. A BMI highly polygenic risk score (hpGRS) optimally powered to detect shared genetic architecture of obesity with other traits was tested for association with RHC parameters including markers of pulmonary vascular remodeling. A BMI strict genetic risk score (sGRS) composed of high-confidence genetic variants was used for Mendelian randomization analyses to assess if higher BMI causes higher PAP.

RESULTS

Among all subjects, both directly measured BMI and hpGRS were positively associated with pulmonary arterial pressures but not markers of pulmonary vascular remodeling. Categorical analyses revealed BMI and hpGRS were associated with PVH but not PAH. Mendelian randomization of the sGRS supported that higher BMI is causal of higher systolic pulmonary artery pressure (sPAP). Sensitivity analyses showed sPAP-BMI sGRS relationship was preserved when either individuals with PAH or PVH were excluded. In the echocardiographic cohort, BMI and hpGRS were positively associated with estimated PAP and markers of left heart remodeling.

INTERPRETATION

BMI is a modifier of pulmonary hypertension severity in both PAH and PVH but is only involved in the pathogenesis of PVH.

The polygenic architecture of left ventricular mass mirrors the clinical epidemiology

Left ventricular (LV) mass is a prognostic biomarker for incident heart disease and all-cause mortality. Large-scale genome-wide association studies have identified few SNPs associated with LV mass. We hypothesized that a polygenic discovery approach using LV mass measurements made in a clinical population would identify risk factors and diseases associated with adverse LV remodeling. We developed a polygenic single nucleotide polymorphism-based predictor of LV mass in 7,601 individuals with LV mass measurements made during routine clinical care. We tested for associations between this predictor and 894 clinical diagnoses measured in 58,838 unrelated genotyped individuals. There were 29 clinical phenotypes associated with the LV mass genetic predictor at FDR q < 0.05. Genetically predicted higher LV mass was associated with modifiable cardiac risk factors, diagnoses related to organ dysfunction and conditions associated with abnormal cardiac structure including heart failure and atrial fibrillation. Secondary analyses using polygenic predictors confirmed a significant association between higher LV mass and body mass index and, in men, associations with coronary atherosclerosis and systolic blood pressure. In summary, these analyses show that LV mass-associated genetic variability associates with diagnoses of cardiac diseases and with modifiable risk factors which contribute to these diseases.

Unbiased Phenome-Wide Association Studies of Red Cell Distribution Width Identifies Key Associations with Pulmonary Hypertension

Rationale: Red cell distribution width (RDW) is a prognostic factor in many diseases; however, its clinical utility remains limited because the relative value of RDW as a biomarker across disease states has not been established. Objectives: To establish an unbiased RDW disease hierarchy to guide the clinical use of RDW and to assess its relationship to cardiovascular hemodynamic and structural parameters. Methods: We performed phenome-wide association studies for RDW in discovery and replication cohorts derived from a deidentified electronic health record in nonanemic individuals. RDW values obtained within 30 days of echocardiogram or right heart catheterization were tested for association with structural and hemodynamic variables. Results: RDW was associated with 263 phenotypes in both men and women in the discovery cohort (n = 121,530), 48 of which replicated in an independent cohort (n = 2,039). The strongest associations were observed with pulmonary arterial hypertension (odds ratio [OR], 2.1; 95% confidence interval [CI], 1.9-2.3), chronic pulmonary heart disease (OR, 2.0; 95% CI, 1.9-2.2), and congestive heart failure (OR, 1.9; 95% CI, 1.8-2.0); P < 1 × 10-74 for all. By echocardiography, RDW was higher in the setting of right ventricular dysfunction than left ventricular dysfunction (P < 0.001). Measured invasively, mean pulmonary arterial pressure was associated with RDW (21 vs. 33 mm Hg at 25th vs. 75th percentile RDW; P < 1 × 10-7) and remained strongly significant even when controlling for mean pulmonary capillary wedge pressure (21 vs. 29 mm Hg at 25th vs. 75th percentile RDW; P < 1 × 10-7). Conclusions: Among 1,364 coded medical conditions, increased RDW was strongly associated with pulmonary hypertension and heart failure. Hemodynamic and echocardiographic phenotyping confirmed these associations and underscored that the most clinically relevant phenotype associated with RDW was pulmonary hypertension. These hypothesis-generating findings highlight the potential shared pathophysiology of pulmonary hypertension and elevated RDW. Elevated RDW in the absence of anemia should alert clinicians to the potential for underlying cardiopulmonary disease.

Phenotypic Refinement of Heart Failure in a National Biobank Facilitates Genetic Discovery

BACKGROUND

Heart failure (HF) is a morbid and heritable disorder for which the biological mechanisms are incompletely understood. We therefore examined genetic associations with HF in a large national biobank, and assessed whether refined phenotypic classification would facilitate genetic discovery.

METHODS

We defined all-cause HF among 488 010 participants from the UK Biobank and performed a genome-wide association analysis. We refined the HF phenotype by classifying individuals with left ventricular dysfunction and without coronary artery disease as having nonischemic cardiomyopathy (NICM), and repeated a genetic association analysis. We then pursued replication of lead HF and NICM variants in independent cohorts, and performed adjusted association analyses to assess whether identified genetic associations were mediated through clinical HF risk factors. In addition, we tested rare, loss-of-function mutations in 24 known dilated cardiomyopathy genes for association with HF and NICM. Finally, we examined associations between lead variants and left ventricular structure and function among individuals without HF using cardiac magnetic resonance imaging (n=4158) and echocardiographic data (n=30 201).

RESULTS

We identified 7382 participants with all-cause HF in the UK Biobank. Genome-wide association analysis of all-cause HF identified several suggestive loci (P<1×10-6), the majority linked to upstream HF risk factors, ie, coronary artery disease (CDKN2B-AS1 and MAP3K7CL) and atrial fibrillation (PITX2). Refining the HF phenotype yielded a subset of 2038 NICM cases. In contrast to all-cause HF, genetic analysis of NICM revealed suggestive loci that have been implicated in dilated cardiomyopathy (BAG3, CLCNKA-ZBTB17). Dilated cardiomyopathy signals arising from our NICM analysis replicated in independent cohorts, persisted after HF risk factor adjustment, and were associated with indices of left ventricular dysfunction in individuals without clinical HF. In addition, analyses of loss-of-function variants implicated BAG3 as a disease susceptibility gene for NICM (loss-of-function variant carrier frequency=0.01%; odds ratio,12.03; P=3.62×10-5).

CONCLUSIONS

We found several distinct genetic mechanisms of all-cause HF in a national biobank that reflect well-known HF risk factors. Phenotypic refinement to a NICM subtype appeared to facilitate the discovery of genetic signals that act independently of clinical HF risk factors and that are associated with subclinical left ventricular dysfunction.

Relationship between very low low-density lipoprotein cholesterol concentrations not due to statin therapy and risk of type 2 diabetes: A US-based cross-sectional observational study using electronic health records

BACKGROUND

Observations from statin clinical trials and from Mendelian randomization studies suggest that low low-density lipoprotein cholesterol (LDL-C) concentrations may be associated with increased risk of type 2 diabetes mellitus (T2DM). Despite the findings from statin clinical trials and genetic studies, there is little direct evidence implicating low LDL-C concentrations in increased risk of T2DM.

METHODS AND FINDINGS

We used de-identified electronic health records (EHRs) at Vanderbilt University Medical Center to compare the risk of T2DM in a cross-sectional study among individuals with very low (≤60 mg/dl, N = 8,943) and normal (90-130 mg/dl, N = 71,343) LDL-C levels calculated using the Friedewald formula. LDL-C levels associated with statin use, hospitalization, or a serum albumin level < 3 g/dl were excluded. We used a 2-phase approach: in 1/3 of the sample (discovery) we used T2DM phenome-wide association study codes (phecodes) to identify cases and controls, and in the remaining 2/3 (validation) we identified T2DM cases and controls using a validated algorithm. The analysis plan for the validation phase was constructed at the time of the design of that component of the study. The prevalence of T2DM in the very low and normal LDL-C groups was compared using logistic regression with adjustment for age, race, sex, body mass index (BMI), high-density lipoprotein cholesterol, triglycerides, and duration of care. Secondary analyses included prespecified stratification by sex, race, BMI, and LDL-C level. In the discovery cohort, phecodes related to T2DM were significantly more frequent in the very low LDL-C group. In the validation cohort (N = 33,039 after applying the T2DM algorithm to identify cases and controls), the risk of T2DM was increased in the very low compared to normal LDL-C group (odds ratio [OR] 2.06, 95% CI 1.80-2.37; P < 2 × 10-16). The findings remained significant in sensitivity analyses. The association between low LDL-C levels and T2DM was significant in males (OR 2.43, 95% CI 2.00-2.95; P < 2 × 10-16) and females (OR 1.74, 95% CI 1.42-2.12; P = 6.88 × 10-8); in normal weight (OR 2.18, 95% CI 1.59-2.98; P = 1.1× 10-6), overweight (OR 2.17, 95% CI 1.65-2.83; P = 1.73× 10-8), and obese (OR 2.00, 95% CI 1.65-2.41; P = 8 × 10-13) categories; and in individuals with LDL-C < 40 mg/dl (OR 2.31, 95% CI 1.71-3.10; P = 3.01× 10-8) and LDL-C 40-60 mg/dl (OR 1.99, 95% CI 1.71-2.32; P < 2.0× 10-16). The association was significant in individuals of European ancestry (OR 2.67, 95% CI 2.25-3.17; P < 2 × 10-16) but not in those of African ancestry (OR 1.09, 95% CI 0.81-1.46; P = 0.56). A limitation was that we only compared groups with very low and normal LDL-C levels; also, since this was not an inception cohort, we cannot exclude the possibility of reverse causation.

CONCLUSIONS

Very low LDL-C concentrations occurring in the absence of statin treatment were significantly associated with T2DM risk in a large EHR population; this increased risk was present in both sexes and all BMI categories, and in individuals of European ancestry but not of African ancestry. Longitudinal cohort studies to assess the relationship between very low LDL-C levels not associated with lipid-lowering therapy and risk of developing T2DM will be important.

Abstract 4229: Preliminary results from the Pharmacogenetics Ovarian Cancer Knowledge to Individualize Treatment (POCKIT) study

Introduction: With an overall five-year survival rate of only 46%, ovarian cancer is the most lethal gynecologic malignancy. Treatment includes surgical cytoreduction (debulking) followed by chemotherapy with platinum and taxane agents, and prognosis generally depends upon clinical characteristics, such as stage of disease at diagnosis, success of surgical debulking, and histologic subtype. However, even among women with comparable clinical characteristics, there can be variation in survival. Part of this variability may be due to inherited genetic variation in genes related to the absorption, distribution, metabolism, and excretion (ADME) of pharmacologic agents commonly used to treat ovarian cancer.

Approach: To test the hypothesis that pharmacogenetic variants influence ovarian cancer prognosis, we assembled a clinical cohort of confirmed epithelial ovarian cancer cases from electronic medical records (EMR) at the Vanderbilt University Medical Center (VUMC) with banked DNA samples available. Clinical characteristics were abstracted from EMR using natural language processing (NLP)-assisted EMR review and a REDCap data collection instrument. Genotyping was conducted on the Sequenom iPLEX ADME PGx panel at the Vanderbilt Technologies for Advanced Genomics (VANTAGE) core facility. Associations with overall survival were evaluated using multivariable proportional hazards regression.

Results: We identified a total of 391 epithelial ovarian cancer cases with banked DNA from VUMC EHR. Clinical characteristic abstracted by NLP followed expected distributions; the majority of cases were Caucasian (87%), with serous histology (63%), late-stage (62%), high-grade (60%) disease. Most common treatments included surgical cytoreduction (93%) and chemotherapy with a platinum (83%) and/or taxane agent (82%) agent. DNA was successfully pulled, plated, and genotyped for 327 cases (81.3%) for 73 common ADME variants in 30 genes. To prevent population stratification, genetic analyses were restricted to 287 Caucasians, where five nominally significant overall survival associations were identified: ABCB1 rs1045642, ABCC2 rs2273697, CYP2A6 rs1801272, CYP2E1 rs2070673, and SLCO2B1 rs2306168. Additional analyses, including for gene and drug scores, are currently under way.

Conclusions: Individual variation in ADME genes may contribute to variation in ovarian cancer survival. Future steps include testing associations for replication, and evaluation of response to treatment. In addition, this research demonstrates that EMR-based study populations, in concert with linked biorepositories, can facilitate research on ovarian cancer.

Citation Format: Ayush Giri, Rebecca T. Levinson, Spencer Keene, Gwendolyn Holman, Stacy D. Smith, Leshaun Clayton, Whitney Lovett, Samantha P. Stansel, Malcolm-Robert Bringhurst Snyder, Jason T. Fromal, Gabriella D. Cozzi, Dineo Khabele, Alicia Beeghly-Fadiel. Preliminary results from the Pharmacogenetics Ovarian Cancer Knowledge to Individualize Treatment (POCKIT) study [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2018; 2018 Apr 14-18; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2018;78(13 Suppl):Abstract nr 4229.

Clinical Features Associated With Nascent Left Ventricular Diastolic Dysfunction in a Population Aged 40 to 55 Years

Diastolic dysfunction (DD), an abnormality in cardiac left ventricular (LV) chamber compliance, is associated with increased morbidity and mortality. Although DD has been extensively studied in older populations, co-morbidity patterns are less well characterized in middle-aged subjects. We screened 156,434 subjects with transthoracic echocardiogram reports available through Vanderbilt's electronic heath record and identified 6,612 subjects 40 to 55 years old with an LV ejection fraction ≥50% and diastolic function staging. We tested 452 incident and prevalent clinical diagnoses for associations with early-stage DD (n = 1,676) versus normal function. There were 44 co-morbid diagnoses associated with grade 1 DD including hypertension (odds ratio [OR] = 2.02, 95% confidence interval [CI] 1.78 to 2.28, p <5.3 × 10-29), type 2 diabetes (OR 1.96, 95% CI 1.68 to 2.29, p = 2.1 × 10-17), tachycardia (OR 1.38, 95% CI 0.53 to 2.19, p = 2.9 × 10-6), obesity (OR 1.76, 95% CI 1.51 to 2.06, p = 1.7 × 10-12), and clinical end points, including end-stage renal disease (OR 3.29, 95% CI 2.19 to 4.96, p = 1.2 × 10-8) and stroke (OR 1.5, 95% CI 1.12 to 2.02, p = 6.9 × 10-3). Among the 60 incident diagnoses associated with DD, heart failure with preserved ejection fraction (OR 4.63, 95% CI 3.39 to 6.32, p = 6.3 × 10-22) had the most significant association. Among subjects with normal diastolic function and blood pressure at baseline, a blood pressure measurement in the hypertensive range at the time of the second echocardiogram was associated with progression to stage 1 DD (p = 0.04). In conclusion, DD was common among subjects 40 to 55 years old and was associated with a heavy burden of co-morbid disease.

Variation in the α2A-adrenergic receptor gene and risk of gestational diabetes

AIM

Sympathetic activation suppresses insulin secretion via pancreatic ADRA2A.  Because sympathetic activity and insulin demand increase during pregnancy, we tested the hypothesis  that ADRA2A variants are associated with gestational diabetes (GDM).

PATIENTS & METHODS

Among Caucasian pregnant women without pre-existing diabetes, we genotyped 458 who had GDM and 1537 without GDM for seven ADRA2A variants.

RESULTS

rs1800038 (OR: 2.34; p = 0.020) and rs3750625 (OR: 1.56; p = 0.010) increased the risk of GDM, and rs11195418 decreased it (OR: 0.62; p = 0.025). The associations remained significant after adjustment for maternal age, maternal BMI, parity and a genetic risk score that included variants previously associated with Type 2 diabetes mellitus and GDM.

CONCLUSION

ADRA2A genetic variation contributes independently to the risk of GDM in Caucasian women.

Replication and fine-mapping of genetic predictors of lipid traits in African-Americans

Circulating lipid concentrations are among the strongest modifiable risk factors for coronary artery disease (CAD). Most genetic studies have focused on Caucasian populations with little information available for populations of African ancestry. Using a cohort of ~2800 African-Americans (AAs) from a biobank at Vanderbilt University (BioVU), we sought to trans-ethnically replicate genetic variants reported by the Global Lipids Genetics Consortium to be associated with lipid traits in Caucasians, followed by fine-mapping those loci using all available variants on the MetaboChip. In AAs, we replicated one of 56 SNPs for total cholesterol (TC) (rs6511720 in LDLR, P=2.15 × 10-8), one of 63 SNPs for high-density lipoprotein cholesterol (HDL-C) (rs3764261 in CETP, P=1.13 × 10-5), two of 46 SNPs for low-density lipoprotein cholesterol (LDL-C) (rs629301 in CELSR2/SORT1, P=1.11 × 10-5 and rs6511720 in LDLR, P=2.47 × 10-5) and one of 34 SNPs for TG (rs645040 in MSL2L1, P=4.29 × 10-4). Using all available variants on MetaboChip for fine mapping, we identified additional variants associated with TC (APOE), HDL-C (LPL and CETP) and LDL-C (APOE). Furthermore, we identified two loci significantly associated with non-HDL-C: APOE/APOC1/TOMM40 and PCSK9. In conclusion, the genetic architecture of lipid traits in AAs differs substantially from that in Caucasians and it remains poorly characterized.

A genetic risk score that includes common type 2 diabetes risk variants is associated with gestational diabetes

OBJECTIVE

Gestational diabetes (GDM) is characterized by maternal glucose intolerance that manifests during pregnancy. Because GDM resembles type 2 diabetes (T2DM), shared genetic predisposition is likely but has not been established. We tested the hypothesis that a genetic risk score (GRS) that included variants known to be associated with T2DM is associated with GDM.

STUDY DESIGN

We conducted a case-control study using the Vanderbilt Medical Center biobank (BioVU) and calculated a simple-count GRS using 34 variants previously associated with T2DM or fasting glucose in the general population, or with GDM or glucose intolerance in pregnancy. We assessed the association of the GRS with GDM adjusting for maternal age, parity, and body mass index (BMI) and calculated the area under the curve for the receiver-operating characteristic curve (c-statistic).

STUDY POPULATION

Among Caucasian women, we identified 458 cases of GDM and 1538 pregnant controls with normal glucose tolerance.

RESULTS

Cases of GDM had a higher number of risk alleles compared to controls (38.9±4.0 vs 37.4±4.0 risk alleles, P=1.6×10^-11 ). The GRS was significantly associated with GDM; the adjusted odds ratio associated with each additional risk allele was 1.10 (95% CI: 1.07-1.13, P=6×10^-11 ). Clinical variables predicted the risk of GDM (c-statistic 0.67, 95% CI: 0.64-0.70), and adding the GRS modestly improved prediction (0.70, 95% CI: 0.67-0.73).

CONCLUSIONS

Among Caucasian women, a GRS that included common T2DM genetic risk variants was associated with increased risk of GDM but showed limited utility in the identification of GDM cases.

Abstract MIP-046: BLOOD TYPE, ABO GENETIC VARIANTS, AND OVARIAN CANCER SURVIVAL

INTRODUCTION: In addition to a fundamental role in transfusion, blood type is also implicated in disease. Blood type antigens on red blood cells are determined by genetic variants in the ABO gene; both phenotype and genotype have been significantly associated with ovarian cancer risk. Meta–analysis of eight case–control studies indicated that women with genetic variants corresponding to blood type A had 9% greater ovarian cancer risk than women with variants corresponding to blood type O. Only one study to date has evaluated ovarian cancer prognosis; among 256 Chinese women, cases with blood type A had more than two–fold worse survival than cases with other blood types (B, AB, and O).

AIMS: To evaluate blood type phenotype and genotype in relation to overall ovarian cancer survival in a predominantly Caucasian study population.

APPROACH: Tumor registry confirmed ovarian or fallopian tube primary malignancies were identified from the Synthetic Derivative, a de–identified mirror of electronic medical records (EMR) from the Vanderbilt University Medical Center (VUMC). Blood type was ascertained from EMR linked laboratory reports. Ten common variants (minor allele frequency ≥0.05) in the ABO gene were ascertained using the Illumina Exome BeadChip. Subject vital status was determined from EMR and linkage to the National Death Index. Associations with overall survival were evaluated with proportional hazards regression in multivariable models that included adjustment for age, stage, grade, histologic subtype of disease, and year of diagnosis.

RESULTS: Blood type phenotype and genotype were available for 556 and 154 tumor registry confirmed ovarian cancer cases, respectively. Among all women, cases with blood type A had significantly better overall survival compared to either blood type O (Hazard ratio (HR): 0.79, 95% confidence interval (CI): 0.63–0.99) or to cases with any blood type other than A (HR: 0.80, 95% CI: 0.65–0.98). A missense variant in exon 7 (rs1053878) with moderate linkage to a variant corresponding to the A phenotype was also associated with better overall survival in a dominant manner (HR: 0.50, 95% CI: 0.25–0.99). While our phenotype association differed by race (p–interaction=0.049) and was evident only among Caucasian cases (HR: 0.75, 95% CI: 0.60–0.93), our genotype association did not vary by race (p–interaction=0.279).

CONCLUSIONS: Women with blood type A had better overall ovarian cancer survival, regardless of whether blood type was directly assayed, or inferred by genotype. These findings contradict the only existing ovarian cancer survival study to date, but include a larger, and predominantly Caucasian study population. Additional research is needed to either replicate or refute our ovarian cancer survival finding, and to determine if ABO variants and blood type are causally related to cancer development and progression.

Citation Format: Hilary Toole, Rebecca T. Levinson, Gabriella D. Cozzi, Angie Deng, Jason T. Fromal, Malcolm–Robert Snyder, Dineo Khabele, Alicia Beeghly–Fadiel . BLOOD TYPE, ABO GENETIC VARIANTS, AND OVARIAN CANCER SURVIVAL [abstract]. In: Proceedings of the 11th Biennial Ovarian Cancer Research Symposium; Sep 12-13, 2016; Seattle, WA. Philadelphia (PA): AACR; Clin Cancer Res 2017;23(11 Suppl):Abstract nr MIP-046.

Blood type, ABO genetic variants, and ovarian cancer survival

OBJECTIVE

Blood type A and the A1 allele have been associated with increased ovarian cancer risk. With only two small studies published to date, evidence for an association between ABO blood type and ovarian cancer survival is limited.

METHODS

We conducted a retrospective cohort study of Tumor Registry confirmed ovarian cancer cases from the Vanderbilt University Medical Center with blood type from linked laboratory reports and ABO variants from linked Illumina Exome BeadChip data. Associations with overall survival (OS) were quantified by hazard ratios (HR) and confidence intervals (CI) from proportional hazards regression models; covariates included age, race, stage, grade, histologic subtype, and year of diagnosis.

RESULTS

ABO phenotype (N = 694) and/or genotype (N = 154) data were available for 713 predominantly Caucasian (89.3%) cases. In multivariable models, blood type A had significantly better OS compared to either O (HR: 0.75, 95% CI: 0.60-0.93) or all non-A (HR: 0.77, 95% CI: 0.63-0.94) cases. Similarly, missense rs1053878 minor allele carriers (A2) had better OS (HR: 0.50, 95% CI: 0.25-0.99). Among Caucasians, this phenotype association was strengthened, but the genotype association was attenuated; instead, four variants sharing moderate linkage disequilibrium with the O variant were associated with better OS (HR: 0.62, 95% CI: 0.39-0.99) in unadjusted models.

CONCLUSIONS

Blood type A was significantly associated with longer ovarian cancer survival in the largest such study to date. This finding was supported by genetic analysis, which implicated the A2 allele, although O related variants also had suggestive associations. Further research on ABO and ovarian cancer survival is warranted.

Mitochondrial Haplogroups as a Risk Factor for Herpes Zoster

Background.  Herpes zoster, or shingles, is a common, painful reactivation of latent varicella zoster virus infection. Understanding host factors that predispose to herpes zoster may permit development of more effective prevention strategies. Our objective was to examine mitochondrial haplogroups as a potential host factor related to herpes zoster incidence. Methods.  Study participants were drawn from BioVU, a deoxyribonucleic acid (DNA) biobank connected to deidentified electronic medical records (EMRs) from Vanderbilt University Medical Center. Our study used 9691 Caucasian individuals with herpes zoster status determined by International Classification of Diseases, Ninth Revision codes 053-053.9. Cases and controls were matched on sex and date of birth within 5 years. Mitochondrial haplogroups were defined from mitochondrial DNA variants genotyped on the Illumina 660W or Illumina Infinium Human-Exome Beadchip. Sex and date of birth were extracted from the EMR. Results.  European mitochondrial haplogroup H had a protective association with herpes zoster status (odds ratio [OR] = .82; 95% confidence interval [CI], .71-.94; P = .005), whereas haplogroup clade IWX was a risk factor for herpes zoster status (OR = 1.38; 95% CI, 1.07-1.77; P = .01). Conclusions.  Mitochondrial haplogroup influences herpes zoster risk. Knowledge of a patient's mitochondrial haplogroup could allow for a precision approach to the management of herpes zoster risk through vaccination strategies and management of other modifiable risk factors.