Genetic Architecture and Clinical Outcomes of Combined Lipid Disturbances

Thomas Gilliland; Jacqueline S Dron; Margaret Sunitha Selvaraj; Mark Trinder; Kaavya Paruchuri; Sarah M Urbut; Sara Haidermota; Rachel Bernardo; Md Mesbah Uddin; Michael C Honigberg; Gina M Peloso; Pradeep Natarajan

doi:10.1161/CIRCRESAHA.123.323973

. Author manuscript; available in PMC: 2025 Jul 5.

Published in final edited form as: Circ Res. 2024 Jun 3;135(2):265–276. doi: 10.1161/CIRCRESAHA.123.323973

Genetic Architecture and Clinical Outcomes of Combined Lipid Disturbances

Thomas Gilliland ^(1),^(2),^(3),^*, Jacqueline S Dron ^(2),^(4),^*, Margaret Sunitha Selvaraj ^(1),^(2),⁽³⁾, Mark Trinder ^(2),⁽⁵⁾, Kaavya Paruchuri ^(1),^(2),⁽³⁾, Sarah M Urbut ^(1),^(2),⁽³⁾, Sara Haidermota ^(1),⁽²⁾, Rachel Bernardo ^(1),⁽²⁾, Md Mesbah Uddin ^(1),⁽²⁾, Michael C Honigberg ^(1),^(2),⁽³⁾, Gina M Peloso ⁽⁶⁾, Pradeep Natarajan ^(1),^(2),^(3),⁽⁴⁾

PMCID: PMC11223949 NIHMSID: NIHMS1996840 PMID: 38828614

Abstract

Background:

Dyslipoproteinemia often involves simultaneous derangements of multiple lipid traits. We aimed to evaluate the phenotypic and genetic characteristics of combined lipid disturbances in a general population-based cohort.

Methods:

Among UK Biobank participants without prevalent coronary artery disease (CAD), we used blood lipid and apolipoprotein B concentrations to ascribe individuals into one of six reproducible and mutually exclusive dyslipoproteinemia subtypes. Incident CAD risk was estimated for each subtype using Cox proportional hazards models. Phenome-wide analyses and genome-wide association studies (GWAS) were performed for each subtype, followed by in silico causal gene prioritization and heritability analyses. Additionally, the prevalence of disruptive variants in causal genes for Mendelian lipid disorders was assessed using whole-exome sequence data.

Results:

Among 450,636 UK Biobank participants: 63 (0.01%) had chylomicronemia; 40,005 (8.9%) had hypercholesterolemia; 94,785 (21.0%) had combined hyperlipidemia; 13,998 (3.1%) had remnant hypercholesterolemia; 110,389 (24.5%) had hypertriglyceridemia; and 49 (0.01%) had mixed hypertriglyceridemia and hypercholesterolemia. Over a median (IQR) follow-up of 11.1 (10.4-11.8) years, incident CAD risk varied across subtypes, with combined hyperlipidemia exhibiting the largest hazard (HR 1.92, 95% CI 1.84-2.01; P=2e-16), even when accounting for non-HDL-C (HR 1.45 95% CI 1.30-1.60; P=2.6e-12). GWAS revealed 250 loci significantly associated with dyslipoproteinemia subtypes, of which 72 (28.8%) were not detected in prior single lipid trait GWAS. Mendelian lipid variant carriers were rare (2.0%) among individuals with dyslipoproteinemia, but polygenic heritability was high, ranging from 23% for remnant hypercholesterolemia to 54% for combined hyperlipidemia.

Conclusions:

Simultaneous assessment of multiple lipid derangements revealed nuanced differences in CAD risk and genetic architectures across dyslipoproteinemia subtypes. These findings highlight the importance of looking beyond single lipid traits to better understand combined lipid and lipoprotein phenotypes and implications for disease risk.

Keywords: Dyslipoproteinemia, chylomicronemia, hypercholesterolemia, hyperlipidemia, hypertriglyceridemia

Subject Terms: Lipids and Cholesterol, Cardiovascular Disease, Genetic, Association Studies

Graphical Abstract

graphic file with name nihms-1996840-f0001.jpg

Introduction

Over 50 years ago, Fredrickson, Levy, and Lees (FLL) introduced a classification scheme for lipid disorders based on patterns of circulating lipoproteins ascertained using ultracentrifugation and gel electrophoresis.¹ In their seminal series of papers, the authors describe six dyslipoproteinemia types involving an excess of one or more lipoprotein fractions and were termed types I, IIa, IIb, III, IV, and V, reflecting chylomicron excess, low-density lipoprotein (LDL) excess, combined LDL and very-low-density lipoprotein (VLDL) excess, remnant cholesterol-rich lipoprotein excess, VLDL excess, and combined chylomicron and VLDL excess, respectively. The authors observed distinguishing clinical characteristics and patterns of heritability for each disorder and hypothesized pathophysiologic mechanisms. Though not routinely clinically applied owing to the cost and time associated with specialized techniques for diagnosis, the FLL framework highlights the importance of simultaneous lipid derangements and their disease-related impacts in clinical practice.

Sniderman and colleagues more recently developed and validated a diagnostic algorithm to group individuals into distinct dyslipoproteinemia subtypes, akin to the FLL scheme, but based on fasting serum levels of total cholesterol (TC), triglycerides (TG), and apolipoprotein B (apoB)². While application to smaller populations reveals a high prevalence of dyslipoproteinemia with relationships to subclinical atherosclerosis,^3,4 associations with clinical outcomes in large contemporary cohorts are presently unknown.

Analyses involving over a million individuals have revealed hundreds of genetic loci associated with traditional lipid measures, including LDL cholesterol (LDL-C), high-density lipoprotein cholesterol (HDL-C), TC and TG.⁵ However, this kind of analysis is limited in its ability to capture concurrent effects on multiple lipid traits that arise from specific disruptions to lipoprotein metabolism. By considering simultaneous lipid abnormalities in a more biologically informed manner, we can better capture genetic variants impacting specific lipoprotein disturbances and improve associated disease risk estimates. For example, a study of combined hyperlipidemia, defined by abnormalities in both LDL-C and TG, revealed that risk associated with cardiovascular disease in individuals with combined hyperlipidemia was equivalent to risk in individuals with monogenic familial hypercholesterolemia.⁶ Carefully curating phenotypes defined by simultaneous lipid derangements provides an opportunity to enhance genetic discovery related to specific lipoprotein metabolic actions and associated clinical consequences.

Here, we apply the apoB-based classification scheme developed by Sniderman and colleagues to a general population-based cohort to evaluate the clinical characteristics and cardiovascular outcomes of dyslipoproteinemia subtypes, including chylomicronemia, hypercholesterolemia, combined hyperlipidemia, remnant hypercholesterolemia, hypertriglyceridemia, and mixed hypertriglyceridemia and hypercholesterolemia. We complement our epidemiological assessments with genome-wide association studies (GWASs) of the dyslipoproteinemia subtypes and perform unbiased phenotype association testing to glean novel mechanistic insights into simultaneous derangements across multiple lipid traits.

Methods

Data Availability Statement

The data underlying this article were accessed from the UKB under Application number 7089. UKB individual-level data are available for request by application to the UKB (www.ukbiobank.ac.uk).

Study participants.

The UK Biobank (UKB) is a prospective cohort of ~500,000 volunteer participants living in the United Kingdom between the ages of 40-69 years. Recruitment occurred across 22 sites between 2006 and 2010.⁷ Laboratory measurements, anthropometrics, and clinical histories were ascertained at study enrollment, herein referred to as “baseline”. Extensive genetic and phenotypic data are available for UKB participants as previously described.⁷ Please see the Major Resources Table in the Supplemental Materials for additional information.

All study participants provided written and informed consent; withdrawn participants were excluded. The analyses described here were conducted under application number 7089, and the secondary use of the data was approved by the Massachusetts General Brigham Institutional Review Board (2013P001840).

Biochemical measurements.

Among UKB participants, fasting and non-fasting blood lipids, apoB, and lipoprotein particle concentrations were measured at baseline. ApoB was quantified by immuno-turbimetric assay, TC and TG by enzymatic assay, HDL-C by enzyme immune-inhibition assay, and direct LDL-C by enzymatic selective protection assay, all using the Beckman Coulter AU5800 analytical platform.⁸ Lipoprotein particle concentration was quantified using the Nightingale Health nuclear magnetic resonance (NMR) biomarker platform.^9-11

The presence of lipid-lowering medications was ascertained at baseline. Consistent with prior studies, we adjusted for the use of statins by dividing TC by 0.8, and both apoB and LDL-C by 0.7.^12,13 These adjusted values represent estimated untreated biochemical measurements estimated from average effects observed in randomized controlled trials.

Dyslipoproteinemia subtype definitions.

UKB participants were assigned one of six dyslipoproteinemia subtypes using the algorithm developed by Sniderman and colleagues that has thresholds for apoB and TG concentrations and ratios of TG/apoB and TC/apoB. The definitions for each dyslipoproteinemia subtype following this algorithm are provided in Table 1. The subtypes include chylomicronemia, hypercholesterolemia, combined hyperlipidemia, remnant hypercholesterolemia, hypertriglyceridemia, and mixed hypertriglyceridemia and hypercholesterolemia. UKB participants not meeting any of the aforementioned criteria were categorized as normolipidemic.² Estimated untreated biochemical measurements were considered when applying these criteria.

Table 1.

Definition and characteristics of dyslipoproteinemia subtypes.

	apoB (mg/dl)	TG (mg/dl)	TG/apoB	TC/apoB	Predicted lipoprotein excess
Chylomicronemia	<75	≥133	≥8.8	-	Chylomicrons
Hypercholesterolemia	≥120	<133	-	-	LDL
Combined hyperlipidemia	≥120	≥133	-	-	LDL and VLDL
Remnant hypercholesterolemia	<120	≥133	<8.8	≥2.4	Remnant particles
Hypertriglyceridemia	<120	≥133	<8.8	<2.4	VLDL
Mixed hypertriglyceridemia and hypercholesterolemia	≥75 and <120	≥133	≥8.8	-	Chylomicrons and VLDL

Open in a new tab

Subtypes were defined using the algorithm developed by Sniderman and colleagues.² ApoB and TG are expressed in mg/dl for both isolated measurements and in defining the thresholds for the ratios. Lipoprotein excess refers to the lipoprotein particle(s) predicted to be found in excess for each phenotype.²

Abbreviations: apoB = apolipoprotein B; LDL = low-density lipoprotein; TC = total cholesterol; TG = triglycerides; VLDL = very low-density lipoprotein.

Clinical phenotypes.

UKB participants were included for analysis if they were free of coronary artery disease (CAD) at baseline. CAD was defined based on the appearance of a qualifying International Classification of Diseases (ICD) code corresponding to acute myocardial infarction and coronary artery revascularization, physician, or patient report, as used previously (Supplemental Table 1).^14-16 Similarly, hypertension and type 2 diabetes—used as covariates in our analytic models—were defined based on a combination of qualifying ICD codes or physician or patient reports.

We coded 1584 phenotypes queried up to March 2020 in the UKB from both prevalent and incident hospital episodes using ICD-9 and ICD-10 diagnosis codes and grouped them into phecodes for the phenome-wide analyses.^17-19 Sex-specific phecodes were coded as “not applicable” for the opposite sex. Clinical phenotypes were defined using the Phecode Map 1.2 ICD-9 and ICD-10 phenotype groupings.¹⁹ Given the inherent limitations of ICD codes in practice, it should be noted that some misdiagnoses may be present within this dataset.²⁰

Genotyping, imputation, and whole-exome sequencing.

DNA samples from UKB participants were genotyped using the UK Biobank Lung Exome Variant Evaluation or Applied Biosystems UK Biobank Axiom Array. Imputation was performed for these genotypes using either the Haplotype Reference Consortium panel or the UK10K + 1000 Genomes panel.⁷

Whole-exome sequence data were generated for ~200,000 UKB participants and were made available in the first released tranche of sequence data.²¹

Protein-coding variants in Mendelian dyslipidemia genes.

Variants disrupting genes associated with Mendelian dyslipidemia phenotypes were identified in the subset of UKB participants with whole-exome sequence data. Genes were defined on whether the associated Mendelian dyslipidemia phenotype followed an autosomal dominant (AD) or autosomal recessive (AR) inheritance pattern. Genes related to high LDL-C included: LDLR, APOB and PCSK9 (AD); LDLRAP1, ABCG5 and ABCG8 (AR). Genes related to low HDL-C included: APOA1 (AD); ABCA1 and LCAT (AR). Genes related to high TG included: LPL and APOA5 (AD); APOC2, GPIBHP1, and LMF1 (AR). Variants were considered if minor allele frequencies (MAFs) were <1% and <10% for AD and AR genes, respectively, and had one of the following annotations: predicted as a high-confidence loss-of-function variant by the Ensemble Variant Effect Predictor LOFTEE plugin, defined as “pathogenic” or “pathogenic/likely pathogenic” in ClinVar (annotations from November 2021), or were predicted “damaging” by metaSVM.^22-24 All variants were annotated following canonical gene transcripts. Variant carriers were defined as those who were homozygous, heterozygotes, or double heterozygous for alternative variant alleles in AD genes or homozygotes for alternative variant alleles in AR genes.

Major genotypes for APOE (i.e., E2/E3/E4) were defined based on allelic combinations from rs429358 and rs7412, where rs429358 and rs7412 were T and T, respectively, for APOE E2; T and C for APOE E3; and C and C for APOE E4.

Statistical analysis.

The primary analysis focused on the association between incident CAD and dyslipoproteinemia subtypes relative to normolipidemic individuals using Cox proportional hazards models (‘survival’ R software package, version 3.2-13). The analysis accounted for study enrollment to March 31, 2020, with participants censored if loss-to-follow-up or death occurred prior to that date. Covariates included age, sex, smoking status, genotyping array, the first 5 principal components of genetic ancestry, hypertension, type 2 diabetes, BMI, and statin usage. Statistical significance was assigned at P<0.05. Two sensitivity analyses were performed. In the first, we included UKB participants not on lipid-lowering medications (including statins, ezetimibe, fibrates, or niacin) (N=78,472). In the second, we included individuals who fasted ≥8 hours before the baseline blood draw (N=18,488).

In the secondary analysis, associations between either hypercholesterolemia, combined hyperlipidemia, remnant hypercholesterolemia, or hypertriglyceridemia and incident CAD relative to normolipidemic individuals were assessed across different strata of non-HDL-C concentrations (130-160 mg/dl, 160-190 mg/dl, 190-220 mg/dl, and 220-250 mg/dl) using Cox proportional hazards models adjusted for age, sex, smoking status, genotyping array, the first 5 principal components of genetic ancestry, hypertension, type 2 diabetes, BMI, and statin usage. Heterogeneity testing was used to assess significant differences across non-HDL-C groups, and a meta-analysis was performed across non-HDL-C groups to calculate an overall effect of dyslipoproteinemia subtypes on CAD risk (‘meta’ R software package, version 4.19-0).²⁵ Statistical significance was assigned at P<0.05.

Phenome-wide association studies (pheWAS) for 1584 phenotypes and each dyslipoproteinemia subtype relative to normolipidemic individuals were performed with logistic regression (‘PheWAS’ R software package, version 0.99.5-4). Covariates included age, sex, genotyping array, and the first 5 principal components of genetic ancestry. After accounting for multiple hypothesis testing, statistical significance was assigned at P<3.2x10⁻⁵ (0.05/1584 phenotypes).

Separate case-control GWASs were performed for dyslipoproteinemia subtypes against normolipidemic individuals using REGENIE (version 1.0.2);²⁶ GWAS was not run for chylomicronemia or mixed hypertriglyceridemia and hypercholesterolemia because there were too few individuals representing cases. First- to third-degree relatives amongst UKB participants were identified using the KING method and excluded from analysis.²⁷ Genotypes with MAF <1%, minor allele count <100, sample missingness >10%, or Hardy-Weinberg equilibrium P-value <1x10⁻¹⁵, and samples with genotype missingness >10% were excluded from analysis. Using REGENIE, logistic regression was performed for each dyslipoproteinemia subtype using imputed variants (N=19,475,096) with MAF ≥1% and an imputation quality INFO score of ≥0.3. Covariates included age, sex, genotyping array, and the first 10 principal components of ancestry. Firth correction was applied to SNPs when P<0.01. Statistical genome-wide significance was assigned at P<5x10^-8. Novel loci were defined as ±500 kb windows around our lead significant SNPs that did not overlap with ±200 kb windows around lead SNPs from prior GWAS reports. In a secondary analysis examining the effect of fasting status, we repeated the above GWASs and additionally adjusted for fasting time and limited to the top independent genome-wide significant SNPs (P<5x10⁻⁸, ±500 kb) for each phenotype.

Using LD score regression (version 1.01) and restricting to SNPs present in HapMap3, we assessed SNP heritability and pair-wise genetic correlation of hypercholesterolemia, combined hyperlipidemia, remnant hypercholesterolemia, and hypertriglyceridemia.^28,29 We utilized a similarity-based method for gene prioritization, Polygenic Priority Score (PoPS), to investigate protein-coding gene contributions to each dyslipoproteinemia subtype.³⁰ PoPS was applied to summary-level GWAS data for each subtype using 1000 Genome Project as a reference ³¹ to yield a ranked set of candidate causal genes.

Analyses were performed using R version 4.0.2 (R Core Team, 2020) unless otherwise specified.

Results

Baseline characteristics of UKB participants.

Overall, 450,636 UKB participants with non-missing apoB and lipid measurements and free of prevalent CAD were included for analysis. The mean (SD) age was 56.3 (8.1) years, 44.4% were male, and 94.3% were white (Table 2). Dyslipoproteinemia subtypes were common, with 259,289 (57.5%) individuals meeting the criteria for one subtype definition. Hypertriglyceridemia was most common, observed in 24.5% (110,389) of individuals, followed by combined hyperlipidemia (21.0%; 94,785), hypercholesterolemia (8.9%; 40,005), remnant hypercholesterolemia (3.1%; 13,998), chylomicronemia (0.01%; 63), and mixed hypertriglyceridemia and hypercholesterolemia (0.01%; 49) (Supplemental Figure 1). The distribution of estimated untreated lipid measurements considered when defining each dyslipoproteinemia subtype is shown in Figure 1. The distribution of lipoprotein particle concentrations for each dyslipoproteinemia subtype is shown in Supplemental Figure 2.

Table 2.

Baseline characteristics of UKB study participants.

	All	Normolipidemic	Chylomicronemia	Hypercholesterolemia	Combined hyperlipidemia	Remnant hypercholesterolemia	Hypertriglyceridemia	Mixed hypertriglyceridemia and hypercholesterolemia
Number (%)	450636 (100)	191347 (42.5)	63 (0.01)	40005 (8.9)	94785 (21.0)	13998 (3.1)	110389 (24.5)	49 (0.01)
Demographics
Age, mean (SD), y	56.3 (8.1)	55.0 (8.3)	55.8 (8.6)	58.4 (7.3)	57.9 (7.5)	56.8 (7.7)	56.6 (8.1)	52.6 (8.6)
Male sex, n (%)	200353 (44.4)	67992 (35.5)	57 (90.5)	14991 (37.5)	48317 (51.0)	4792 (34.2)	64163 (58.1)	41 (83.7)
White, n (%)	424851 (94.3)	179080 (93.6)	59 (93.7)	37725 (94.3)	90518 (95.5)	13408 (95.8)	104015 (94.2)	46 (93.9)
Black, n (%)	7114 (1.6)	4635 (2.4)	0 (0.0)	847 (2.1)	635 (0.7)	128 (0.9)	869 (0.8)	0 (0.0)
Baseline lipids
TC, median (IQR), mg/dl	225.8 (199.5-253.9)	208.4 (186.7-229.6)	176.4 (155.8-200.2)	265.5 (247.1-285.7)	267.4 (248.2-289.9)	221.6 (197.1-247.0)	214.5 (195.0-232.5)	230.4 (206.8-256.5)
LDL-C, median (IQR), mg/dl	143.0 (122.7-165.2)	127.6 (111.6-142.9)	90.5 (76.4-98.6)	176.2 (166.0-189.5)	178.5 (166.2-194.7)	126.1 (109.8-143.1)	137.2 (123.5-149.2)	114.2 (107.0-129.6)
HDL-C, median (IQR), mg/dl	54.5 (45.7-65.1)	60.8 (51.5-71.2)	32.7 (27.7-39.0)	60.6 (51.9-70.7)	49.5 (43.0-57.2)	64.4 (54.6-74.1)	46.9 (40.3-54.5)	32.6 (29.5-37.4)
TG, median (IQR), mg/dl	130.6 (92.3-188.9)	90.3 (72.1-109.4)	605.7 (512.2-696.1)	105.1 (88.6-119.5)	205.0 (165.5-267.6)	179.3 (151.2-235.5)	181.9 (153.6-232.7)	839.3 (786.0-903.3)
ApoB, median (IQR), mg/dl	107.3 (91.9-123.7)	95.0 (83.5-106.0)	57.4 (48.4-68.6)	129.9 (124.2-139.7)	135.4 (126.7-148.3)	86.2 (75.0-97.1)	104.7 (94.8-112.5)	84.2 (78.4-94.3)
Lipid-lowering therapy
Statin, n (%)	62990 (14.0)	15903 (8.3)	5 (7.9)	9984 (25.0)	22888 (24.1)	522 (3.7)	13680 (12.4)	8 (16.3)
Ezetimibe, n (%)	1861 (0.4)	437 (0.2)	1 (1.6)	175 (0.4)	623 (0.6)	33 (0.2)	592 (0.5)	0 (0.0)
Fibrate, n (%)	1024 (0.2)	245 (0.1)	2 (3.2)	57 (0.1)	333 (0.4)	30 (0.2)	357 (0.3)	0 (0.0)
Niacin, n (%)	39 (0.009)	14 (0.007)	0 (0.0)	1 (0.002)	11 (0.01)	2 (0.01)	11 (0.01)	0 (0.0)
Any therapy, n (%)	78477 (17.4)	22718 (11.8)	9 (14.3)	11393 (28.5)	25869 (27.3)	1054 (7.5)	17426 (15.8)	8 (16.3)
Medical history
Smoking status, n (%)	199707 (44.3)	77564 (40.5)	36 (57.1)	17689 (44.2)	46826 (49.4)	6353 (45.4)	51207 (46.4)	32 (65.3)
Never	248706 (55.2)	112920 (59.0)	27 (42.9)	22145 (55.4)	47466 (50.1)	7573 (54.1)	58559 (53.0)	16 (32.7)
Previous	152861 (33.9)	60920 (31.8)	27 (42.9)	13759 (34.4)	34631 (36.5)	4920 (35.1)	38581 (35.0)	23 (46.9)
Current	46846 (10.4)	16644 (8.7)	9 (14.3)	3930 (9.8)	12195 (12.9)	1433 (10.2)	12626 (11.4)	9 (18.4)
Hypertension, n (%)	126092 (28.0)	41155 (21.5)	30 (47.6)	11699 (29.2)	33336 (35.2)	3573 (25.5)	36279 (32.9)	20 (40.8)
Diabetes mellitus, type 2, n (%)	9316 (2.1)	2624 (1.4)	6 (9.5)	467 (1.2)	2366 (2.5)	174 (1.2)	3674 (3.3)	5 (10.2)
BMI, mean (SD), kg/m ²	27.4 (4.8)	25.9 (4.4)	28.9 (3.8)	26.9 (4.3)	28.9 (4.5)	26.9 (4.5)	28.8 (4.8)	29.3 (4.1)

Open in a new tab

Abbreviations: apoB = apolipoprotein B; BMI = body-mass index; HDL-C = high-density lipoprotein cholesterol; IQR = interquartile range; LDL-C = low-density lipoprotein cholesterol; SD = standard deviation; TC = total cholesterol; TG = triglycerides.

Figure 1. — Distributions are presented as mg/dl on a log(10) scale, with the vertical line in each distribution indicating the median value for that lipid trait. ApoB, LDL-C, and TC are corrected for baseline statin prescription, reflecting estimated untreated values. Abbreviations: apoB = apolipoprotein B; HDL-C = high-density lipoprotein cholesterol; LDL-C = low-density lipoprotein cholesterol; TC = total cholesterol; TG = triglycerides.

Mean BMI was lowest among normolipidemic individuals (25.9 kg/m²) and highest in individuals with mixed hypertriglyceridemia and hypercholesterolemia (29.3 kg/m²). Type 2 diabetes was most common in individuals with mixed hypertriglyceridemia and hypercholesterolemia (10.2%), while its prevalence in individuals with hypercholesterolemia (1.2%) and remnant hypercholesterolemia (1.2%) was similar to normolipidemic individuals (1.4%). Statin treatment was most common in individuals with hypercholesterolemia (25.1%) and combined hyperlipidemia (24.3%), compared to normolipidemic individuals (8.3%).

Risk of incident CAD by dyslipoproteinemia subtype.

Over a median (IQR) 11.1 (10.4-11.8) years of follow-up, individuals with combined hyperlipidemia had the highest risk of incident CAD (hazard ratio [HR] 1.92, 95% CI 1.84-2.01; P=2e-16) compared to normolipidemic individuals, followed by individuals with hypercholesterolemia (HR 1.51, 95% CI 1.42-1.61; P=2e-16) and hypertriglyceridemia (HR 1.28, 95% CI 1.22-1.34; P=2e-16) (Figure 2, Supplemental Table 2). Individuals with remnant hypercholesterolemia were not at significantly increased risk for CAD relative to normolipidemic individuals (HR 1.05, 95% CI 0.93-1.18; P=0.44). Significant differences in CAD risk were observed between individuals with combined hyperlipidemia, remnant hypercholesterolemia, and hypertriglyceridemia (Supplemental Table 3). In a sensitivity analysis, participants who were on lipid-lowering medications at baseline were excluded, which revealed a consistent pattern of increased hazard for individuals with combined hyperlipidemia compared to other dyslipoproteinemia subtypes when contrasted with normolipidemic individuals (Supplemental Table 4).

Figure 2. — Each line reflects the crude cumulative incidence rates of CAD for individuals with combined hyperlipidemia (N=94,785), remnant hypercholesterolemia (N=13,998), hypertriglyceridemia (N=110,389), hypercholesterolemia (N=40,005), and normolipidemic individuals (N=191,347) over a median follow-up of 11.1 years. The shaded areas represent the 95% CI. Abbreviations: CAD = coronary artery disease.

Adjusting for either apoB or non-HDL-C attenuated the effect of the dyslipoproteinemia subtype on incident CAD, with preservation of increased hazard among individuals with combined hyperlipidemia compared to other common dyslipoproteinemia subtypes (Supplemental Tables 5 and 6 for models additionally adjusting for apoB and non-HDL-C, respectively). Remnant hypercholesterolemia reached nominal significance only in the model adjusted for apoB (HR 1.15, 95% CI 1.02-1.29; P=0.02).

Across strata of non-HDL-C (130-160 mg/dl, 160-190 mg/dl, 190-220 mg/dl, and 220-250 mg/dl), combined hyperlipidemia was the only dyslipoproteinemia subtype associated with increased risk of incident CAD when compared to stratum-matched normolipidemic individuals. In random-effects meta-analysis across the non-HDL-C strata, combined hyperlipidemia conferred a 1.45-fold independent hazard (95% CI 1.30-1.60; P=2.6e-12) for incident CAD (Figure 3). Risk was elevated for individuals with either hypertriglyceridemia (HR 1.18, 95% CI 1.10-1.30; P=5.5e-6) or hypercholesterolemia (HR 1.17, 95% CI 1.04-1.30; P=6.8e-3) but not for individuals with remnant hypercholesterolemia (Supplemental Figure 3).

Figure 3. — Individuals with combined hyperlipidemia were compared against stratum-matched normolipidemic individuals across four strata of non-HDL-C. The random-effects meta-analysis is reported for the overall combined hyperlipidemia effect. Covariates included age, sex, smoking status, genotyping array, the first 5 principal components of genetic ancestry, hypertension, type 2 diabetes, BMI, and statin usage. Abbreviations: CI = confidence interval; HR = hazard ratio; non-HDL-C = non-high-density lipoprotein cholesterol; τ² = effect size variance estimated by restricted maximum likelihood method.

Phenome-wide association testing in UKB.

Unbiased pheWAS was performed to screen for associations between dyslipoproteinemia subtypes and 1584 phecodes among UKB participants. Effect estimates (odds ratios) between each dyslipoproteinemia subtype and phecode for the top 50 associations for each are shown in Supplemental Figure 4, but do not reflect direction of causality; further testing is required. Representative top associations across the dyslipoproteinemia subtypes are shown in Figure 4. Hypercholesterolemia was strongly associated with increased odds of atherosclerotic cardiovascular disease phenotypes, although the associations were stronger for combined hyperlipidemia. Combined hyperlipidemia was also strongly associated with increased odds of diabetes mellitus, liver disease, and obesity. Similarly, hypertriglyceridemia was associated with increased odds of diabetes mellitus, liver disease, and obesity, but had a weaker association with atherosclerotic cardiovascular disease. Remnant hypercholesterolemia was most strongly associated with increased odds of alcohol-related disorders and malnutrition along with liver disease but did not show significant associations with diabetes mellitus or CAD.

Figure 4. — A PheWAS of UKB phenotypes was conducted for combined hyperlipidemia, hypercholesterolemia, hypertriglyceridemia, and remnant hypercholesterolemia relative to normolipidemic individuals. Odds ratios are only presented when a nominal P<0.05 is reached and are marked with “*” if phenome-wide significant (Bonferroni adjusted P<3.2x10⁻⁵). Blank cells reflect exposure-phenotype associations P>0.05. Covariates included age, sex, genotyping array, and the first 5 principal components of genetic ancestry. Abbreviations: OR = odds ratio.

Genomic architecture of dyslipoproteinemia subtypes.

Only 2,020 of 98,587 (2.0%) UKB participants with a dyslipoproteinemia subtype and whole-exome sequence data carried a disruptive protein-coding variant in a Mendelian lipid gene. The hypercholesterolemia subtype had the highest carrier prevalence of variants in LDL-C genes, with 479 of 15,309 (3.1%) individuals carrying a variant. Variants were most commonly identified in LDLR (Supplemental Table 7, Supplemental Figure 5). When assessing APOE genotype status, the E2/E2 genotype, which is classically associated with FLL type III,³² was markedly enriched among individuals with remnant hypercholesterolemia (11.4%; 619) relative to the other subtypes and was rare among normolipidemic individuals (0.5%; 344) (Supplemental Table 8, Supplemental Figure 6).

Case-control GWASs for individuals with either hypercholesterolemia (N=37,211), combined hyperlipidemia (N=88,007), remnant hypercholesterolemia (N=13,030), or hypertriglyceridemia (N=102,677) against normolipidemic individuals (N=177,576) revealed 77, 210, 31, and 99 SNPs associated with each subtype, respectively (Supplemental Figure 7, Supplemental Tables 9-13), representing 250 independent loci. Of these loci, 13 were shared across all phenotypes (Supplemental Table 14) and were near lipid-associated genes. A remaining 12, 97, 5, and 21 loci were uniquely associated with hypercholesterolemia, combined hyperlipidemia, remnant hypercholesterolemia, and hypertriglyceridemia, respectively (Supplemental Table 15).

When compared to meta-analyzed GWAS results from the Million Veterans Program and Global Lipids Genetics Consortium for lipid traits, 72 of our 250 (28.8%) loci are novel, having not previously reached genome-wide significance.³³ Furthermore, when compared against published studies in the GWAS Catalog,³⁴ 14 of our 250 (5.6%) loci have not been previously associated with HDL-C, LDL-C, TC, TG, or apoB concentrations (Supplemental Table 16). Examples of these novel loci include: rs17576576, which associates with hypercholesterolemia and is upstream of peroxisome proliferator-activated receptor gamma, coactivator 1 alpha (PPARGC1A); rs3856522, which associates with combined hyperlipidemia and is an exonic variant of ATPase homolog 2 pseudogene (AHSA2P); and rs141955254, which is an intergenic variant in proximity to BMP/retinoic acid inducible neural specific 1 (BRINP1) and the sole novel purported association with remnant hypercholesterolemia.

We also noted an overlap between genome-wide significant SNPs in a recent CARDIoGRAM-C4D Consortium CAD GWAS meta-analysis excluding the UKB and genome-wide significant SNPs from our UKB dyslipoproteinemia GWASs.³⁵ Of the 241 significant CAD SNPs, combined hyperlipidemia had the largest overlap compared to other dyslipoproteinemia subtypes, at 16.9% (Figure 5, Supplemental Table 17). The direction of effect generally aligned between CAD SNPs and SNPs associated with hypercholesterolemia and combined hyperlipidemia, but not for remnant hypercholesterolemia and hypertriglyceridemia (Supplemental Figure 9). For example, variants in the APOB locus were associated with an increased risk of CAD and an increased risk of both hypercholesterolemia and combined hyperlipidemia but were associated with a decreased risk of remnant hypercholesterolemia and hypertriglyceridemia. Further, we observe that variants in certain loci, such as LPA, are strongly associated with increased CAD risk but only have a modest associated effect on risk of dyslipoproteinemia subtypes.

Figure 5. — The Venn diagram shows the overlap of 250 genetic loci significantly associated with either hypercholesterolemia, combined hyperlipidemia, remnant hypercholesterolemia, or hypertriglyceridemia. Numbers are presented as loci counts (% of total loci).

Using external multi-omics data, PoPS ranked protein-coding genes near GWAS loci for hypercholesterolemia, combined hyperlipidemia, remnant hypercholesterolemia, and hypertriglyceridemia. Supplemental Figure 10 shows the union of the top 20 genetic contributors to each phenotype according to PoPS score. APOE was the top-ranked gene across all dyslipoproteinemia subtypes, and common variants linked to Mendelian lipid-related genes were prioritized for all subtypes, including LDLR, APOB, SCARB1, and the APOA5/A4/C3/A1 locus. LPL was prioritized more highly for subtypes with elevated TG levels. Some prioritized novel genes were PDGFB, encoding platelet-derived growth factor subunit B and IGF1, encoding insulin-like growth factor 1 for hypertriglyceridemia; and BRCA1, encoding BRCA1 DNA repair associated and B4GALT1, encoding beta-1,4-galactosyltransferase 1 for remnant hypercholesterolemia.

LD score regression was used to estimate heritabilities and genetic correlations among the dyslipoproteinemia subtypes (Supplemental Figure 11). The strongest pairwise genetic correlations were between hypertriglyceridemia and remnant hypercholesterolemia, and the weakest genetic correlations were between hypercholesterolemia and remnant hypercholesterolemia. Observed SNP heritabilities (standard errors [SE]) for hypercholesterolemia, combined hyperlipidemia, remnant hypercholesterolemia, and hypertriglyceridemia were 0.37 (0.1), 0.54 (0.09), 0.23 (0.29), and 0.28 (0.04), respectively.

Accounting for fasting time among dyslipoproteinemia subtypes.

Among individuals who fasted ≥8 hours prior to baseline blood draw (N=18,488), 52.3% had a dyslipoproteinemia subtype, similar to the 57.5% of all individuals (fasting and non-fasting) with a dyslipoproteinemia subtype. No significant differences were observed in the prevalence of chylomicronemia, combined hyperlipidemia, remnant hypercholesterolemia, or mixed hypertriglyceridemia and hypercholesterolemia (Supplemental Table 18). Repeating the GWAS with fasting time as a covariate and restricting to top independent SNPs for each phenotype did not change effect estimate directions or association strength of the dyslipoproteinemia subtypes (Supplemental Figure 12, Supplemental Table 19).

Discussion

In a large contemporary prospective cohort, we identified 57.5% of individuals with one of six dyslipoproteinemia subtypes defined using an apoB-based classification algorithm. Extensive analyses looking at the association between each subtype and various clinical outcomes, as well as simultaneous assessment for both common SNPs and protein-coding variants provided the opportunity for a more nuanced understanding of these complex phenotypes and related implications.

First, there is a clinical benefit of defining individuals by combinations of lipid derangements to refined risk estimates for CAD and other diseases. Of the dyslipoproteinemia subtypes, we showed combined hyperlipidemia conferred the highest risk of incident CAD, independent of non-traditional lipid parameters such as apoB or non-HDL-C; this subtype also had the largest number of overlapping SNPs with the CARDIoGRAM-C4D Consortium CAD GWAS.³⁵ Although hypercholesterolemia and hypertriglyceridemia were also associated with increased CAD risk, these associations were attenuated after adjusting for apoB. When considering other disease outcomes, both combined hyperlipidemia and hypertriglyceridemia were associated with increased odds of diabetes and liver disease, but hypercholesterolemia was not. These differences between subtypes have important implications for treatment guidelines. For example, while individuals with combined hyperlipidemia and hypercholesterolemia would benefit from lipid-lowering strategies to lower LDL-C levels, additional measures to reduce risk for liver disease could be considered in individuals with combined hyperlipidemia. This added clinical consideration may be overlooked if only single lipid abnormalities are considered.

Second, the genetic architecture of dyslipoproteinemia subtypes is strongly polygenic and has shared features between subtypes with similar lipid abnormalities. Early studies reported associations between candidate SNPs and FLL phenotypes in small, clinically ascertained cohorts.^36,37 Our advanced experimental design using a contemporary population-based cohort and genome-wide assessment of SNPs allowed us to robustly demonstrate the polygenic nature of dyslipoproteinemia subtypes. And while there are indeed Mendelian dyslipidemia phenotypes with lipid profiles that align with our subtype definitions, polygenic contributors are far more common than variants in Mendelian lipid genes. This is consistent with prior genetic studies of individuals with severe hypercholesterolemia and with combined hyperlipidemia.³⁸ Of the 250 independent genetic loci we detected across the dyslipoproteinemia subtypes, 72 (28.8%) were not detectable through conventional single lipid trait GWAS, highlighting the necessity of considering lipids in combination. Further, we show high SNP heritability and strong genetic correlation between subtypes with elevated TG measures, including combined hyperlipidemia, remnant hypercholesterolemia, and hypertriglyceridemia. We also observe overlapping loci between subtypes and various diseases, strengthening the associations observed in our pheWAS. For example, we identified a SNP (rs2943641) upstream from IRS1 that is significantly associated with both combined hyperlipidemia and hypertriglyceridemia. This SNP was previously associated with a 35% increased risk of type 2 diabetes³⁹ and we observed that these two subtypes had the strongest risk association with type 2 diabetes.

Third, combining our genetic and phenotypic findings reveals a more in-depth understanding of the dyslipoproteinemia subtypes, particularly remnant hypercholesterolemia. Reminiscent of the FLL type III phenotype, we recapitulated the increased prevalence of the APOE E2/E2 genotype in individuals with remnant hypercholesterolemia compared to all other groups and identified only 5 loci uniquely associated with this subtype. We also found remnant hypercholesterolemia was associated with alcohol-related disorders and both malnutrition and hyperalimentation, which might suggest that this subtype is more heavily influenced by environmental exposures, as previously reported.³² However, we did not observe any enrichment for exogenous hormone usage in individuals with remnant hypercholesterolemia (Supplemental Table 20).

Lastly, shared and distinct pathological mechanisms underlying dyslipoproteinemia subtypes can be identified. For example, from our GWAS we identified SNPs around RBM47 that are uniquely associated with combined hyperlipidemia. The RBM47 locus was recently reported as a putative novel CAD risk locus among individuals of Middle Eastern ancestry,⁴⁰ and protein-coding variants were not only shown to be associated with TG/HDL-C ratio, but specific variants can significantly alter apoC-III concentrations.⁴¹ This may indicate that alterations to apoC-III function may contribute towards phenotypic expression of combined hyperlipidemia and individuals may benefit from apo-CIII-lowering therapies in addition to LDL-C-lowering therapies. We also found the FAM13A locus to be uniquely associated with combined hyperlipidemia (lead variant rs6824451, OR 1.04; P=7.84x10⁻¹⁰), with nominal significance for hypertriglyceridemia (OR 1.03; P=7.48x10⁻⁷). FAM13A encodes family with sequence similarity 13 member A, which has been implicated in body fat distribution in humans and visceral to subcutaneous adipose tissue ratio in mice via effects on adipocyte differentiation;⁴² a functional genomic study in adipocytes prioritized FAM13A as a causal gene in insulin resistance.⁴³ Overall, these findings point to a common mechanism connecting abnormal adipocyte homeostasis with dyslipoproteinemia subtypes defined by elevated TG levels.

This study has limitations. First, the UKB participants are predominantly of European ancestry, meaning a more ancestrally diverse cohort is necessary to test whether our findings can be generalized to non-European groups. Second, we relied on primarily non-fasting lipid values for dyslipoproteinemia subtype definitions. Reassuringly, when limiting UKB participants to those with ≥8-hour fast prior to the baseline lab draw, a similar proportion of participants met the criteria for a dyslipoproteinemia subtype, suggesting this is a high prevalence population. Further evidence of high baseline dyslipoproteinemia is that hypercholesterolemia prevalence was doubled compared to contemporary cohorts (8.9% versus 4-5%), despite apoB not being significantly influenced by the post-prandial state.^17,44 And importantly, though isolated hypertriglyceridemia is classically defined in the fasting state, post-prandial hypertriglyceridemia has equal implications for cardiovascular risk.⁴⁵ Our approach of applying the Sniderman apoB-based criteria to non-fasting samples permits clinical and genetic discovery on a scale not feasible when restricting to fasting-only samples, and we show that genetic results are unaffected by fasting time. Another limitation due to the paucity of fasted participants was that when further restricted to individuals with NMR-derived lipoprotein particle concentrations, there were not enough individuals with either chylomicronemia or mixed hypertriglyceridemia and hypercholesterolemia to effectively assess lipoprotein excess, particularly for chylomicrons and VLDL (Supplemental Figure 2).

An important consideration for the work presented here is that the dyslipoproteinemia subtypes we defined using the apoB-based classification algorithm do not necessarily coincide with classic FLL phenotypes. While there are similarities in the lipoprotein fractions defining each phenotype, it is important to recognize that FLL phenotypes are diagnosed in individuals ascertained clinically using specialized ultracentrifugation and gel electrophoresis techniques. The dyslipoproteinemia subtypes we defined are based on accessible lipid measurements and allow for easier and more practical clinical assessments.

Strengthened by large sample size, availability of baseline lipid metrics, comprehensive disease outcome data, and extensive genetic information, we were able to deeply characterize both the genetic architecture and clinical relevance of six dyslipoproteinemia subtypes. Accounting for patterns of simultaneous lipid derangements allows for more specific descriptions of underlying pathophysiologic disturbances, which individual lipid measures alone cannot provide.

Supplementary Material

323973 Data Set

NIHMS1996840-supplement-323973_Data_Set.xlsx^{(176.7KB, xlsx)}

323973 Data Supplement

NIHMS1996840-supplement-323973_Data_Supplement.pdf^{(3.3MB, pdf)}

323973 Major Resources Table

NIHMS1996840-supplement-323973_Major_Resources_Table.pdf^{(361.8KB, pdf)}

323973 STROBE Statement

NIHMS1996840-supplement-323973_STROBE_Statement.pdf^{(588.3KB, pdf)}

Novelty and Significance.

What is known?

The Fredrickson, Levy, and Lees (FLL) classification scheme for dyslipoproteinemia subtypes is based on patterns of circulating lipoproteins.
Dyslipoproteinemia subtypes have unique clinical characteristics, patterns of heritability, and hypothesized pathophysiologic mechanisms.
Diagnostic challenges related to both time and cost have prevented the routine clinical assessment of dyslipoproteinemia subtypes.

What new information does this article contribute?

Associations with clinical outcomes vary for each dyslipidemia subtype, with combined hyperlipidemia having the strongest associated risk for coronary artery disease (CAD) and strong associations with type 2 diabetes and obesity.
Genome-wide association studies (GWASs) revealed 250 loci significantly associated with dyslipoproteinemia subtypes, of which 72 (28.8%) were not detected in prior single lipid trait GWASs.
Clinical and genetic association assessments can be leveraged to uncover possible mechanisms underlying dyslipoproteinemia subtypes, such as the shared FAM13A GWAS association that links abnormal adipocyte homeostasis and insulin resistance to subtypes defined by elevated triglyceride levels (i.e., combined hyperlipidemia and hypertriglyceridemia).

To our knowledge, this is the largest and most comprehensive assessment of dyslipoproteinemia subtypes. We identified that over half (57.5%) of the participants in the UK Biobank (N=450,636) have dyslipoproteinemia, with associated risk for various clinical outcomes. Emphasizing the importance of looking beyond single lipid traits to better understand combined lipid and lipoprotein phenotypes and implications for disease risk, we showed that combined hyperlipidemia had the highest associated risk for CAD (HR: 1.92), independent of non-traditional lipid parameters such as apolipoprotein (apo) B or non-high-density lipoprotein cholesterol (non-HDL-C). In contrast, both hypercholesterolemia and hypertriglyceridemia showed attenuated CAD risk after apoB adjustment. Our findings also revealed varying levels of associated odds for comorbidities such as type 2 diabetes, obesity, and liver disease across different subtypes. From our GWAS results, we identified 250 loci significantly associated with dyslipoproteinemia subtypes, many of which were not detected in GWASs of single lipid traits. From these results, we gained mechanistic insights across dyslipoproteinemia subtypes and mechanisms linking multiple subtypes. Notably, sensitivity analyses incorporating fasting status yielded nearly identical genetic effects for our dyslipoproteinemia subtypes, suggesting comparability to the traditional FLL classification scheme that utilizes fasting lipid values for their definitions.

Acknowledgments

The authors would like to acknowledge and thank the staff, investigators, and participants of the UK Biobank.

Sources of Funding

Dr. Gilliland is partially supported by the National Heart, Lung, and Blood Institute T32 grant 5T32HL125232. Dr. Dron is supported by the Canadian Institute of Health Research as a Banting Postdoctoral Research Fellow. Dr. Selvaraj is supported by a TOPMed Fellowship (2022-6842.02). Drs. Peloso and Natarajan are supported by grants from the National Heart, Lung, and Blood Institute (R01HL142711, R01HL127564). In addition, Dr. Natarajan is supported by grants from the National Human Genome Research Institute (U01HG011719) and Massachusetts General Hospital (Paul and Phyllis Fireman Endowed Chair in Vascular Medicine).

Non-standard Abbreviations and Acronyms

AD: autosomal dominant
apoB: apolipoprotein B
AR: autosomal recessive
CAD: coronary artery disease
CI: confidence interval
FLL: Fredrickson, Levy, and Lees
GWAS: genome-wide association study
HDL-C: high-density lipoprotein cholesterol
HDL: high-density lipoprotein
HR: hazard ratio
ICD: International Classification of Diseases
IQR: interquartile range
LDL-C: low-density lipoprotein cholesterol
LDL: low-density lipoprotein
MAF: minor allele frequency
pheWAS: phenome-wide association study
PoPS: Polygenic Priority Score
SE: standard error
TC: total cholesterol
TG: triglyceride
UKB: UK Biobank
VLDL: very-low-density lipoprotein

Footnotes

Disclosures

P.N. reports research grants from Allelica, Amgen, Apple, Boston Scientific, Genentech / Roche, and Novartis, personal fees from Allelica, Apple, AstraZeneca, Blackstone Life Sciences, Creative Education Concepts, CRISPR Therapeutics, Eli Lilly & Co, Foresite Labs, Genentech / Roche, GV, HeartFlow, Magnet Biomedicine, Merck, and Novartis, scientific advisory board membership of Esperion Therapeutics, Preciseli, and TenSixteen Bio, scientific co-founder of TenSixteen Bio, equity in MyOme, Preciseli, and TenSixteen Bio, and spousal employment at Vertex Pharmaceuticals, all unrelated to the resent work. The other authors report no disclosures.

References

1.Classics Fredrickson DS, Levy RI, Lees RS. Fat transport in lipoproteins; an integrated approach to mechanisms and disorders. N Engl J Med. 1967;276:34–44. [DOI] [PubMed] [Google Scholar]
2.de Graaf J, Couture P, Sniderman A. A diagnostic algorithm for the atherogenic apolipoprotein B dyslipoproteinemias. Nat Clin Pract Endocrinol Metab. 2008;4:608–618. [DOI] [PubMed] [Google Scholar]
3.Holewijn S, Sniderman AD, den Heijer M, Swinkels DW, Stalenhoef AFH, de Graaf J. Application and validation of a diagnostic algorithm for the atherogenic apoB dyslipoproteinemias: ApoB dyslipoproteinemias in a Dutch population-based study. Eur J Clin Invest. 2011;41:423–433. [DOI] [PubMed] [Google Scholar]
4.Sathiyakumar V, Pallazola VA, Park J, Vakil RM, Toth PP, Lazo-Elizondo M, Quispe R, Guallar E, Banach M, Blumenthal RS, Jones SR, Martin SS. Modern prevalence of the Fredrickson-Levy-Lees dyslipidemias: findings from the Very Large Database of Lipids and National Health and Nutrition Examination Survey. Arch Med Sci. 2020;16:1279–1287. [DOI] [PMC free article] [PubMed] [Google Scholar]
5.Graham SE, Clarke SL, Wu KHH, Kanoni S, Zajac GJM, Ramdas S, Surakka I, Ntalla I, Vedantam S, Winkler TW, et al. The power of genetic diversity in genome-wide association studies of lipids. Nature. 2021;600:675–679. [DOI] [PMC free article] [PubMed] [Google Scholar]
6.Trinder M, Vikulova D, Pimstone S, Mancini GBJ, Brunham LR. Polygenic architecture and cardiovascular risk of familial combined hyperlipidemia. Atherosclerosis. 2022;340:35–43. [DOI] [PubMed] [Google Scholar]
7.Bycroft C, Freeman C, Petkova D, Band G, Elliott LT, Sharp K, Motyer A, Vukcevic D, Delaneau O, O’Connell J, et al. The UK Biobank resource with deep phenotyping and genomic data. Nature. 2018;562:203–209. [DOI] [PMC free article] [PubMed] [Google Scholar]
8.Fry D, Almond R, Moffat S, Gordon M, Singh P. UK Biobank Biomarker Project: Companion Document to Accompany Serum Biomarker Data.; 2019. https://biobank.ndph.ox.ac.uk/showcase/ukb/docs/serum_biochemistry.pdf [Google Scholar]
9.Julkunen H, Cichońska A, Tiainen M, Koskela H, Nybo K, Mäkelä V, Nokso-Koivisto J, Kristiansson K, Perola M, Salomaa V, et al. Atlas of plasma NMR biomarkers for health and disease in 118,461 individuals from the UK Biobank. Nat Commun. 2023;14:604. [DOI] [PMC free article] [PubMed] [Google Scholar]
10.Würtz P, Kangas AJ, Soininen P, Lawlor DA, Davey Smith G, Ala-Korpela M. Quantitative Serum Nuclear Magnetic Resonance Metabolomics in Large-Scale Epidemiology: A Primer on -Omic Technologies. Am J Epidemiol. 2017;186:1084–1096. [DOI] [PMC free article] [PubMed] [Google Scholar]
11.Soininen P, Kangas AJ, Würtz P, Suna T, Ala-Korpela M. Quantitative serum nuclear magnetic resonance metabolomics in cardiovascular epidemiology and genetics. Circ Cardiovasc Genet. 2015;8:192–206. [DOI] [PubMed] [Google Scholar]
12.Natarajan P, Peloso GM, Zekavat SM, Montasser M, Ganna A, Chaffin M, Khera AV, Zhou W, Bloom JM, Engreitz JM, et al. Deep-coverage whole genome sequences and blood lipids among 16,324 individuals. Nat Commun. 2018;9:3391. [DOI] [PMC free article] [PubMed] [Google Scholar]
13.Baigent C, Blackwell L, Emberson J, Holland LE, Reith C, Bhala N, Peto R, Barnes EH, Keech A, Simes J, et al. Efficacy and safety of more intensive lowering of LDL cholesterol: a meta-analysis of data from 170,000 participants in 26 randomised trials. Lancet. 2010;376:1670–1681. [DOI] [PMC free article] [PubMed] [Google Scholar]
14.Honigberg MC, Zekavat SM, Aragam K, Finneran P, Klarin D, Bhatt DL, Januzzi JL Jr, Scott NS, Natarajan P. Association of Premature Natural and Surgical Menopause With Incident Cardiovascular Disease. JAMA. 2019;322:2411–2421. [DOI] [PMC free article] [PubMed] [Google Scholar]
15.Klarin D, Zhu QM, Emdin CA, Chaffin M, Horner S, McMillan BJ, Leed A, Weale ME, Spencer CCA, Aguet F, et al. Genetic analysis in UK Biobank links insulin resistance and transendothelial migration pathways to coronary artery disease. Nat Genet. 2017;49:1392–1397. [DOI] [PMC free article] [PubMed] [Google Scholar]
16.Honigberg MC, Zekavat SM, Niroula A, Griffin GK, Bick AG, Pirruccello JP, Nakao T, Whitsel EA, Farland LV, Laurie C, et al. Premature Menopause, Clonal Hematopoiesis, and Coronary Artery Disease in Postmenopausal Women. Circulation. 2021;143:410–423. [DOI] [PMC free article] [PubMed] [Google Scholar]
17.Denny JC, Ritchie MD, Basford MA, Pulley JM, Bastarache L, Brown-Gentry K, Wang D, Masys DR, Roden DM, Crawford DC. PheWAS: demonstrating the feasibility of a phenome-wide scan to discover gene-disease associations. Bioinformatics. 2010;26:1205–1210. [DOI] [PMC free article] [PubMed] [Google Scholar]
18.Wei WQ, Bastarache LA, Carroll RJ, Marlo JE, Osterman TJ, Gamazon ER, Cox NJ, Roden DM, Denny JC. Evaluating phecodes, clinical classification software, and ICD-9-CM codes for phenome-wide association studies in the electronic health record. PLoS One. 2017;12:e0175508. [DOI] [PMC free article] [PubMed] [Google Scholar]
19.Wu P, Gifford A, Meng X, Li X, Campbell H, Varley T, Zhao J, Carroll R, Bastarache L, Denny JC, et al. Mapping ICD-10 and ICD-10-CM Codes to Phecodes: Workflow Development and Initial Evaluation. JMIR Med Inform. 2019;7:e14325. [DOI] [PMC free article] [PubMed] [Google Scholar]
20.Mainor AJ, Morden NE, Smith J, Tomlin S, Skinner J. ICD-10 Coding Will Challenge Researchers: Caution and Collaboration may Reduce Measurement Error and Improve Comparability Over Time. Med Care. 2019;57:e42–e46. [DOI] [PMC free article] [PubMed] [Google Scholar]
21.Szustakowski JD, Balasubramanian S, Kvikstad E, Khalid S, Bronson PG, Sasson A, Wong E, Liu D, Wade Davis J, Haefliger C, et al. Advancing human genetics research and drug discovery through exome sequencing of the UK Biobank. Nat Genet. 2021;53:942–948. [DOI] [PubMed] [Google Scholar]
22.Liu X, Wu C, Li C, Boerwinkle E. dbNSFP v3.0: A One-Stop Database of Functional Predictions and Annotations for Human Nonsynonymous and Splice-Site SNVs. Hum Mutat. 2016;37:235–241. [DOI] [PMC free article] [PubMed] [Google Scholar]
23.Landrum MJ, Lee JM, Riley GR, Jang W, Rubinstein WS, Church DM, Maglott DR. ClinVar: public archive of relationships among sequence variation and human phenotype. Nucleic Acids Res. 2014;42:D980–5. [DOI] [PMC free article] [PubMed] [Google Scholar]
24.Karczewski KJ, Francioli LC, Tiao G, Cummings BB, Alföldi J, Wang Q, Collins RL, Laricchia KM, Ganna A, Birnbaum DP, et al. The mutational constraint spectrum quantified from variation in 141,456 humans. Nature. 2020;581:434–443. [DOI] [PMC free article] [PubMed] [Google Scholar]
25.Schwarzer G, Carpenter JR, Rücker G. Meta-Analysis with R. Springer International Publishing [Google Scholar]
26.Mbatchou J, Barnard L, Backman J, Marcketta A, Kosmicki JA, Ziyatdinov A, Benner C, O’Dushlaine C, Barber M, Boutkov B, et al. Computationally efficient whole-genome regression for quantitative and binary traits. Nat Genet. 2021;53:1097–1103. [DOI] [PubMed] [Google Scholar]
27.Manichaikul A, Mychaleckyj JC, Rich SS, Daly K, Sale M, Chen WM. Robust relationship inference in genome-wide association studies. Bioinformatics. 2010;26:2867–2873. [DOI] [PMC free article] [PubMed] [Google Scholar]
28.International HapMap 3 Consortium. Integrating common and rare genetic variation in diverse human populations. Nature. 2010;467:52–58. [DOI] [PMC free article] [PubMed] [Google Scholar]
29.Bulik-Sullivan BK, Loh PR, Finucane HK, Ripke S, Yang J, Schizophrenia Working Group of the Psychiatric Genomics Consortium, Patterson N, Daly MJ, Price AL, Neale BM. LD Score regression distinguishes confounding from polygenicity in genome-wide association studies. Nat Genet. 2015;47:291–295. [DOI] [PMC free article] [PubMed] [Google Scholar]
30.Weeks EM, Ulirsch JC, Cheng NY, Trippe BL, Fine RS, Miao J, Patwardhan TA, Kanai M, Nasser J, Fulco CP, et al. Leveraging polygenic enrichments of gene features to predict genes underlying complex traits and diseases. Nat Genet. 2023;55:1267–1276. [DOI] [PMC free article] [PubMed] [Google Scholar]
31.1000 Genomes Project Consortium, Auton A, Brooks LD, Durbin RM, Garrison EP, Kang HM, Korbel JO, Marchini JL, McCarthy S, McVean GA, Abecasis GR. A global reference for human genetic variation. Nature. 2015;526:68–74. [DOI] [PMC free article] [PubMed] [Google Scholar]
32.de Beer F, Stalenhoef AFH, Hoogerbrugge N, Kastelein JJP, Gevers Leuven JA, van Duijn CM, Havekes LM, Smelt AHM. Expression of type III hyperlipoproteinemia in apolipoprotein E2 (Arg158 --> Cys) homozygotes is associated with hyperinsulinemia. Arterioscler Thromb Vasc Biol. 2002;22:294–299. [DOI] [PubMed] [Google Scholar]
33.Klarin D, Damrauer SM, Cho K, Sun YV, Teslovich TM, Honerlaw J, Gagnon DR, DuVall SL, Li J, Peloso GM, et al. Genetics of blood lipids among ~300,000 multi-ethnic participants of the Million Veteran Program. Nat Genet. 2018;50:1514–1523. [DOI] [PMC free article] [PubMed] [Google Scholar]
34.Buniello A, MacArthur JAL, Cerezo M, Harris LW, Hayhurst J, Malangone C, McMahon A, Morales J, Mountjoy E, Sollis E, et al. The NHGRI-EBI GWAS Catalog of published genome-wide association studies, targeted arrays and summary statistics 2019. Nucleic Acids Res. 2019;47:D1005–D1012. [DOI] [PMC free article] [PubMed] [Google Scholar]
35.Aragam KG, Jiang T, Goel A, Kanoni S, Wolford BN, Atri DS, Weeks EM, Wang M, Hindy G, Zhou W, et al. Discovery and systematic characterization of risk variants and genes for coronary artery disease in over a million participants. Nat Genet. 2022;54:1803–1815. [DOI] [PMC free article] [PubMed] [Google Scholar]
36.Johansen CT, Wang J, Lanktree MB, McIntyre AD, Ban MR, Martins RA, Kennedy BA, Hassell RG, Visser ME, Schwartz SM, et al. An increased burden of common and rare lipid-associated risk alleles contributes to the phenotypic spectrum of hypertriglyceridemia. Arterioscler Thromb Vasc Biol. 2011;31:1916–1926. [DOI] [PMC free article] [PubMed] [Google Scholar]
37.Hegele RA, Ban MR, Hsueh N, Kennedy BA, Cao H, Zou GY, Anand S, Yusuf S, Huff MW, Wang J. A polygenic basis for four classical Fredrickson hyperlipoproteinemia phenotypes that are characterized by hypertriglyceridemia. Hum Mol Genet. 2009;18:4189–4194. [DOI] [PMC free article] [PubMed] [Google Scholar]
38.Gill PK, Dron JS, Berberich AJ, Wang J, McIntyre AD, Cao H, Hegele RA. Combined hyperlipidemia is genetically similar to isolated hypertriglyceridemia. J Clin Lipidol. 2021;15:79–87. [DOI] [PubMed] [Google Scholar]
39.Rung J, Cauchi S, Albrechtsen A, Shen L, Rocheleau G, Cavalcanti-Proença C, Bacot F, Balkau B, Belisle A, Borch-Johnsen K, et al. Genetic variant near IRS1 is associated with type 2 diabetes, insulin resistance and hyperinsulinemia. Nat Genet. 2009;41:1110–1115. [DOI] [PubMed] [Google Scholar]
40.Kunji K, El-Menyar A, Ullah E, Kullo I, Saad M, Al Suwaidi J. Whole genome sequencing of a middle eastern cohort replicates known loci for coronary heart disease and identifies potential novel ones. J Am Coll Cardiol. 2021;77:13. [Google Scholar]
41.Sandholm N, Hotakainen R, Haukka JK, Jansson Sigfrids F, Dahlström EH, Antikainen AA, Valo E, Syreeni A, Kilpeläinen E, Kytölä A, et al. Whole-exome sequencing identifies novel protein-altering variants associated with serum apolipoprotein and lipid concentrations. Genome Med. 2022;14:132. [DOI] [PMC free article] [PubMed] [Google Scholar]
42.Fathzadeh M, Li J, Rao A, Cook N, Chennamsetty I, Seldin M, Zhou X, Sangwung P, Gloudemans MJ, Keller M, et al. FAM13A affects body fat distribution and adipocyte function. Nat Commun. 2020;11:1465. [DOI] [PMC free article] [PubMed] [Google Scholar]
43.Chen Z, Yu H, Shi X, Warren CR, Lotta LA, Friesen M, Meissner TB, Langenberg C, Wabitsch M, Wareham N, et al. Functional Screening of Candidate Causal Genes for Insulin Resistance in Human Preadipocytes and Adipocytes. Circ Res. 2020;126:330–346. [DOI] [PMC free article] [PubMed] [Google Scholar]
44.Mora S, Chang CL, Moorthy MV, Sever PS. Association of Nonfasting vs Fasting Lipid Levels With Risk of Major Coronary Events in the Anglo-Scandinavian Cardiac Outcomes Trial-Lipid Lowering Arm. JAMA Intern Med. 2019;179:898–905. [DOI] [PMC free article] [PubMed] [Google Scholar]
45.Langsted A, Nordestgaard BG. Nonfasting versus fasting lipid profile for cardiovascular risk prediction. Pathology. 2019;51:131–141. [DOI] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials