Subtyping Severe Hypercholesterolemia by Genetic Determinant to Stratify Risk of Coronary Artery Disease

Abstract

Background:

Severe hypercholesterolemia, defined as low-density lipoprotein cholesterol (LDL-C) measurement ≥ 190 mg/dl, is associated with increased risk for coronary artery disease (CAD). Causes of severe hypercholesterolemia include: monogenic familial hypercholesterolemia (FH), polygenic hypercholesterolemia, elevated lipoprotein(a) [Lp(a)] hypercholesteremia, polygenic hypercholesterolemia with elevated Lp(a) (two-hit), or non-genetic hypercholesterolemia. The added value of using a genetics approach to stratifying risk of incident CAD among those with severe hypercholesterolemia versus using LDL-C levels alone for risk stratification is not known.

Methods:

To determine whether risk stratification by genetic cause provided better 10-year incident CAD risk stratification than LDL-C level, a retrospective cohort study comparing incident CAD risk among severe hypercholesterolemia subtypes (genetic and non-genetic causes) was performed among 134,185 UK Biobank participants. Analyses were limited to unrelated, White British or Irish participants with available exome sequencing data. Participants with cardiovascular disease at baseline were excluded from analyses of incident CAD.

Results:

Of 134,185 individuals, 70,637 (52.6%) were female, and the mean (SD) age was 56.7 (8.0) years. 9.0% of the cohort met severe hypercholesterolemia criteria. Participants with LDL-C between 210-229 mg/dL and LDL-C ≥230 mg/dL showed modest increases in incident CAD risk relative to those with LDL-C between 190-209 mg/dL (hazard ratio [HR] (95% confidence interval [CI]) 210-229 mg/dL: 1.3 (1.1-1.6); ≥230 mg/dL: 1.3 (1.0-1.7). In contrast, when risk was stratified by genetic subtype, monogenic FH, elevated Lp(a), and two-hit hypercholesterolemia subtypes had increased rates of incident CAD relative to the non-genetic hypercholesterolemia subtype (HR (95% CI): monogenic FH: 2.3 (1.4-4.0); elevated Lp(a): 1.6 (1.2-2.0); two-hit: 1.9 (1.4-2.6)), while polygenic hypercholesterolemia did not.

Conclusions:

Genetics-based subtyping for monogenic FH and Lp(a) in those with severe hypercholesterolemia provided better stratification of 10-year incident CAD risk than LDL-C-based stratification.

Graphical Abstract

Introduction

Severe hypercholesterolemia is associated with increased risk for coronary artery disease (CAD)^1,2. Current guidelines define severe hypercholesterolemia as the presence of a low-density lipoprotein cholesterol (LDL-C) measurement ≥ 190 mg/dl. Several distinct genetic determinants are known to cause the disorder including the presence of a monogenic familial hypercholesterolemia (FH) causing variant^3,4, an elevated serum concentration of lipoprotein (a) (Lp(a)) which is primarily controlled by genetic variation in and near the LPA gene⁵, and polygenic hypercholesterolemia caused by the inheritance of a large burden of common alleles that are each associated with incremental increases of LDL-C^6,7.

There is controversy in the field regarding the clinical value of measuring genetic determinants of severe hypercholesterolemia⁸. The controversy stems from the hypothesis that LDL-C levels alone carry sufficient information to guide therapy without measuring genetic markers of chronic, lifelong exposure to elevated LDL-C. Supporting the argument for identifying genetic determinants is the observation that patients with a monogenic FH variant have three times higher CAD risk than those without a variant at any given level of LDL-C⁹ and the knowledge that Lp(a) is a predictor of CAD risk independent of other risk factors⁵. Polygenic hypercholesterolemia is also thought to increase CAD risk, but its independence from LDL-C is less well established⁶.

In this study, we examined four genetic determinants of severe hypercholesterolemia in the UK Biobank. For each determinant, we established its contribution to disease incidence and the proportion of risk among severe hypercholesterolemia cases directly attributable to exposure. Next, we created a severe hypercholesterolemia cohort and classified participants into five mutually exclusive subtypes defined by the four genetic determinants: 1) monogenic FH caused by a pathogenic or likely pathogenic (P/LP) variant in a canonical FH gene 2) polygenic hypercholesterolemia caused by an LDL-C polygenic score in the top decile of the population¹⁰, 3) elevated Lp(a) hypercholesterolemia defined as a serum Lp(a) concentration ≥125 nmol/L (50mg/dL)^11,12 4) two-hit hypercholesterolemia defined as coexisting elevated Lp(a) and polygenic hypercholesterolemia, and 5) non-genetic, defined as the absence of a genetic or molecular determinant¹³. We also investigated the impact of lipid lower medication (LLM) on outcomes by genetic subtype. Time-to-event analyses were used to test if subtyping provides risk stratification of 10-year incident CAD beyond stratification by LDL-C levels.

Methods

Availability of Data

All anonymized electronic health record (EHR), self-report, and genetic data used here can be obtained directly from the UK Biobank. Analysis scripts are available upon reasonable request to the authors.

Participants

This study was approved by the Geisinger Institutional Review Board. The UK Biobank is a large epidemiological cohort with genetic data, self-reported health data, and linkage to hospital inpatient records¹⁴. UK Biobank participants were 40-69 years old when recruited. This study was conducted from January 2021 to January 2023, under UK Biobank project number 49945. The UK Biobank has ethical approval from the North West Multi-Centre Ethics Committee. All participants provided informed consent to participate in UK Biobank projects.

Several exclusion criteria were applied to the 200,602 participants with exome sequences available for analysis (Figure 1). To reduce confounding by genetic ancestry and relatedness between participants, the cohort was filtered to only include individuals of white British or Irish ancestry. First-degree relatives were removed using the UK Biobank tools package in R¹⁵. Those with missing LDL-C, high-density lipoprotein cholesterol (HDL), systolic blood pressure measurements, or smoking status were also excluded. Exome sequences of the remaining 149,326 UK Biobank participants were screened for FH-causing variants in canonical FH genes (Table S1; Supplemental Methods)¹⁶. 65 unique FH-causing variants were identified in 357 individuals, 204 of whom had LDL-C ≥ 190mg/dL. Individuals who did not have a monogenic FH variant but had a secondary cause of hypercholesterolemia (Table S2) or were missing imputed genotype data or Lp(a) measurements were also excluded.

Figure 1.

Flow chart of study design. 130,091 UK Biobank participants with exome sequences available were classified by molecular and genetic determinants of severe hypercholesterolemia. 11,738 participants with severe hypercholesterolemia were classified into subtypes based on the presence of corresponding determinants. Those without a determinant were classified as the non-genetic subtype. LDL-C = low-density lipoprotein cholesterol; HDL = high-density lipoprotein cholesterol; BP = blood pressure; FH = familial hypercholesterolemia; PRS = polygenic risk score; Lp(a) = lipoprotein (a); CAD = coronary artery disease.

Genetic instruments for prediction of polygenic hypercholesterolemia and elevated Lp(a) were calculated from imputed genotype array data. This study used genotypes imputed from the Haplotype Reference Consortium reference panel¹⁷. Polygenic risk for hypercholesterolemia was estimated in the filtered cohort using the PRS-CS software. For variant weights, we used summary statistics from a genome-wide association study of quantitative LDL-C levels performed on a European ancestry sample reported by the Global Lipids Genetics Consortium⁷. Genetically-predicted Lp(a) was calculated using the same genetic instrument for predicted Lp(a) levels created by Burgess et al.¹⁸ and reported in the UK Biobank by Trinder et al.⁵ This Lp(a) score includes 43 variants in the LPA gene region (Supplemental Methods).

Classification of Severe Hypercholesterolemia Determinants and Subtypes

Those with lipid-lowering medication (LLM) use at baseline (defined by self-reported data collected during an in-person interview and questionnaire) (Table S3) had LDL-C values adjusted by dividing direct LDL-C by 0.7 to approximate the untreated state (Supplemental Methods)⁹. Those with LDL-C ≥ 190 mg/dL after adjustment were considered to have severe hypercholesterolemia. This approach was supported by sensitivity analyses (Table S4, Table S5).

The severe hypercholesterolemia cohort, containing only individuals with LDL-C ≥ 190 mg/dL, was stratified using two approaches: 1) LDL-C-based and 2) subtyping-by genetic determinant. In the LDL-C-based stratification, the cohort was divided into three groups: LDL-C ≥ 190 and < 210 mg/dL, LDL-C ≥ 210 and < 230 mg/dL, and ≥ 230 mg/dL. For the genetic stratification, each participant was classified into one of five mutually exclusive subtypes based on genetic determinants of severe hypercholesterolemia: 1) monogenic FH caused by a P/LP variant in a canonical FH gene 2) polygenic hypercholesterolemia defined by an LDL-C polygenic score in the top decile of the population, a commonly used threshold for designating high genetic risk in analyses of polygenic risk scores, including hypercholesterolemia^10,19 and atherosclerotic cardiovascular disease (ASCVD)⁶ 3) elevated Lp(a) hypercholesterolemia defined as a serum Lp(a) concentration ≥125 nmol/L (50 mg/dL) based on the 2018 American Heart Association guidelines for high risk¹² 4) two-hit hypercholesterolemia defined as coexisting elevated Lp(a) and polygenic hypercholesterolemia, and 5) non-genetic, defined as the absence of a genetic or molecular determinant¹³ (Table S6).

Baseline Clinical Characteristics and Outcomes

CAD and cardiovascular disease (CVD) phenotypes were defined by self-reported disease and inpatient diagnoses as described in Elliot et al.²⁰. CAD was defined as heart attack or myocardial infarction and related sequelae, identified using ICD-9 and ICD-10 codes and self-report; and coronary angioplasty or coronary artery bypass grafts, identified using OPCS-4 codes and self-report (Table S7). The CVD definition includes CAD in addition to angina, transient ischemic attack, and ischemic stroke. Peripheral artery disease (PAD) and aortic valve stenosis were defined following Bjornsson et al.²¹. The American Heart Association’s (AHA) Lifestyle Score (poor, intermediate, or ideal) was determined for UK Biobank participants following Said et al.²² (Supplemental Methods).

Statistical Analyses

To calculate the likelihood of having severe hypercholesterolemia at baseline for each determinant relative to the remainder of the cohort, generalized linear models stratified by age binned in 5 year increments, sex, and controlling for the first five principal components of ancestry were performed in R version 4.1.1²³. Population attributable fraction (PAF) is the proportional reduction in disease that would occur if exposure to a risk factor were removed from the population. Attributable risk proportion (ARP) is an estimate of the proportion of risk among the exposed that can be directly attributed to the exposure of interest. PAF and ARP were calculated using the twoxtwo library in R²⁴.

Linear and logistic regression were used to determine whether each genetically-defined subtype differed from the non-genetic subtype in baseline CAD risk factors, comorbidities, and self-reported family history of heart disease. All regression analyses controlled for age, sex, and the first five principal components of ancestry. Forest plots were visualized using the ggplot2 package in R²⁵. P-values were adjusted for multiple tests using Benjamini-Hochberg correction and considered significant if adjusted P<0.05²⁶.

Due to the increased risk of myocardial infarction (defined here as CAD) in patients with a history of myocardial infarction, angina, transient ischemic attack, or ischemic stroke (defined here as CVD), individuals with pre-existing CVD were removed prior to the analysis of incident CAD²⁷. Incident CAD was defined as a first-time CAD event occurring over the 10 years immediately following assessment at the UK Biobank assessment center. Cox proportional hazard regression models were fitted using the survival package in R²⁸, stratifying by age, sex, baseline LLM use, smoking status, diabetes status, and controlling for systolic blood pressure (BP), HDL, LDL-C and the first five principal components of ancestry, where appropriate. A sensitivity analysis additionally controlling for body mass index (BMI) and C-reactive protein was also performed. The model was stratified by covariates when necessary to meet proportionality assumptions. The proportionality assumption of every Cox model was tested with the cox.zph function, and all results were non-significant. Kaplan-Meier plots were generated using the survminer package in R²⁹. Absolute risk of 10-year incident CAD was calculated following the method described by Austin (2010)³⁰.

RESULTS

Prevalence of Determinants of Severe Hypercholesterolemia

The UK Biobank cohort comprised 130,091 individuals with exome sequence and genotype array data that met study inclusion criteria at the time of study (Table 1; Figure 1). 357 individuals (1 in 418; 0.24%) had a P/LP variant in an FH gene (Table S8); 3,325 (1 in 39; 2.6%) had a coexisting high LDL-C polygenic risk score (PRS_LDL-C) and serum concentration of Lp(a) ≥ 125 nmol/L; 9,657 (1 in 13; 7.4%) had a high PRS_LDL-C alone; and 17,940 (1 in 7; 13.8%) had a serum concentration of Lp(a) ≥ 125 nmol/L alone. Of these, 11,738 individuals (1 in 11, 9.0%) met criteria for severe hypercholesterolemia (LDL-C ≥ 190 mg/dL). The proportion of the five subtypes in the severe hypercholesterolemia cohort was 1.7% monogenic FH (n=204), 7.8% two-hit hypercholesterolemia (n=913), 17.6% polygenic hypercholesterolemia (n=2,070), 17.2% elevated Lp(a) (n=2,023), and 55.6% non-genetic (n=6,528) (Table 2).

Table 1.

Demographics

Sample Size (N)	130,091
Age (Mean (SD))	56.68 (8.01)
Sex (% Female)	52.6
LDL-C (mg/dL) ((Mean (SD))	137.91 (33.27)
LDL-C (est. untreated) (mg/dL) ((Mean (SD))	145.73 (33.01)
BMI (Mean (SD))	27.22 (4.62)
Systolic BP (mmHg) (Mean (SD))	137.72 (18.49)
HDL-C (mg/dL) (Mean (SD))	56.71 (14.73)
Smoking (%)	9.3
Cholesterol-lowering medication (%)	16.8
Years follow-up ((Mean (IQR))	11.31 (1.63)

Table 2.

Clinical characteristics and comorbidities of subtypes in the severe hypercholesterolemia cohort (n = 11,738).

	Non-genetic	Monogenic	Two-hit	Polygenic	Elevated Lp(a)
Sample Size (N)	6528	204	913	2070	2023
Prevalence	1 in 2	1 in 58	1 in 13	1 in 6	1 in 6
Age (Mean (SD))	58.88 (6.99)	56.9 (7.23)	58.1 (7.31)	58.02 (7.21)	59.07 (6.67)
Sex (% Female)	55.5	64.7	57.6	54.9	61.2
LDL-C (est. untreated) (mg/dL) ((Mean (SD))	210.19 (21.46)	236.1 (40.46)	214.74 (24.74)	212.68 (23.37)	211.45 (22.58)
BMI (Mean (SD))	28.13 (4.24)	27.57 (4.92)	27.68 (4.5)	28.03 (4.3)	27.99 (4.43)
Systolic BP (mmHg) (Mean (SD))	143.04 (18.42)	137.98 (18.82)	141.21 (18.71)	141.47 (18.52)	143.02 (18.7)
HDL-C (mg/dL) (Mean (SD))	57.14 (13.11)	55.67 (11.78)	59.47 (14.45)	57.33 (12.75)	59.18 (13.59)
Triglycerides (mg/dL) (Mean (SD))	194.44 (74.71)	146.53 (71.81)	173.15 (72.44)	187.83 (74.6)	182.92 (72.35)
BP meds (%)	23.7	20.1	20.4	19.6	26.1
Smoking (%)	11.3	8.3	10.5	9.6	9.7
Cholesterol-lowering medication (%)	29.8	56.4	32	28.1	32.5
AHA Lifestyle Score (Mean (SD)) 1	0.99 (0.44)	1.06 (0.49)	1.01 (0.45)	1.02 (0.43)	1.03 (0.44)
Townsend Deprivation Index (Mean (SD)) 2	−1.58 (2.9)	−1.47 (3.09)	−1.61 (2.94)	−1.6 (2.89)	−1.55 (2.82)
Prevalent T1D (%)	14/6528 (0.2)	2/204 (1)	5/913 (0.5)	2/2070 (0.1)	4/2023 (0.2)
Prevalent T2D (%)	18/6514 (0.3)	0/202 (0)	3/908 (0.3)	3/2068 (0.1)	4/2019 (0.2)
Prevalent CAD (%)	2	3.9	1.4	1.7	2.9
Prevalent CVD (%)	6.7	8.3	4.9	4.7	6.6
Prevalent PAD (%)	0.4	1	0.5	0.1	0.5
Prevalent AVS (%)	0.2	0.5	0.1	0.1	0.2

¹AHA Lifestyle Score: A higher score indicates a more ideal lifestyle (0 = poor, 1 = intermediate, 2 = ideal).

²Townsend Deprivation Index: A higher (less negative) score corresponds to a lower socioeconomic status.

Attributable Risk of Determinants to Severe Hypercholesterolemia

Individuals with an FH-causing variant were at the highest risk of severe hypercholesterolemia relative to the remainder of the cohort (adjusted odds ratio [OR], 13.9; 95% confidence interval [CI], 11.2-17.2) (Figure 2). Individuals with two determinants: both a PRS_LDL-C in the top decile and a serum concentration of Lp(a) ≥ 125 nmol/L (OR, 4.1; 95% CI, 3.7-4.4), a PRS_LDL-C in the top decile alone (OR, 3.2; 95% CI, 3.0-3.4), and serum concentration of Lp(a) ≥ 125 nmol/L alone (OR, 1.3; 95% CI, 1.2-1.4) were also at increased risk of having severe hypercholesterolemia. The PAF of FH-causing variants, two determinants, PRS_LDL-C in the top decile alone, and a serum concentration Lp(a) ≥ 125 nmol/L alone were 1.5%, 5.4%, 11.0%, and 4.0% respectively; and the ARP were 84.4%, 68.9%, 62.5%, and 23.2% respectively (Table S9).

Figure 2.

Genetic determinants of severe hypercholesterolemia. A fan plot shows prevalence, odds ratio, and population attributable fraction for each determinant of severe hypercholesterolemia. The x and y coordinates of each point indicate the prevalence and odds ratio (OR), respectively, of each determinant. The population attributable fraction is represented by the y-axis intercept of the blue line connected to each point. The dashed horizontal black line demarcates an odds ratio of 1.

Clinical Characteristics of Severe Hypercholesterolemia Subtypes

The monogenic FH subtype was associated with the most striking differences in clinical characteristics relative to the non-genetic subtype including younger age, more likely to be female, higher LDL-C, lower HDL, lower triglycerides, higher rates of lipid-lowering medication at baseline, and use of a stronger statin compound (atorvastatin or rosuvastatin) (Figure S1A, Figure S2). The non-genetic subtype was associated with a poor AHA lifestyle score compared to polygenic and elevated Lp(a) hypercholesterolemia subtypes. The prevalence of comorbidities at baseline were similar among the subtypes with these exceptions: the monogenic FH and elevated Lp(a) hypercholesterolemia subtypes were associated with a higher prevalence of CAD at baseline (monogenic OR, 3.0; 95% CI, 1.3-6.0; elevated Lp(a) OR, 1.6; 95% CI, 1.2-2.2), and polygenic hypercholesterolemia was associated with a lower prevalence of type 2 diabetes (OR, 0.6; 95% CI, 0.5-0.8) and CVD (OR, 0.7; 95% CI, 0.6-0.9) (Figure S1B). The monogenic FH and elevated Lp(a) hypercholesterolemia subtypes were associated with the highest odds of reporting a first-degree relative with heart disease (monogenic OR, 2.0; 95% CI, 1.5-2.7; elevated Lp(a) OR, 1.3; 95% CI, 1.2-1.4) (Figure S3).

LDL-C Level Versus Subtyping for Risk Stratification of 10-Year Incident CAD

We observed 485 incident CAD events after removing 729 participants with prevalent CVD at baseline, corresponding to an event rate of 0.46% per person-year. Using the LDL-C-based stratification of the cohort, the LDL-C 210-229 mg/dL group (n=2,588) and the LDL-C ≥ 230 mg/dL group (n=1,593) were both associated with increased risk of incident CAD relative to the LDL-C 190-209 mg/dL group (n=6,828) (210-229 HR, 1.3; 95% CI, 1.1-1.7; ≥ 230 HR, 1.3; 95% CI, 1.0-1.7) (Figure 3A).

Figure 3.

Severe hypercholesterolemia stratified by LDL-C and genetic subtype. Kaplan-Meier plots show 10-year incident CAD risk among UK Biobank participants A) binned into three groups of increasing LLM-adjusted LDL-C: 190-210 mg/dL (reference), 210-230 mg/dL and >230 mg/dL and B) binned by non-genetic (reference) and genetic subtypes. Cox proportional-hazard ratios and 95% confidence intervals adjusted for CAD risk factors and principal components of ancestry are shown.

In the genetic determinant subtype analysis, the monogenic FH (n=187), two-hit (n=868), and elevated Lp(a) subtypes (n=1,889) were associated with increased risk of incident CAD relative to the non-genetic subtype (monogenic FH HR, 2.3; 95% CI, 1.4-4.0; two-hit HR, 1.9; 95% CI, 1.4-2.6; elevated Lp(a) HR, 1.5; 95% CI, 1.2-2.0) (Figure 3B). Within the monogenic FH subtype, those with an LDLR loss-of-function variant (n=17) had the highest incident CAD risk (HR, 8.9; 95% CI, 2.4-32.7), however the clinical significance cannot be determined due to the small sample size (Figure S4). The polygenic hypercholesterolemia subtype (n=1,972) was not associated with increased incident CAD. Having an LDL-C ≥ 230 mg/dL was associated with a 1.2 absolute 10-year incident CAD risk difference relative to having an LDL-C between 190-209 mg/dL (Table S10). In contrast, having the monogenic subtype was associated with a 4.8% absolute 10-year incident CAD risk difference relative to having the non-genetic subtype. The results were not sensitive to whether the polygenic and two-hit subtypes were defined by the top decile or top quartile of LDL-C PRS, another threshold that has been used for designating high genetic risk of hypercholesterolemia³¹ (Table S11). Using genetically predicted Lp(a) instead of measured Lp(a) did not meaningfully affect the results (Figure S5).

The previous analyses were repeated separately in males and females. In males, LDL-C-based stratification did not result in a significantly increased risk of incident CAD in either the LDL-C 210-229 mg/dL group or the LDL-C ≥ 230 mg/dL group relative to the LDL-C 190-209 mg/dL group (Figure S6A). When using genetic subtyping of males, the monogenic FH (HR, 2.3; 95% CI, 1.1-4.5) and two-hit (HR, 1.8; 95% CI, 1.3-2.6) subtypes were associated with increased risk of incident CAD relative to non-genetic (Figure S6B). In females, LDL-C-based stratification was incoherent, with a significantly increased CAD risk observed in the LDL-C 210-229 mg/dL group relative to the LDL-C 190-209 mg/dL group (HR, 1.8; 95% CI, 1.3-2.7), but no significant association between the LDL-C ≥ 230 mg/dL group relative to the LDL-C 190-209 mg/dL group (Figure S6C). When using genetic subtyping of females, the monogenic FH (HR, 3.2; 95% CI, 1.2-8.1), two-hit (HR, 2.3; 95% CI, 1.3-4.0), and elevated Lp(a) (HR, 2.2; 95% CI, 1.5-3.3) subtypes were associated with increased risk of incident CAD relative to non-genetic (Figure S6D).

LLM Sensitivity Analyses

Relative to those in the severe hypercholesterolemia cohort who were untreated with LLM at baseline, the LLM treated subgroup had an overall higher burden of CAD risk factors (Table S12), a higher prevalence of more severe LDLR loss-of-function variants (Figure S7), a higher incidence of CAD when not accounting for clinical risk factors, and a lower risk of incident CAD (HR, 0.7; 95% CI, 0.6-0.9) when controlling for clinical risk factors (Table S13). To address these complexities, the analysis of incident CAD risk by severe hypercholesterolemia subtype was repeated separating the LLM-treated and untreated subgroups. In the LLM-treated subgroup, the monogenic FH and the two-hit subtypes were associated with increased risk of incident CAD relative to the non-genetic subtype (monogenic HR, 3.1; 95% CI, 1.6-5.9; two-hit HR, 1.8; 95% CI, 1.1-3.0) (Figure 4). The remaining subtypes were not significantly different in risk from the non-genetic subtype. In the LLM-untreated subgroup, the two-hit subtype and the elevated Lp(a) subtype were associated with increased risk of CAD (two-hit HR, 1.9; 95% CI, 1.3-2.8; elevated Lp(a) HR, 1.7; 95% CI, 1.3-2.3). Too few individuals with the monogenic FH subtype were untreated at baseline (n=88), therefore while higher rates of CAD were observed relative to the non-genetic subtype, the difference did not achieve statistical significance (HR=1.5; 95% CI, 0.5-4.2). Polygenic hypercholesterolemia was not different from the non-genetic subtype with respect to CAD risk in either subgroup. Controlling for BMI and C-reactive protein did not affect the results (Table S14).

Figure 4.

10-year incident CAD risk stratified by LLM use. Forest plot shows 10-year incident CAD risk among LLM-treated (blue) and LLM-untreated (red) UK Biobank participants with a severe hypercholesterolemia subtype compared to non-genetic severe hypercholesterolemia. Cox proportional-hazard ratios with 95% confidence intervals are shown.

Discussion

Our results show that subtyping severe hypercholesterolemia by genetic determinants adds substantial information for incident CAD risk stratification compared to observable LDL-C levels. Furthermore, observable LDL-C levels are a poor proxy for the excess CAD risk conferred by genetic determinants as the risk gradient for 20 mg/dL incremental increases in LDL-C is much less than that observed by the presence of a monogenic FH variant or the combination of elevated Lp(a) and polygenic hypercholesterolemia. Our results show that polygenic hypercholesterolemia by itself does not confer substantial excess risk compared to LDL-C measurement and isolated Lp(a) is intermediate between these risk groups.

The investigation of polygenic and monogenic contributions to severe hypercholesterolemia and incident CAD reveals important distinctions in their prevalence and risk profile. Among the genetic determinants examined, the largest fraction (PAF = 11.0%) of severe hypercholesterolemia cases was attributable to polygenic risk in the top decile of the population, consistent with high heritability (h2 = 40-50%) estimates of LDL-C³². However, the rate of incident CAD among those with polygenic hypercholesterolemia was not different from that in individuals with non-genetic hypercholesterolemia, indicating that polygenic risk does not provide additional risk stratification information. In contrast, the monogenic FH and the two-hit subtypes were associated with similar increases in incident CAD risk (HR=2.3 vs. 1.9). This equivalence in risk is supported by data from the Copenhagen General Population Study which identified levels of Lp(a) that were associated with a similar future risk of CAD as those with FH³³.

Genetic determinants are a major cause (ARP > 20%) of severe hypercholesterolemia among individuals with a genetic subtype, and the primary cause (ARP > 60%) in all severe hypercholesterolemia subtypes except elevated Lp(a). Comorbidities, CAD risk factors, medication use, and family history were different between subtypes of severe hypercholesterolemia. The monogenic FH subtype was associated with the highest prevalence of CAD at baseline, consistent with lifelong exposure to elevated LDL-C. Individuals with the monogenic FH subtype had fewer CAD risk factors relative to other subtypes, including a higher prevalence of female sex. The bias towards female sex may reflect premature mortality in men with an FH variant. This study confirms an association between low triglyceride levels and monogenic FH among severe hypercholesterolemia cases³⁴. This finding supports the use of low triglyceride levels for phenotype-based identification of FH in EHR data³⁵. Similarly, the higher family history of heart disease among those with the monogenic FH subtype demonstrates the importance of family history as a component of the FH clinical definition. An association with the elevated Lp(a) subtype and a family history of heart disease was also observed. A high serum concentration of Lp(a) is known to mimic the FH clinical phenotype through increased risk of hypercholesterolemia, CAD, and family history. This reveals a limitation in the specificity of using phenotypic definitions and diagnostic scores to identify FH in population-cohorts³⁶.

The sensitivity analysis by LLM use revealed differences between subtypes that may inform treatment strategies. In the LLM treated group, the most excess CAD risk was observed among those with monogenic FH, further evidence of pervasive undertreatment of this disorder³⁷. Despite the highest proportion of individuals reporting LLM use among genetic subtypes, this result may indicate that monogenic FH patients have treatment initiated too late in life^38,39. Those with monogenic FH have been exposed to lifelong high levels of cholesterol, contributing to higher rates of incident CAD despite statin treatment; those with polygenic inheritance may have less chronic exposure⁴⁰. The two-hit hypercholesterolemia subtype was associated with a two-fold increase of CAD risk in both the LLM-treated and untreated subgroups. Similar to monogenic FH, the excess risk associated with two-hit hypercholesterolemia in the LLM-treated group underscores the importance of early measurement of Lp(a) for risk stratification.

The elevated Lp(a) genetic subtype was defined using a widely accepted threshold of high Lp(a) serum concentration instead of a genetic instrument based on LPA genotypes. A previous study of the UK Biobank demonstrated that elevated Lp(a) defined with LPA genetic scores or serum concentration of Lp(a) have equivalent CAD risk⁵, a finding that was validated in a sensitivity analysis here.

Limitations

LLM use reported in the UK Biobank includes incomplete data on strength, frequency, and timeframe of use, as well as adherence to prescribed therapy, making it difficult to determine the precise therapy regimen the patient was taking when the LDL-C was measured. Adjusted LDL-C estimates are based on a single LDL-C measurement taken at baseline. The cumulative lifetime exposure to severe hypercholesterolemia was not accounted for. Estimates of FH variant prevalence and incident CAD risk in the monogenic FH subtype were likely underestimated here due to 1) a high rate of premature death in patients with an FH variant²⁷ 2) the high prevalent CAD event rate in the monogenic FH subtype, 3) the omission of structural variations in LDLR including loss-of-function copy number variants (CNV) from analyses, and 4) the healthy volunteer selection bias in the UK Biobank⁴¹. The monogenic FH subtype was not stratified by polygenic score or Lp(a) as reported in other studies due to sample size constraints^42-44. Further, we did not analyze the impact on outcomes related to the presence of an FH variant in those with an LDL-C < 190 mg/dl. Due to low sample sizes of non-European populations, analyses were performed exclusively on the European subset of participants. These results will need to be confirmed in populations of non-European ancestry.

Conclusions

Incorporating genetic causes of severe hypercholesterolemia, particularly monogenic FH and Lp(a), into CAD risk assessment can be used to identify those severe hypercholesterolemia patients at highest risk for an imminent coronary artery disease event.

Highlights.

• A large proportion of severe hypercholesterolemia cases were attributable to genetic determinants.

• Among severe hypercholesterolemic participants, genetic subtyping stratified incident CAD risk more than LDL-C-based subtyping.

• Individuals with an FH-causing variant had high incidence of CAD even when treated and would benefit from earlier diagnosis and treatment.

• Those with elevated Lp(a) and two-hit hypercholesterolemia had high incidence of CAD if untreated and would benefit from earlier diagnosis and targeted treatment.

Abbreviations

AHA

American Heart Association

ARP

attributable risk proportion

ASCVD

atherosclerotic cardiovascular disease

BMI

body mass index

BP

blood pressure

CAD

coronary artery disease

CI

confidence interval

CVD

cardiovascular disease

EHR

electronic health record

FH

familial hypercholesterolemia

HDL

high-density lipoprotein cholesterol

HR

hazard ratio

LDL-C

low-density lipoprotein cholesterol

LLM

lipid lower medication

Lp(a)

lipoprotein(a)

OR

odds ratio

P/LP

pathogenic or likely pathogenic

PAD

peripheral artery disease

PAF

population attributable fraction

PRS

polygenic risk score

SD

standard deviation

Data availability statement

The data underlying this article were accessed from the UK Biobank (biobank.ndph.ox.ac.uk). The derived datasets can be generated with permission from the UK Biobank.