An 8-SNP LDL Cholesterol Polygenic Score: Associations with Cardiovascular Risk Traits, Familial Hypercholesterolemia Phenotype, and Premature Coronary Heart Disease in Central Romania

Abstract

Familial hypercholesterolemia (FH) is the most significant inherited risk factor for coronary heart disease (CHD). Current guidelines focus on monogenic FH, but the polygenic form is more common and less understood. This study aimed to assess the clinical utility of an 8-SNP LDLC polygenic score in a central Romanian cohort. The cohort included 97 healthy controls and 125 patients with premature (P)CHD. The weighted LDLC polygenic risk score (wPRS) was analyzed for associations with relevant phenotypic traits, PCHD risk, and clinical FH diagnosis. The wPRS positively correlated with LDLC and DLCN scores, and LDLC concentrations could be predicted by wPRS. A trend of increasing LDLC and DLCN scores with wPRS deciles was observed. A +1 SD increase in wPRS was associated with a 36% higher likelihood of having LDLC > 190 mg/dL and increases in LDLC (+0.20 SD), DLCN score (+0.16 SD), and BMI (+0.15 SD), as well as a decrease in HDLC (-0.14 SD). Although wPRS did not predict PCHD across the entire spectrum of values, individuals above the 90th percentile were three times more likely to have PCHD compared to those within the 10th or 20th percentiles. Additionally, wPRS > 45th percentile identified "definite" clinical FH (DLCN score > 8) with 100% sensitivity and 45% specificity. The LDLC polygenic score correlates with key phenotypic traits, and individuals with high scores are more likely to have PCHD. Implementing this genetic tool may enhance risk prediction and patient stratification. These findings, the first of their kind in Romania, are consistent with the existing literature.

Keywords: LDL cholesterol; coronary heart disease; familial hypercholesterolemia; phenotype; polygenic score; single-nucleotide polymorphism.

Figure 1

Relationship between PRSs and phenotypic traits: eLDLC and DLCN scores. All figures display median values with error bars representing the interquartile range. (A): comparison of median aPRS between HC, nFH, and dFH groups. (B): comparison of median wPRS between HC, nFH, and dFH groups. (C): trend analysis of eLDLC concentrations based on wPRS decile in the whole cohort. (D): trend analysis of DLCN scores based on wPRS decile in the whole cohort. (E): trend analysis of eLDLC concentrations based on wPRS decile in the HC group. (F): trend analysis of eLDLC concentrations based on wPRS decile in the PCHD group. The trend in (C–F) was evaluated using the Jonckheere–Terpstra statistical test. Abbreviations: dFH, definite FH (DLCN score > 8); eLDLC, estimated pre-treatment LDL cholesterol; FH, familial hypercholesterolemia; HC, healthy control; nFH, non-definite FH (DLCN score ≤ 8); NS, nonsignificant; PCHD, premature coronary heart disease (patients); PRS, polygenic risk score (additive/a or weighted/w).

Figure 2

(A–C) represent eLDLC-wPRS regression equations for the whole cohort, PCHD group, and HC group, respectively. (D) illustrates phenotypic trait variations with wPRS: variables are color-coded on a gradient from green (low) to red (high), with the exception of HDLC, which is color-coded from red (low) to green (high). Abbreviations: BMI, body mass index; DLCN, Dutch Lipid Clinic Network (score); eLDLC, estimated pre-treatment LDL cholesterol; HC, healthy controls; PCHD, premature coronary heart disease (patients); wPRS, weighted polygenic risk score.

Figure 3

The performance of wPRS in predicting premature coronary heart disease (PCHD). (A): ROC analysis to differentiate between HC and PCHD based on wPRS. (B): comparison of PCHD prevalence between those with the lowest (first decile) and highest (tenth decile) wPRS scores. Abbreviations: AUC, area under the curve; HC, healthy control; wPRS, weighted polygenic risk score.

Figure 4

Sample of a multilayer perceptron artificial neural network algorithm for predicting PCHD. “History” stands for family history of CHD. “Alcohol” stands for alcohol consumption. Abbreviations: AHT, arterial hypertension; BMI, body mass index; eLDLC, estimated pre-treatment LDL cholesterol; HC, healthy controls; PCHD, premature coronary heart disease.

Figure 5

Normalized importance of variables for predicting PCHD. Both models display the average normalized importance calculated after 100 iterations of multilayer perceptron neural network analysis. Medians are shown with error bars representing the 10–90% value interval. “History” stands for family history of CHD. “Alcohol” stands for alcohol consumption. Abbreviations: AHT, arterial hypertension; BMI, body mass index; eLDLC, estimated pre-treatment LDL cholesterol; PCHD, premature coronary heart disease; wPRS, weighted polygenic risk score.

Figure 6

The performance of wPRS in predicting clinical FH. (A): whole cohort trend analysis of wPRS based on DLCN score categories. (B): whole cohort comparison of wPRS scores between those with “unlikely FH” (DLCN < 3) and the other participants. (C): whole cohort ROC analysis to differentiate between those with “unlikely FH” (DLCN < 3) and the other participants, based on wPRS. (D): ROC analysis to differentiate between dFH patients and HC based on wPRS. (E): ROC analysis to differentiate between dFH and nFH patients based on wPRS. (F): ROC analysis to differentiate between dFH patients and the other participants based on wPRS. Classification for DLCN score categories in panel A is as follows: unlikely (DLCN < 3), possible (DLCN 3-5), probable (DLCN 6-8), and definite (DLCN > 8). Abbreviations: AUC, area under the curve; dFH, definite FH (DLCN score > 8); DLCN, Dutch Lipid Clinic Network (score); FH, familial hypercholesterolemia; HC, healthy controls; nFH, non-definite FH (DLCN score ≤ 8); PCHD, premature coronary heart disease (patients); wPRS, weighted polygenic risk score.