Lipoprotein(a) (Lp[a]) has gained attention as a heritable coronary artery disease (CAD) risk factor and therapeutic target. Two genetic variants in the LPA gene have been reported to influence Lp(a) levels and increase CAD risk. The aim of this study was to prospectively test these variants for their associations with Lp(a) and CAD risk. Participants (n = 1,400) in the Intermountain Heart Collaborative Study Registry who had Lp(a) cholesterol levels determined at coronary angiography were genotyped for rs3798220 and rs1045587 in LPA. Variants were detected by Taqman polymerase chain reaction. Chi-square and linear and logistic regression tests were used as appropriate among genotypes for Lp(a) and angiographic CAD. Age averaged 63 years; 65% were men; and severe CAD was present in 57%, mild CAD in 12%, and no CAD in 31%. Minor allele frequencies were 0.023 for rs3798220 and 0.090 for rs10455872. In multivariate modeling, only rs10455872 (odds ratio [OR] 2.33, 95% confidence interval [CI] 1.67 to 3.33, p = 1.75 × 10 −9 ) and rs3798220 (OR 1.99, 95% CI 0.99 to 4.00, p = 0.065) contributed to the prediction of elevated Lp(a) cholesterol. Lp(a) cholesterol was weakly associated with CAD (OR 1.17, 95% CI 1.00 to 1.37, p = 0.055). Rs10455872 strongly predicted prevalent CAD (per allele OR 1.43, 95% CI 1.07 to 1.91, p = 0.0172); the effect size for the rare rs3798220 variant was similar (dominant OR 1.47, 95% CI 0.81 to 2.67, p = 0.20), but power was limited to demonstrate significance. The combined genotype explained only a small percentage (≤4%) of variability in Lp(a) cholesterol and prevalence of angiographic CAD. In conclusion, heritable contributions of LPA rs10455872 and rs3798220 to Lp(a) cholesterol levels and to angiographic CAD were prospectively assessed in this study. The percentage of intersubject variability in Lp(a) cholesterol and the percentage of prevalent CAD explained were small.
Lipoprotein(a) (Lp[a]) is a complex circulating particle consisting of a large glycoprotein (apolipoprotein[a]) molecule, structurally resembling plasminogen, attached by a disulfide bond to an apolipoprotein B molecule. Accumulating evidence suggests that Lp(a) is an independent risk factor for coronary artery disease (CAD), peripheral artery disease, and stroke in whites and especially in blacks, who have a larger range of Lp(a), leading some societies to endorse Lp(a) measurement to refine risk in selected populations. Preclinical models suggest that Lp(a) is prothrombotic, proinflammatory, and proatherogenic, and Lp(a) enters the arterial intima in humans. Lp(a) concentrations are highly heritable, with a broad, bimodal concentration range. Genomewide association studies have discovered 2 common variants of the LPA gene, which encodes Lp(a) lipoprotein, that are associated with Lp(a) lipoprotein and increased CAD risk. Despite these findings, the incremental value of Lp(a) measurement applied to cardiovascular risk refinement remains controversial. Moreover, past experience with genetic studies suggests the need for multiple replications in distinct populations.
Methods
To better define their clinical value for cardiovascular risk assessment, we tested 2 LPA variants (rs3798220 and rs10455872) for quantitative associations with Lp(a) levels and angiographic CAD in a prospectively enrolled and tracked patient cohort in the Intermountain Heart Collaborative Study Registry. On the basis of recent reports, we hypothesized (1) that Lp(a) cholesterol concentrations would be significantly associated with angiographic CAD and (2) that these 2 LPA single-nucleotide polymorphisms (SNPs) would be associated with Lp(a) concentrations and the risk for angiographic CAD. Our further intent was to prospectively quantify these associations in an independent and high-risk patient population with precise (angiographic) CAD phenotyping to determine the potential utility of genotyping for clinical CAD risk assessment and decision making.
Participants (n = 1,400) in the Intermountain Heart Collaborative Study Registry who had Lp(a) cholesterol levels determined at the time of coronary angiography were entered into the study and were genotyped for rs3798220 and rs10455872 in LPA. All participants in the Intermountain Heart Collaborative Study Registry have given written informed consent for the use of plasma and deoxyribonucleic acid as well as demographic, medical, and angiographic characteristics in approved cardiovascular investigations. The Intermountain Heart Collaborative Study Registry and the present study were approved by the Intermountain Medical Center Institutional Review Board.
The Vertical Auto Profile test (Atherotech Diagnostic Laboratories, Birmingham, Alabama) was used to determine Lp(a) cholesterol. Unlike Lp(a) immunoassays, the Vertical Auto Profile test measures Lp(a) cholesterol directly and does not measure total particle mass. Normal values for Vertical Auto Profile Lp(a) cholesterol are <10 mg/dl. (As a result, Lp[a] cholesterol values are less than immunoassay numeric values.) An overview of previous study comparisons indicates that Lp(a) immunoassays and Lp(a) cholesterol correlate categorically (i.e., high vs normal or low) approximately 80% of the time (M. Cobble, MD, Atherotech, personal communication, August 27, 2012).
Genetic variants were detected by Taqman real-time polymerase chain reaction. Assays were obtained from Applied Biosystems (Foster City, California). The allele discrimination protocol was run on a Vii 7, real-time thermal cycler (Applied Biosystems). Five percent of samples were regenotyped for quality assurance. Successful genotyping was obtained for 94.6% of samples.
Coronary angiography was performed for clinical indications by experienced, board-certified cardiologists, who also visually assessed the presence and severity of CAD at the time of angiography and entered the results into a standardized electronic medical record form. Determination was blinded to Lp(a) cholesterol and genotype results. Three CAD categories were prospectively defined: none or minimal (<10% stenosis), mild to moderate (10% to 69% stenosis), and severe (≥70% stenosis in ≥1 major artery or a major [≥2 mm in diameter] branch).
Two methods were used to test for associations with Lp(a) cholesterol: one based on continuous Lp(a) cholesterol concentrations and another based on categorical Lp(a) cholesterol (normal, <10 mg/dl; abnormal, ≥10 mg/dl). Multivariate linear regression was used to test for associations between Lp(a) cholesterol (which was square-root transformed to achieve normality) and the LPA variants (additive model for rs10455872, dominant model for rs3798220) while adjusting for age, gender, history of hyperlipidemia (highly correlated with lipid-lowering medication or statin use), history of CAD, previous myocardial infarction, diabetes, smoking history, family history of early (age <60 years) CAD, and the natural logarithm of body mass index. Univariate chi-square tests and multivariate logistic regression were used as appropriate to test for associations between normal versus abnormal Lp(a) cholesterol and the LPA variants (additive model for rs10455872, dominant model for rs3798220). The multivariate model was again adjusted for age, gender, history of hyperlipidemia, history of CAD, previous myocardial infarction, diabetes, smoking history, family history of early CAD, and the natural log of body mass index. Chi-square tests were used to examine univariate associations between the variants and angiographic CAD, and logistic regression was applied to evaluate their associations in multivariate modeling, which adjusted for age, gender, hyperlipidemia, hypertension, diabetes, family history of premature CAD, smoking history, and the natural log of body mass index. In the latter models, CAD was defined as angiographic diameter stenosis ≥70% and non-CAD as the absence of visual angiographic stenosis (i.e., diameter stenosis <10%) and absent history of CAD, myocardial infarction, or coronary revascularization. (Those with mild to moderate CAD [diameter stenosis 10% to 69%] were excluded as an indeterminate phenotype for purposes of the present study.)
Results
Demographics of the study population are listed in Table 1 . Age averaged 63 years; 65% were men; and severe CAD was present in 57%, mild CAD in 12%, and no CAD in 31%. Minor allele frequencies were 0.023 for rs3798220 and 0.090 for rs10455872 ( Table 2 ). The 2 SNPs were minimally associated with each other (r = −0.045, p = 0.095).
| Variable | Value |
|---|---|
| Age (yrs) | 62.8 ± 12.8 |
| Men | 915 (65.4%) |
| Race | |
| White | 1,261 (90.1%) |
| Unknown race | 70 (5.0%) |
| Hispanic | 30 (2.1%) |
| Asian and Pacific Islander | 17 (1.2%) |
| African American | 11 (0.8%) |
| Other | 11 (0.8%) |
| Body mass index (kg/m 2 ) | 29.1 ± 6.6 |
| Hyperlipidemia by history | 771 (55.1%) |
| Hypertension by history | 842 (60.1%) |
| Diabetes mellitus | 314 (22.4%) |
| Smokers | 244 (17.4%) |
| Previous CAD | 583 (41.6%) |
| Previous myocardial infarction | 87 (6.2%) |
| Previous heart failure | 255 (18.2%) |
| Previous stroke | 52 (3.7%) |
| Presentation with acute myocardial infarction | 176 (12.6%) |
| Presentation with probable unstable angina pectoris | 860 (61.4%) |
| Stable angina pectoris or non–chest pain presentation | 364 (26.0%) |
| Systolic blood pressure (mm Hg) | 145.9 ± 24.2 |
| Diastolic blood pressure (mm Hg) | 82.3 ± 13.1 |
| Total cholesterol (mg/dl) | 173.0 ± 41.8 |
| High-density lipoprotein cholesterol (mg/dl) | 41.2 ± 13.3 |
| Low-density lipoprotein cholesterol (mg/dl) | 102.5 ± 35.7 |
| Triglycerides (mg/dl) | 131 (94–183) |
| Lp(a) cholesterol (mg/dl) | 5.57 ± 3.64 |
| Extent of CAD at entry angiography | |
| None | 434 (31.0%) |
| Mild to moderate ∗ | 165 (11.8%) |
| Severe | 801 (57.2%) |
∗ Mild to moderate CAD (i.e., 10% to 69% maximal coronary artery stenosis) was excluded from association analyses as an indeterminate phenotype.
| SNP | Chromosome | Position/Type | Gene | Minor Allele | Minor Allele Frequency | No. Genotyped |
|---|---|---|---|---|---|---|
| rs3798220 | 6 | 160880877 missense | LPA | C | 2.3% | 1,397 |
| rs10455872 | 6 | 161010118 intronic | LPA | G | 9.0% | 1,358 |
Univariate associations of genotype with Lp(a) cholesterol are listed in Table 3 . Carriage of ≥1 minor alleles for each SNP predicted higher levels of Lp(a) cholesterol. The power of significance testing for the rare rs3798220 was limited by small numbers of carriers, resulting in statistically modest trend values in univariate testing (p = 0.15). Carriage of the more common rs10455872 SNP was associated with a greater mean increase in Lp(a) cholesterol and a highly significant result (p = 2 × 10 −10 ). In multivariate linear regression modeling, forcing in the 2 SNPs and 9 clinical variables, rs10455872 (p = 1.75 × 10 −9 ) and (marginally) rs3798220 (p = 0.065) contributed to the prediction of elevated Lp(a) cholesterol, whereas none of the other variables contributed significantly ( Table 4 ). Nevertheless, the combined multivariate model explained only 3.75% of the variation in Lp(a) levels. These genetic associations were unchanged if the few blacks were excluded or if only whites were included.
| Lp(a) Cholesterol | rs3798220 | rs10455872 | |||
|---|---|---|---|---|---|
| 0 | 1–2 | 0 | 1 | 2 | |
| (n = 1,334) | (n = 63) | (n = 1,134) | (n = 203) | (n = 21) | |
| Mean (mg/dl) | 5.53 | 6.24 | 5.29 | 6.78 | 8.86 |
| SD | 3.63 | 3.86 | 3.46 | 3.97 | 5.30 |
| 25th percentile | 3 | 3 | 3 | 3 | 5 |
| Median (mg/dl) | 4 | 5 | 4 | 6 | 8 |
| 75th percentile | 7 | 9 | 7 | 9 | 13 |
| Linear regression p value | 0.15 ∗ | 2.0 × 10 −10 | |||
| R 2 | 0.0015 ∗ | 0.0293 (combined: 0.0358) | |||
∗ Combining CT and CC, given n = 1 for CC. Nonparametric testing used Wilcoxon’s and Kruskal-Wallis tests (for rs3798220 and rs10455872, respectively). Linear regression analyses across genotypes used the square root of Lp(a) cholesterol.
| Variable | Parameter Estimate | SE | t Value | p Value |
|---|---|---|---|---|
| Intercept | 2.42078 | 0.35858 | 6.75 | <0.0001 |
| rs10455872 | 0.28967 | 0.048 | 6.03 | <0.0001 |
| rs37988220 | 0.17047 | 0.09399 | 1.81 | 0.07 |
| Hyperlipidemia | 0.07744 | 0.0464 | 1.67 | 0.0953 |
| Diabetes | −0.06172 | 0.05129 | −1.2 | 0.229 |
| ln(body mass index) | −0.1066 | 0.09779 | −1.09 | 0.2759 |
| Age | 0.00173 | 0.00167 | 1.04 | 0.3003 |
| Male | −0.04306 | 0.04285 | −1.01 | 0.3151 |
| Family history of CAD | 0.03968 | 0.04293 | 0.92 | 0.3555 |
| Previous myocardial infarction | −0.07958 | 0.08774 | −0.91 | 0.3646 |
| Hypertension | −0.0344 | 0.04557 | −0.75 | 0.4505 |
| Previous CAD | 0.03138 | 0.04495 | 0.7 | 0.4853 |
| Smoking history | −0.00741 | 0.0532 | −0.14 | 0.8892 |
Additional categorical analyses tested the association of these SNPs and clinical variables with an abnormal Lp(a) cholesterol level (i.e., ≥10 mg/dl). Results were similar to those of linear modeling: with univariate analysis, using chi-square trend tests, p = 0.20 for rs3798220 (dominant model) and p = 7.3 × 10 −8 for rs10455872 (additive model). In multivariate logistic regression modeling (forcing in the 9 clinical variables), the odds ratio (OR) for elevated Lp(a) cholesterol was 2.30 per allele (95% confidence interval [CI] 1.699 to 3.13, p = 1.22 × 10 −7 ) for rs10455872 and 1.85 for dominant carriage (95% CI 0.95 to 3.60, p = 0.068) of rs3798220. Again in this model, none of the 9 clinical variables contributed significantly.
Lp(a) cholesterol itself was weakly associated with CAD (OR 1.69 per square root of Lp[a] mg/dl, 95% CI 1.00 to 1.37, p = 0.055). Univariately, rs10455872 was significantly associated with CAD (per allele OR 1.43, 95% CI 1.07 to 1.91, p = 0.0172); the effect size for the rare rs3798220 variant was similar (dominant OR 1.47, 95% CI 0.81 to 2.67, p = 0.20), but power was limited to demonstrate significance. These associations were not attenuated when adjusting for sets of clinical factors (see Figure 1 ). The final multivariate risk model, which adjusted for other clinical factors including Lp(a), is listed in Table 5 , with adjusted ORs for rs10455872 and rs3798220 of 1.77 and 1.74, respectively. However, the combined genotypes explained only a small percentage (3.1%) above chance in the prediction of prevalent CAD (C-statistic = 0.531). Additional predictors of prevalent angiographic CAD included age, male gender, family history of CAD, hyperlipidemia, hypertension, and smoking history. Genotype did not predict prevalent myocardial infarction apart from CAD (data not shown).
