Background: The relationship of lipoprotein (a) [Lp(a)] concentrations with risk of coronary heart disease needs clarification, especially for threshold values for increased risk and for possible interactions with LDL-cholesterol concentrations and apolipoprotein (a) [apo(a)] size polymorphism. This study was designed to examine the ability of baseline Lp(a) concentration and apo(a) size to predict future severe angina pectoris in apparently healthy men.
Methods: Baseline Lp(a) concentration and apo(a) size were determined in 195 men who subsequently developed angina and in 195 men who remained free of cardiovascular disease for 5 years.
Results: Cases had higher median Lp(a) concentrations than did controls (30.6 vs 22.5 nmol/L; P = 0.02). Lp(a) concentration was predictive of angina [relative risk (RR) from lowest to highest quintiles: 1.0, 1.5, 1.0, 1.8, and 2.6; P for trend = 0.015]. The increased risk was ∼4-fold (95% confidence interval, 1.4- to 11-fold) among men who had Lp(a) above the 95th percentile (>158 nmol/L). Men with Lp(a) concentrations in the highest quintile and LDL-cholesterol concentrations >1600 mg/L had a 12-fold increased risk (95% confidence interval, 1.5- to 43-fold). Small apo(a) size isoforms also significantly predicted risk of angina (RR for lowest quintile = 4.1; P for trend = 0.004). When the independent effect of Lp(a) concentration and apo(a) size was assessed by including them in the same multivariate model, only the association between apo(a) size and risk remained significant.
Conclusions: High Lp(a) predicts risk of angina, and the risk is substantially increased with high concomitant LDL-cholesterol. Small apo(a) size predicts angina with greater strength and independence than Lp(a) concentration.
Lipoprotein(a) [Lp(a)]1 is composed of LDL particles additionally containing apolipoprotein(a) [apo(a)], a glycoprotein highly heterogeneous in size, covalently linked to apo B. Apo(a), because of its homology to plasminogen, may add thrombogenicity properties to the atherogenicity of LDL. Although there are variations in findings among prospective studies of Lp(a) as a risk factor for coronary heart disease (CHD), which may in part be attributed to methodologic differences, several studies, including a recent metaanalysis, have demonstrated that CHD risk is increased as Lp(a) concentration increases (1)(2).
Several aspects of the relationship between Lp(a) and CHD remain unclear. Lp(a) is unique among the lipid risk factors in that its concentrations vary from <0.1 to >180 nmol/L. Unlike other lipid risk factors, the Lp(a) risk may not be continuous but increases at high concentrations, such as the top 25% or above (3)(4)(5)(6)(7)(8). Additionally, several studies have found that Lp(a) concentration contributed to risk only when LDL-cholesterol (LDL-C) was concomitantly increased (4)(5)(7)(8)(9), although other studies did not find such an interaction (6)(10)(11). These two issues are quite relevant to potential clinical use of Lp(a).
Finally, small apo(a) isoform size may also be associated with increased CHD risk. However, discordant results have been reported concerning the contribution of apo(a) size to CHD risk independent of Lp(a) concentrations (3)(4)(11)(12)(13)(14)(15)(16)(17)(18). Substantial methodologic variation in determining Lp(a) concentrations and apo(a) size may explain the difference in findings.
In the Physicians’ Health Study, we previously reported no association between Lp(a) concentrations and risk of future myocardial infarction, stroke, or peripheral vascular disease (19)(20)(21). In these as in other studies, Lp(a) values were determined by methods that underestimated concentrations of Lp(a) with small apo(a) sizes, i.e., the type of Lp(a) that is most strongly associated with CHD. In the present study, we continued our evaluation of the relationship between Lp(a) and CHD in this cohort by determining Lp(a) concentration and apo(a) size in participants who developed angina pectoris. We used highly standardized and validated methods for the determination of Lp(a) concentration and apo(a) isoform size.
Materials and Methods
We performed a nested case–control analysis of incident angina within the Physicians’ Health Study, a randomized double-blinded, placebo-controlled primary prevention trial of aspirin and β-carotene for the prevention of cardiovascular disease and cancer among 22 071 male physicians (40–84 years of age in 1982) and without a previous history of myocardial infarction, stroke, transient ischemic attacks, unstable angina, cancer (except nonmelanoma skin cancer), recurrent renal or liver disease, peptic ulcer disease or gout, contraindications to aspirin use, or current use of aspirin or other platelet-active agents or vitamin A supplements (22). Study participants completed two questionnaires before randomization in 1982, providing information on smoking history, alcohol intake, physical activity, and disease diagnoses. Medical history was updated at 6 months, 12 months, and annually thereafter.
Men with angina and additional evidence of severe coronary atherosclerosis were selected as cases. Men with a baseline history of angina or coronary revascularization and those with myocardial infarction during follow-up before diagnosis of angina were excluded from this analysis. The evidence was coronary artery bypass surgery, significant occlusion (>50%) at coronary angiography, an exercise tolerance test with or without thallium, thallium scan, or exercise radionuclide ventriculogram. After reviewing additional evidence in all available medical records, the Endpoints Committee of Physicians, which was blinded to treatment status, confirmed the reported angina pectoris.
Lp(a) data were available for 195 of the 218 (90%) all-incident angina cases in the first 5 years of follow-up. Each of the 195 cases was matched with a control who was free from angina or other cardiovascular disease at the time of the case’s diagnosis. Matching criteria were age (± 1 year), smoking status (never, past, current), and length of time since randomization (in 6-month intervals).
Before randomization, blood samples were collected, and plasma was aliquoted and stored at −80 °C from 14 916 participants (68%) until analysis. There were no freeze–thaw cycles between initial storage in 1982–1984 until current testing.
Triglycerides and total cholesterol, LDL-C, and HDL-cholesterol (HDL-C) were measured in a CDC-certified laboratory for lipid testing. Lp(a) protein concentration was measured by a double monoclonal antibody-based ELISA. The detecting monoclonal antibody in this assay is directed to a unique epitope in apo(a) kringle 4 type 9; therefore, the assay is independent of apo(a) size polymorphisms (9). The method is recognized as the “reference method” for Lp(a) (23). Lp(a) concentrations are expressed in nmol/L. Concomitantly, Lp(a) total lipoprotein mass was measured by a latex-enhanced nephelometric commercial method (Dade Behring), and values are expressed in mg/L. The apo(a) size isoforms were determined by high-resolution sodium dodecyl sulfate-agarose gel electrophoresis followed by immunoblotting (24), and the apo(a) size isoforms are designated by the relative number of kringle 4 motifs. In individuals heterozygous for apo(a) isoform size, when one of the two isoforms was predominant, that isoform was used for statistical analysis. When the two isoforms appeared to be present in the same proportion, the mean of the kringle 4 number of the two isoforms was used.
The significance of any difference in means between cases and controls was tested by use of a paired Student t-test and, for proportions, the McNemar test. Because of the skewed distribution of Lp(a) and triglyceride concentrations and apo(a) size, we used natural-log-transformed data for the paired t-test and calculated Spearman correlation coefficients. The distribution of plasma concentrations in the controls was used to define quintiles. We used conditional logistic regression accounting for the matching variables of age and smoking to compute odds ratios as estimates of relative risks (RRs) according to quintile distribution of the controls. On the basis of an a priori hypothesis, we also assessed the RRs using baseline Lp(a) concentrations below and above the 95th percentile of the control value. To test for the trend of the association, we constructed an indicator variable for the quintiles of Lp(a) concentration and apo(a) size by assigning the quintile-specific median Lp(a) concentration or apo(a) size as the value for each quintile. The P value for trend across quintiles is the P value that is generated when this indicator variable is entered into the model as a continuous variable. We tested for interactions between Lp(a) concentration or apo(a) size and LDL-C concentration, using those with Lp(a) concentrations in the lowest quintile or apo(a) size in the highest quintile and LDL-C <1600 mg/L as the reference group. In addition, we tested for an interaction between Lp(a) concentration and apo(a) size, using those with Lp(a) in the lowest quintile and apo(a) size in the higher quintiles (quintiles 2–5) as the reference group. Multivariate adjustment was made for aspirin treatment assignment, triglycerides, LDL-C and HDL-C concentrations, and other baseline risk factors, including body mass index, personal history of diabetes and hypertension, and family history of myocardial infarction before 60 years. RRs are presented with 95% confidence intervals (CI), and all P values are two-tailed.
association of Lp(a) concentration with risk of angina
The median Lp(a) concentration at baseline was significantly higher in men who subsequently developed angina than in those who did not (30.6 vs 22.5 nmol/L; P = 0.02; Table 1⇓ ). The mean apo(a) size was smaller among cases than controls (P = 0.04). Lp(a) concentration was associated with increased RR for angina (Table 2⇓ ). This association was similar after adjusting for aspirin and CHD risk factors but strengthened after further controlling for lipid risk factors. Because it has previously been shown to be associated with significantly increased risk (6), we examined the RR of future angina at the 95th percentile cut-point of the control distribution. Above this cut-point (158.8 nmol/L), the RRs were 3.3 (95% CI, 1.4–7.9, age- and smoking-adjusted) to 3.6 (95% CI, 1.3–10.3, also adjusted for risk factors; Fig. 1⇓ ) compared with those with Lp(a) concentration below the 20th percentile. We found that Lp(a) and LDL-C concentrations synergistically interacted to increase risk. Individuals with LDL-C ≥1600 mg/L and a Lp(a) concentration in the highest quintile had a markedly increased risk (age- and smoking-matched RR = 12.2; 95% CI, 3.5–42; P for interaction = 0.06; Fig. 2A⇓ ). A similar pattern was seen after adjustment for other CHD risk factors.
No statistically significant correlation between Lp(a) concentration and any of the examined lipids was seen in both groups [Spearman correlation coefficient (r) between LDL-C and Lp(a) concentrations = 0.05 (P = 0.49) for cases and r = 0.001 (P = 0.99) for controls], except for a modest inverse correlation between concentrations of Lp(a) and triglycerides among cases (r = −0.18; P <0.015).
To evaluate the impact of assay differences on interpretation of clinical data, we also measured the concentrations of Lp(a) mass by a commercially available method. Overall, we found a good correlation of values between the ELISA and the commercial method through much of the range (r = 0.97; P = 0.0001). However, the median Lp(a) mass measured by this commercial method was not significantly different in cases than in controls (114 vs 96 mg/L; P = 0.16), and the association with angina was also not significant [age- and smoking-adjusted RRs from the 1st to the 5th quintile of Lp(a), 1.0 (referent), 0.94, 0.77, 1.14, and 1.33; P for trend = 0.14; Table 3⇓ ]. Additional controlling for lipids and other risk factors slightly increased the association: the corresponding RRs were 1.0 (referent), 1.35, 1.05, 1.24, and 1.88 (P for trend = 0.11). In addition, individuals with LDL-C concentrations ≥1600 mg/L and Lp(a) in the highest quintile (≥300 mg/L) had a significantly increased risk (age- and smoking-matched RR = 11.9, 95% CI, 3.2–45; P for interaction = 0.08) compared with those with LDL-C <1600 mg/L and Lp(a) in the lowest quintile (<26 mg/L).
association of apo(a) size with risk of angina
Small apo(a) size was not significantly associated with angina in analyses controlling for age and smoking (RR = 1.87 for 1st vs 5th quintile; P for trend = 0.10). However, after controlling for lipids and other CHD risk factors, the association became stronger and significant (RR = 4.09 for 1st vs 5th quintile; P for trend = 0.004; Table 4⇓ ). No statistically significant correlation between apo(a) size and lipids, lipoproteins, and other CHD risk factors was seen in the control group. In a separate analysis, we grouped the cases and controls into small (≤22 kringle 4 repeats) and large (>22 repeats) apo(a) phenotypes. The RRs were 1.43 (95% CI, 0.96–2.12) in the matched model and 1.98 (95% CI, 1.18–3.32) in the multivariate adjusted model.
We performed additional analyses to determine which of the covariates were responsible for the association between apo(a) size and risk of angina. After we adjusted for LDL-C, HDL-C, and triglyceride concentrations separately, the RRs (between the opposite quintiles) changed to 2.08 (95% CI, 0.99–4.38), 2.20 (95% CI, 1.06–4.57), and 2.90 (95% CI, 1.31–6.44), respectively, suggesting that the three lipids are the strongest confounders in the model. Adjusting for nonlipid risk factors did not affect the risks. Only individuals with increased LDL-C concentrations and apo(a) size in the lowest quintile had a markedly increased risk (age- and smoking-matched RR = 8.9; 95% CI, 2.6–30; P for interaction = 0.18; Fig. 2B⇑ ). A similar pattern was seen after adjustment of CHD risk factors.
relation of Lp(a) concentration and apo(a) size to risk of angina
Lp(a) concentration and apo(a) size were inversely correlated (r = −0.55; P = 0.0001). After adjustment for apo(a) size, the association between Lp(a) concentration and risk became null (Table 2⇑ ). In contrast, the RR among men with apo(a) size in the lowest quintile became stronger (from 4.9 to 6.37) after adjustment for Lp(a) concentration (Table 4⇑ ). These observations suggest that the association between Lp(a) concentrations and risk of angina could be completely explained by the variation in apo(a) size.
We further demonstrated this combined effect of the two analytes by cross-tabulating Lp(a) concentrations [lower (quintiles 1–4) vs higher (quintile 5)] and apo(a) size [smaller (quintile 1) vs larger (quintiles 2–5)] in a multivariate model. As shown in Fig. 3⇓ , small apo(a) size increased risk when Lp(a) concentration was low to average or high, whereas Lp(a) concentrations had no effect on risk beyond that of apo(a) size. We found that 66% of the cases and 75% of the controls were in the category of larger apo(a) size (quintiles 2–5) and lower Lp(a) concentration (quintiles 1–4), whereas 26% of the cases and 15% of the controls were in the smaller apo(a) size (quintile 1) and higher Lp(a) concentration (quintile 5), leading to a total of 90–92% concordance in these two categories and suggesting that we could reasonably classify individuals into high vs low risk by measuring either analyte.
In this study, we found that Lp(a) concentration, when measured by an assay unaffected by apo(a) size heterogeneity, predicts future angina. The increased risk was primarily associated with higher concentrations [twofold risk in men with Lp(a) concentrations in the 80th percentile and an almost fourfold risk among those with Lp(a) above the 95th percentile]. Men with Lp(a) concentrations in the 80th percentile and concomitant high LDL-C concentrations (>1600 mg/L) had a >12-fold increased risk. Small apo(a) was also a strong predictor of angina in individuals with high LDL-C, and the effect of apo(a) size eclipsed that of Lp(a) concentration in multivariate analysis. That is, when the independent effects of apo(a) size and Lp(a) concentration were evaluated by including them in the same multivariate model, only apo(a) size remained a significant strong independent predictor of future angina. However, because determination of apo(a) size is not readily available for clinical use, a high Lp(a) concentration can identify individuals at risk because of its inverse correlation with apo(a) size. The reference Lp(a) method gave higher RRs for Lp(a) concentration than the commercial method, which was able to identify most people at increased risk only at very high Lp(a) concentrations (above the 95th percentile, or >600 mg/L).
Many studies have also found that mainly high Lp(a) concentrations are associated with increased risk of vascular disease (3)(4)(5)(6)(7)(8)(25)(26), with the top 20–25% of Lp(a) associated with two- to threefold increased risk. In the Caerphilly study in Wales (6), twofold increased risk was associated with the top 5% of the range. However, considering the inaccuracy of most of the methods used to measure Lp(a), a threshold Lp(a) concentration to define a high CHD risk needs to be established.
Our study found that Lp(a) concentration strongly contributed to CHD risk when LDL-C was concomitantly increased, consistent with several other studies (4)(5)(7)(9). The PROCAM study found almost threefold increased CHD risk for high Lp(a) when LDL-C was >1580 mg/L (7), and the PRIME study found fourfold increased risk with high Lp(a) at LDL-C concentrations >1630 mg/L (5). However, some inconsistent results have also been reported (6)(10)(11). The Bruneck study demonstrated that the association of Lp(a) concentrations and early carotid stenosis was confined to those with increased LDL-C (>1280 mg/L), whereas there was no interaction with LDL-C concentrations in the association with advanced coronary atherosclerosis (4). Larger studies performed using well-validated methods are needed to assess whether high Lp(a) concentrations indeed interact with high LDL-C. If it is confirmed, the Lp(a) concentration could be used to identify a small group of people with high LDL-C with especially high CHD risk who might benefit from more intensive risk management.
Recently, it has been suggested that increased Lp(a) concentrations and small apo(a) isoforms may synergistically contribute to Lp(a) pathogenicity (27). The Bruneck study showed that small apo(a) sizes are a strong risk predictor of advanced stenotic atherosclerosis when associated with increased Lp(a). However, for early stages of atherosclerosis, Lp(a) concentrations but not apo(a) isoforms were found to be predictive (4). Using multiple regression analysis, some studies found that apo(a) size was more important than Lp(a) concentrations (13), whereas others found that only Lp(a) concentrations were predictive (3)(11)(14)(15)(16)(17)(18). We found that small apo(a) size is an independent predictor of angina regardless of Lp(a) concentration and that no additivity or synergism was present. Thus, studies are not in agreement on the relative roles of Lp(a) concentration and apo(a) size in CHD. Because these variables are moderately correlated, resolution of this issue would require well-designed, large clinical studies using validated methods.
The size heterogeneity of apo(a) among individuals presents a unique challenge in the determination of Lp(a) concentrations. Antibodies that recognize the kringle 4 type 2 repeated epitopes will have variable immunoreactivity depending on the size of apo(a) (28)(29). Monoclonal antibodies directed to apo(a) antigenic sites in other than kringle 4 type 2 have been suggested as a possible solution to this problem. As indicated earlier, the ELISA used in this study uses a monoclonal antibody specifically directed to an epitope present in apo(a) kringle 4 type 9, whereas the commercial latex method uses polyclonal antibodies that recognize kringle 4 type 2. Therefore, this ELISA method has been considered accurate and recognized by experts in the field as the reference procedure for Lp(a) measurement (23)(28). In contrast, the latex method tends to underestimate Lp(a) in individuals with apo(a) of a smaller size than the apo(a) present in the assay calibrator and overestimate Lp(a) in individuals with larger apo(a) particles (23)(28). Unfortunately, a recent survey demonstrated that almost all commercially available methods for Lp(a) were affected by the kringle 4 type 2 variably repeated epitopes, reaffirming the need for developing accurate Lp(a) methods for the clinical laboratory (23)(28). In the present study, we found that the commercial Lp(a) assay indeed underestimated the predicting risk power of Lp(a). It seems likely that Lp(a) methods affected by apo(a) size that were previously used in the Physicians’ Health Study (19)(20)(21) may have underestimated or even obscured the true relationship between Lp(a) concentration and CHD.
Because Lp(a) concentrations vary greatly among ethnic groups and this variability cannot always be explained by the differences in apo(a) size (27), the association between Lp(a) concentrations and risk of future coronary events could be different by racial or ethnic groups. Although the conclusion of this study may be limited to Caucasians, the association of Lp(a) concentration and apo(a) related to the risk of arteriosclerosis remains valid.
Plasma samples had prolonged storage before assay. However, the frequency distributions and the median concentrations seen here were almost identical to those reported for samples stored for shorter duration or analyzed immediately after collection (27). Furthermore, because samples from cases and controls were stored and handled similarly, it is unlikely that any systemic bias as a result of storage have affected the findings. Finally, any specimen deterioration would have been detected during the electrophoretic separation used in assessing apo(a) size, but no breakdown or aggregation of apo(a) was seen on immunoblotting.
In conclusion, our findings indicate that Lp(a) concentration and apo(a) isoform size predict angina. People with high Lp(a) concentrations, especially with high LDL-C concentrations, are at 4- to 12-fold higher risk than those with lower concentrations of both, reflecting a synergistic relationship. However, good clinical application of these findings will need the availability of standardized commercial assays validated to accurately measure Lp(a) concentrations independent of apo(a) size. Although apo(a) size was superior to Lp(a) concentration as a predictor of angina, this finding needs to be confirmed by additional larger studies.
This work was supported by research grants (HL-26490, HL-34595, CA-34944, CA-40360, CA-58684, CA-42182, HL-30086, HV-88175, and CA-78293) from the NIH (Bethesda, MD).
1 The follow-up period was 5 years.
2 MC, matching criteria; MI, myocardial infarction.
3 In individuals heterozygous for apo(a) isoform size, when one of the two isoforms was predominant, that isoform was used for statistical analysis. When the two isoforms appeared to be present in the same proportion, the mean kringle 4 number of the two isoforms was used.
1 Matched for age and smoking.
2 Matched for age and smoking and controlled for aspirin, CHD risk factors, and lipids.
3 Matched for age and smoking and controlled for aspirin, CHD risk factors, lipids, and apo(a) size.
1 Data were available for 187 cases and 187 controls.
2 Matched for age and smoking.
3 Matched for age and smoking and controlled for aspirin, CHD risk factors, and lipids.
4 Matched for age and smoking and controlled for aspirin, CHD risk factors, lipids, and apo(a) size.
1 Matched for age and smoking.
2 Matched for age and smoking and controlled for aspirin, CHD risk factors, and lipids.
3 Matched for age and smoking and controlled for aspirin, CHD risk factors, lipids, and Lp(a) concentration.
↵1 Nonstandard abbreviations: Lp(a), lipoprotein(a); apo(a), apolipoprotein(a); CHD, coronary heart disease; LDL-C and HDL-C, LDL- and HDL-cholesterol, respectively; RR, relative risk; and CI, confidence interval.
- © 2004 The American Association for Clinical Chemistry