Jump to
Abstract
Abstract An interaction between high plasma lipoprotein(a) [Lp(a)], unfavorable plasma lipids, and other risk factors may lead to very high risk for premature CAD. Plasma Lp(a), lipids, and other coronary risk factors were examined in 170 cases with early familial CAD and 165 control subjects to test this hypothesis. In univariate analysis, relative odds for CAD were 2.95 (P<.001) for plasma Lp(a) above 40 mg/dL. Nearly all the risk associated with elevated Lp(a) was found to be restricted to persons with historically elevated plasma total cholesterol (6.72 mmol/L [260 mg/dL] or higher) or with a total/HDL cholesterol ratio >5.8. Nonlipid risk factors were also found to at least multiply the risk associated with Lp(a). When Lp(a) was over 40 mg/dL and plasma total/HDL cholesterol >5.8, relative odds for CAD were 25 (P=.0001) in multiple logistic regression. If two or more nonlipid risk factors were also present (including hypertension, diabetes, cigarette smoking, high total homocysteine, or low serum bilirubin), relative odds were 122 (P<1×10−12). The ability of nonlipid risk factors to increase risk associated with Lp(a) was dependent on at least a mildly elevated total/HDL cholesterol ratio. In conclusion, high Lp(a) was found to greatly increase risk only if the total/HDL cholesterol ratio was at least mildly elevated, an effect exaggerated by other risk factors. Aggressive lipid lowering in those with elevated Lp(a) therefore appears indicated.
-
Reprint requests to Dr Paul N. Hopkins, Cardiovascular Genetics Research Clinic, 410 Chipeta Way, Room 161, Salt Lake City, UT 84108.
- Received December 16, 1996.
- Accepted May 8, 1997.
Lipoprotein(a) [Lp(a)] refers to particles formed by the covalent binding of the glycoprotein apoprotein(a) to apoprotein B of LDL by disulfide linkage. Increased plasma concentration of Lp(a) has been associated with increased risk for premature coronary disease in numerous retrospective case-control studies among white subjects1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 and in most prospective studies,47 48 49 50 51 52 53 54 55 56 57 58 59 with three notable exceptions.56 60 61 High Lp(a) is apparently not a risk factor among blacks40 62 or, in some studies, among the elderly.10 11 46 63 High Lp(a) has been found to be predictive of more rapid progression of coronary artery disease (CAD) in serial angiograms.64 65 66 It was also associated with more rapid occlusion of coronary bypass grafts in one study67 but not in another.68 A higher frequency of restenosis after angioplasty in those with high Lp(a) has been noted in most69 70 71 72 73 74 75 but not all studies,35 and endothelial dysfunction has been associated with high Lp(a).76 Risk for stroke12 37 77 or carotid atherosclerosis78 79 and peripheral vascular disease80 81 82 is increased in those with high Lp(a). In children, high Lp(a) is found more frequently in those with a positive family history of coronary disease, at least among whites.83 84 85 86 87
In most of these studies, simple univariate relative odds for CAD have been modest, usually 2 to 4, when Lp(a) is above 30 to 40 mg/dL. In contrast, the risk associated with high Lp(a) has been reported to be much higher in persons with familial hypercholesterolemia.16 20 Yet, an elevated Lp(a) imparted little excess risk among participants of the Physicians Health Study, a group with relatively low serum cholesterol concentrations.61 Furthermore, Maher et al,88 found that in persons with a 10% or greater reduction in LDL cholesterol during a vigorous lipid-lowering intervention, Lp(a) was not predictive of atherosclerosis progression or CAD events (only 9% incidence of new events in those with high Lp(a) and at least 10% LDL reduction), while in those with little change in LDL, Lp(a) remained a strong predictor of progression and events (with a 39% incidence in those with high Lp(a) and <10% LDL reduction). Lp(a) concentration was unaffected by this intervention, consistent with other studies using most cholesterol-lowering agents and observations that plasma Lp(a) levels are normally determined almost entirely by genetic factors at the apo(a) locus.89 90 These findings led to the hypothesis that Lp(a) is primarily a risk factor when serum cholesterol or LDL cholesterol is elevated. Results from several observational and intervention studies were consistent with this hypothesis when, in meta-analysis, relative odds reported in each study were plotted against average study LDL cholesterol levels.91 Nevertheless, few studies have examined this issue directly,9 27 78 and none have examined potential interactions between Lp(a) and nonlipid risk factors.
We have measured plasma Lp(a) concentration, lipids, and other cardiovascular risk factors in patients with early familial coronary disease and control subjects. Nearly all the excess risk associated with Lp(a) was found among those with currently or historically elevated plasma total cholesterol concentration or in those with an increased ratio of total/HDL cholesterol, supporting the hypothesis that plasma lipid levels strongly influence risk from high Lp(a). In addition, several other nonlipid risk factors were found to markedly increase the risk associated with high Lp(a), again dependent on the presence of a moderately high total/HDL cholesterol ratio.
Methods
Study Participants
Early familial CAD cases included 124 men and 46 women who had survived a myocardial infarction, percutaneous transluminal angioplasty, or coronary artery bypass grafting by age 55 for men or age 65 for women. Patients were seen at least 6 months after their acute event. Each of these cases had a sibling confirmed to have early CAD by the same definition. Only unrelated CAD cases were used in the present study. These were selected as the sibling with the earliest onset of CAD in families with multiple screened siblings.
Control subjects included 85 men and 80 women who were ascertained either from a random population sampling (35% of control subjects) or who were spouses of hypertensive siblings who had participated in previous studies in our clinic (65% of control subjects). The former population was selected randomly by computer from a database of families collected as part of the Family Health Tree program, an ongoing collaboration between Cardiovascular Genetics, the Utah Department of Health, and area high schools.92 93 High school students, with the help of their parents, collect health information on family members. The data are computerized and scored for familial disease tendency. Control subjects were selected randomly from the parent generation. Other samples identified from Family Health Trees have been further studied and confirmed to be representative of the Utah population.94 Hypertensive probands whose spouses were used as control subjects were selected for prior studies because they and one or more of their siblings had hypertension onset before age 60.95 All screened spouses were used as control subjects for this study. Comparability of the two groups of control subjects was tested. There were no significant differences for any cardiovascular risk factor between the two control groups, except for a slightly higher plasma total cholesterol among the spouses of hypertensive siblings (P=.05). Since a slightly higher serum cholesterol level could only provide a more conservative comparison with the early familial coronary cases, the two control groups were combined for all further analyses.
In this report, all available cases and control subjects were used. In prior studies of this group, we had constrained the age range to 38 to 68 years so that cases and control subjects might be more completely overlapping in age (these studies also provide justification for including homocyst(e)ine [H(e)] and bilirubin as risk factors in the present study).96 97 However, since Lp(a) is only slightly increased after menopause in women and is unaffected by age in men,98 we deemed such restriction unnecessary. Nevertheless, we have confirmed all findings reported here using the constrained age range as well. This study was approved by the Institutional Review Board of the University of Utah Medical Center. All subjects signed informed consent before participating.
A participant was considered to have hypertension if taking antihypertensive medication with a prior physician diagnosis of hypertension or if the mean of two supine diastolic blood pressures taken with a Critikon Dynamap automated blood pressure machine was greater than 95 mm Hg. Diabetes was considered present if a prior physician diagnosis had been made or if the fasting serum glucose on screening was >7.77 mmol/L (140 mg/dL). Cigarette smoking was dichotomized into “ever” or “never,” with ever smoking defined as having smoked daily for 1 year or more. Many patients had quit after onset of their CAD; hence the designation as ever smoking rather than current and former.
Laboratory Methods
Blood samples were collected in the morning after 12 to 16 hours of fasting and prepared according to guidelines of the Lipid Research Clinic’s Program Manual of Laboratory Operations.99 Lipids were measured by a microscale procedure developed in our laboratory.100 Briefly, HDL was measured in the supernatant after precipitation of apolipoprotein B–containing particles with dextran sulfate-MgCl2 and centrifugation in an Eppendorf microcentrifuge. Triglyceride-rich lipoproteins, primarily VLDL, were separated from LDL and HDL by use of a Beckman TL100 ultracentrifuge. The value for VLDL cholesterol was taken as the measured cholesterol in the top fraction. This value was compared with the total cholesterol minus the cholesterol in the bottom fraction containing LDL+HDL and verified to yield virtually identical results. Cholesterol and triglycerides in total plasma and subfractions were measured with a Roche FARA II automated analyzer. Our lipid laboratory participates in the standardization program of the Centers for Disease Control in Atlanta, Georgia.
Lp(a) was measured with a commercial enzyme-linked immunosorbent assay that uses a monoclonal antibody linked to a solid phase to trap Lp(a) particles and a polyclonal anti-Lp(a) antibody conjugated with horseradish peroxidase as the reporter antibody (Macra Lp(a) kit, Strategic Diagnostics Inc).101 102 Units refer to whole particle mass per unit volume. The polyclonal anti-Lp(a) used in this assay yields Lp(a) concentrations somewhat higher than an assay using an anti-apoB reporter antibody, the discrepancy being most apparent at Lp(a) concentrations under 10 mg/dL and minimal at values above 10 mg/dL.102 This is in contrast to the underestimation bias at higher Lp(a) concentrations observed for a monoclonal anti-Lp(a) antibody that recognized an epitope on the kringle 4 type 2 repeating subunit.103 In our laboratory, the interassay variation (coefficient of variation) was 6% for samples having over 40 mg/dL Lp(a) and 11% for samples with 18 mg/dL Lp(a). These values are comparable to those reported in the kit documentation. Precision decreased further at lower concentrations with a coefficient of variation of 21% at 2.0 mg/dL Lp(a).
Samples were stored at −70°C using small-volume storage vials that were thawed only once at the time of assay to avoid the differential loss of Lp(a) antigenicity seen at lower storage temperatures.75 104 105 Indeed, an overly long storage time at −20°C has been suggested as an explanation for the lack of association between Lp(a) and coronary disease in the Helsinki Heart Study60 and another Finnish study.56
Total H(e) was measured after reduction of disulfide bonds and detection of released homocysteine by high-pressure liquid chromatography as described by Malinow et al106 with minor modifications.107
Statistical Analysis
The SAS Statistical Software Package was used for data analysis (SAS Institute, Inc, Cary, NC). Statistical analyses on triglycerides were done after logarithmic transformation. Statistical tests included Student’s t test, χ2, Fisher’s exact test, Pearson’s correlation, and stepwise multiple logistic regression. The Wilcoxon rank sum test was used to compare Lp(a) values in cases and control subjects. Calculation of pooled relative odds after stratification by gender was performed by the Mantel-Haenszel method as described by Rothman.108
Results
Clinical characteristics of unrelated early familial coronary cases and control subjects are shown in Table 1⇓. The male cases were, on average, 5 years older, while age difference in women was about 15 years. In addition, the prevalence rates of hypertension and diabetes were markedly higher in both male and female cases than control subjects. Case women had significantly higher body mass index (BMI, as kg/m2), blood pressures, plasma total cholesterol, measured LDL cholesterol, VLDL cholesterol, triglycerides, and H(e) levels than control women. Plasma HDL cholesterol and bilirubin were significantly lower. Among men, similar relationships were seen, except that BMI, blood pressures, and total and measured LDL cholesterol were not significantly higher in cases than control subjects. Measured plasma LDL cholesterol levels were not different in male cases compared with control subjects, probably because of treatment. About half of the cases were being treated with diet and/or medication for high cholesterol (with about one third of all cases taking lipid-lowering medication), while fewer than 8% of the control subjects were on a diet and none were taking lipid-lowering medication. Maximum total cholesterol levels (see below) were significantly higher in cases of both genders.
Clinical Characteristics of Cases and Control Subjects by Gender
To estimate the effect of increasing Lp(a) on risk, cases and control subjects were categorized by plasma Lp(a) quintile (based on the control population), and relative odds were calculated. Control population percentile levels are given in Table 2⇓. The Mantel-Haenszel test for trend was significant in men (P=.045) but failed reach significance in women (P=.3). Pooled estimates of relative odds after stratification by gender are given in Table 3⇓. Clearly, most of the excess risk associated with Lp(a) was observed in the highest quintile (ie, above the 80th percentile). Relative odds for Lp(a) above 40 mg/dL (roughly the 90th percentile in control subjects) was 2.95 (P=.001) in univariate analysis.
Lipoprotein(a) Percentiles in Control Subjects
Univariate Relative Odds by Lipoprotein(a) Quintile
Risks associated with Lp(a) at different lipid levels were then examined. Two-way classification by screened plasma total cholesterol and Lp(a) is shown in Table 4⇓. In this analysis, only a slight trend for greater effects of Lp(a) among those with higher measured total cholesterol was observed. Classification by measured LDL cholesterol at the time of screening showed almost identical trends (data not shown). Classification by total/HDL cholesterol showed stronger evidence for interaction. However, the 13.4-fold increase in risk among those with plasma Lp(a) >40 mg/dL and total/HDL cholesterol >5.8 would also be consistent with a multiplicative effect between these two variables.
Relative Odds for Combinations of Plasma Lipoprotein(a) Concentration and Plasma Total Cholesterol at Screening
Many of the cases were on lipid-lowering therapy at the time of screening. Most patients treated for lipids began lipid treatment after their coronary events. Historical lipid levels would therefore be expected to be more closely associated with risk than levels measured at the time of screening. To examine this hypothesis, historical maximum cholesterol levels were estimated. The historical maximum cholesterol was taken as the maximum of either the reported prior highest plasma cholesterol level (from questionnaires filled out by participants), the measured total cholesterol level at the time of screening, or the measured total cholesterol level multiplied by 1.2 for those currently taking cholesterol-lowering medication. The 20% adjustment was considered conservative given that many of those treated were taking statin medications and were also following cholesterol-lowering diets. Among the 170 cases, 96 provided a prior highest cholesterol value on their questionnaires (48 of 56 on lipid-lowering medication did so). The actual maximum cholesterol used was the currently measured cholesterol for 79 cases, the historically high cholesterol for 67 cases, and the adjusted high cholesterol for 24 cases. Only 11 of 165 control subjects provided a prior highest cholesterol value on their questionnaires and in five instances it was higher than the measured cholesterol level. The total cholesterol measured at screening among the 48 patients taking lipid-lowering medication who also reported a historical high cholesterol was 5.81±1.06 mmol/L (225±41 mg/dL) (mean±SD) compared with a reported prior maximum total cholesterol of 7.78±1.45 mmol/L (301±56 mg/dL). An adjustment of 20% therefore appeared to be reasonable and conservative.
Risks associated with the various combinations of plasma Lp(a) and maximum total plasma cholesterol were then calculated as shown in Table 6⇓. If maximum total cholesterol level was under 5.68 (220 mmol/L), an Lp(a) level of 40 mg/dL or greater was associated with only a small, nonsignificant increase in risk. However, a marked increase in risk (relative odds=13.8, P=.0000001) was seen among those with both high cholesterol (above 6.72 mmol/L [260 mg/dL]), and high Lp(a). In persons with maximum total cholesterol levels between these values, Lp(a) effects were intermediate.
Relative Odds for Combinations of Plasma Lipoprotein(a) Concentration and Plasma Total/HDL Cholesterol Ratio at Screening
Relative Odds for Combinations of Plasma Lipoprotein(a) Concentration and Maximum Total Cholesterol
We then calculated the ratio of plasma maximum total/HDL cholesterol. The HDL levels at the time of screening were used in this analysis. Results are shown in Table 7⇓. A strong interactive effect is apparent with a pooled relative odds estimate of 35 in those with maximum total/HDL cholesterol ratio above 5.8 and Lp(a) more than 40 mg/dL compared with those with a maximum total/HDL cholesterol ratio under 4.5 and Lp(a) under 15 mg/dL. These results suggest more significant interaction associated with high Lp(a) when combined with the more predictive maximum total/HDL cholesterol ratio.
Relative Odds for Combinations of Plasma Lipoprotein(a) Concentration and Maximum Total/HDL Cholesterol Ratio
Multiple stepwise logistic regression was then performed to examine the effects of other risk factors on the relationship between plasma Lp(a) and maximum total/HDL cholesterol ratio. Three dummy variables (each with possible values 0 or 1) were defined for the various combinations of high Lp(a) and high maximum total cholesterol/HDL ratio (both high, only maximum total/HDL cholesterol high, only Lp(a) high). Other risk factors included in the model were age, gender, BMI, cigarette smoking, hypertension, diabetes, and triglycerides (natural log transformed). Results are given in Table 8⇓. As in the stratified analysis above, multiple stepwise logistic regression suggested a much stronger risk associated with high Lp(a) if maximum total/HDL cholesterol ratio was also high, with relative odds of 25.1 (P=.0001) compared with a nonsignificant risk for high Lp(a) if maximum plasma total cholesterol was <5.8. Similar results were found when plasma bilirubin and total H(e) were entered into the regression, when the analysis was restricted to persons 38 to 68 years of age, or when only reported maximum plasma total cholesterol levels were used without adjustment for drug treatment (results not shown).
Results of Multiple Logistic Regression With Early CAD (Case/Control) as the Dependent Variable
When continuous levels of maximum total/HDL cholesterol ratio and Lp(a) were examined in multiple logistic regression without an interaction term, Lp(a) entered sixth (behind age, cigarette smoking, maximum total/HDL cholesterol ratio, diabetes, and hypertension), with relative odds of 2.1 associated with a 40 mg/dL increase (P=.015). However, when an interaction term between continuous levels of Lp(a) and maximum total/HDL cholesterol was entered into the regression, the interaction term and maximum total/HDL cholesterol ratio entered, while Lp(a) alone did not, suggesting that most if not all the risk attributable to Lp(a) occurred in conjunction with an elevated maximum total/HDL cholesterol ratio (results not shown).
We then examined potential interaction between high Lp(a) and other, nonlipid risk factors as shown in Table 9⇓. The effect of Lp(a) >40 mg/dL versus <40 mg/dL was examined in the presence or absence of hypertension, cigarette smoking, elevated homocysteine (above the sex-specific 90th percentile for control subjects), and low bilirubin (below the sex-specific 40th percentile for control subjects). Diabetes was examined together with hypertension, since too few control subjects were affected to calculate reliable odds estimates. Marked elevations in risk were seen when high Lp(a) was combined with hypertension, high H(e), or low bilirubin, with observed relative odds for these combinations greater than the product of each risk factor alone. The risk associated with high Lp(a) and cigarette smoking appeared approximately multiplicative. A multiple logistic model was then tested, which included age and gender, together with three dummy variables representing combinations of Lp(a) and other risk factors (Lp(a) >40 mg/dL with zero to two other risk factors, Lp(a)=40 mg/dL with three or more other risk factors, and Lp(a) >40 mg/dL with three or more other risk factors). Other risk factors included maximum total/HDL cholesterol ratio >5.8, together with factors listed in Table 9⇓. The combination of high Lp(a) and three or more other risk factors was associated with relative odds of 43 (P<.0001), while for three or more risk factors without high Lp(a), the relative odds were 13 (P<.0001). High Lp(a) with fewer than three risk factors did not enter into the regression.
Risks Associated With High Lipoprotein(a) in the Presence and Absence of Nonlipid Risk Factors
To determine whether the results with multiple risk factors were dependent on lipid levels, the risk associated with high Lp(a) was calculated for exposure to different numbers of nonlipid risk factors in the presence or absence of a maximum total/HDL cholesterol ratio >5.8. Results are shown in Table 10⇓. In this analysis, the increased risk imparted by high Lp(a) in the presence of nonlipid risk factors was clearly dependent on an elevation of the maximum total/HDL cholesterol ratio. Exceedingly high risks were evident when Lp(a) was >40 mg/dL, maximum total/HDL cholesterol was >5.8, and two or more nonlipid risk factors were present (odds ratio=122, P<1.0×10−12 compared with those having lower levels of all these factors).
Relative Odds for Combinations of Plasma Lipoprotein(a) Concentration, Increased Total/HDL Cholesterol (>5.8), and Number of Other (Nonlipid) Risk Factors
Discussion
We have observed very high risks (relative odds of 13.8, P=1.0×10−7) associated with a combination of high plasma Lp(a) (>40 mg/dL) and elevated historical maximum total cholesterol concentrations. Even higher risks (relative odds of 35, P<1.0×10−7) were associated with the combination of high Lp(a) and an increased ratio of maximum total/HDL cholesterol (>5.8). These risks were further multiplied by other, nonlipid risk factors, but primarily in the presence of an elevated maximum total/HDL cholesterol ratio. In contrast, a high Lp(a) level without historically high plasma total cholesterol or an elevated maximum total/HDL cholesterol ratio was associated with only a small, nonsignificant increase in risk. The risk associated with a high Lp(a) and high plasma maximum total/HDL cholesterol ratio were more than multiplicative, as shown by a significant interaction in multiple logistic regression analysis. These results corroborate and extend more limited findings from other studies.9 27 88 91 Moreover, this is the first investigation to directly examine potential interactions between Lp(a) and nonlipid risk factors.
Limitations of our study should be considered. In some subsets, because of limited numbers, we were unable to make a strong case for true interaction between Lp(a) and lipids versus a simple multiplicative effect. Also, findings from our study are limited to persons with early familial CAD. Extension of these findings to persons with sporadic early coronary disease and confirmation in larger studies would be of great interest. In any retrospective case-control study, inference to a healthy population requires the assumption that levels of the risk factor (in this case Lp(a) levels) are not affected by the disease in question. Recent studies do indicate that Lp(a) concentration may be increased by a number of inflammatory conditions and that Lp(a) concentration shows low-level correlations with other acute-phase reactants.109 Acute-phase reactants together with Lp(a) are known to rise in the first few days to weeks after a myocardial infarction, but these acute elevations resolve after 1 month.110 111 Because our cases were sampled more than 6 months after any acute event, it seems unlikely that the significant differences in our study (and in most other case-control studies) can be attributed to changes in Lp(a) induced by coronary atherosclerosis. However, to our knowledge, no long-term determinations of prospective change in Lp(a) have been made among those who eventually do or do not develop CAD. Finally, we did not determine Lp(a) genotype and are unable to address the question of genotype-specific risk raised by some authors.112 Nevertheless, the preponderance of evidence strongly suggests increased risk associated with quantitative elevation of plasma Lp(a) concentration. Differences in risk associated with qualitative differences remain more speculative.
Several mechanisms whereby Lp(a) may promote atherosclerosis have been proposed. These include inhibition of clot lysis by Lp(a), leading to a thrombogenic state;113 114 115 increased binding to proteoglycans, thereby promoting increased uptake by macrophages;116 and promotion of smooth muscle cell proliferation by blocking the plasmin-dependent activation of transforming growth factor-β.117 118 119 Though Lp(a) is, if anything, less subject to oxidation than native LDL,120 oxidized Lp(a) may contribute directly to accumulation of lipid in macrophages.121 122
Our findings provide additional insight into how high Lp(a) may increase coronary disease risk. If Lp(a) itself contributed substantially to the lipid pool in foam cells, one might expect a graded risk that was additive with other lipid risk factors. The striking absence of risk associated with high Lp(a) when total/HDL cholesterol was low speaks against this hypothesis. Alternatively, aggravation of late sequelae such as thrombosis superimposed on preexistent, complicated plaques would be consistent with the dependency of Lp(a)-associated risk on other risk factors. The striking interaction between H(e) and Lp(a) found here is of particular interest in this regard, since increased H(e) is also associated with greater thrombotic risk. This interpretation may also help explain the association of Lp(a) with late sequelae of atherosclerosis in other clinical studies. Examples include more rapid progression of CAD in serial angiograms64 65 66 ; higher Lp(a) levels in myocardial infarction survivors who had infarct-related arteries that remained occluded compared with those with recanalization123 ; and increased 4-year mortality in persons with angiograms showing significant carotid atherosclerosis who also had high Lp(a).124
Our findings may have significant public health implications, especially if confirmed by other studies. High Lp(a) may serve as an important risk factor to identify individuals who would especially benefit from aggressive lipid lowering.
Acknowledgments
This study was supported in part by grants HL47651, HL21088, and HL47466 from the National Heart, Lung, and Blood Institute, Bethesda, Md.
References
This Issue
Jump to
Article Tools
Share this Article
- Lipoprotein(a) Interactions With Lipid and Nonlipid Risk Factors in Early Familial Coronary Artery Disease
