A Prospective Case-Control Study of Lipoprotein(a) Levels and Apo(a) Size and Risk of Coronary Heart Disease in Stanford Five-City Project Participants
Lipoprotein(a) [Lp(a)] is formed by the assembly of LDL particles and a carbohydrate-rich protein, apolipoprotein(a) [apo(a)], which has a high degree of structural homology with plasminogen. While the majority of retrospective studies have found an association between Lp(a) level and cardiovascular disease (CVD), the few prospective studies to date have reported contradictory results. We conducted a nested case-control study using the participants in the Stanford Five-City Project, a long-term CVD prevention trial. Participants with an incident possible or definite myocardial infarction or coronary death were matched to a single control subject for age, sex, ethnicity, residence in a treatment or control city, and time of survey. This process yielded 134 case-control pairs, 90 male and 44 female, for whom plasma was available for analysis of Lp(a). Lp(a) values in nanomoles per liter were determined by an enzyme-linked immunoassay that measures Lp(a) independently of apo(a) size polymorphism. Apo(a) size isoforms were determined by SDS–agarose gel electrophoresis. Median Lp(a) level in male cases was almost double that in control subjects (41.8 versus 21.2 nmol/L; P<.01); in female cases, median Lp(a) was 34% higher than in control subjects (32.5 versus 21.2 nmol/L), but this difference was not statistically significant. Among the male cases, there was an increased frequency of small apo(a) isoforms, while no significant difference was found in apo(a) size between female cases and control subjects. The association between Lp(a) level and case-control status in men was independent of total, HDL, and non-HDL cholesterol levels, as well as apo(a) size isoform, cigarette smoking, blood pressure, and obesity. In men, the most efficient threshold value of Lp(a) concentration for separating cases and control subjects was 35 nmol/L; the odds ratio for being a case above this level compared with below was 2.84 (95% confidence interval: 1.53-5.27, P<.001). This study provides strong evidence that Lp(a) level is a prospective, independent risk factor for developing coronary artery disease in men and indicates that the size of apo(a) may also play a role. The lack of a significant association in women deserves further evaluation in larger studies.
- Received May 5, 1996.
- Revision received September 20, 1996.
Lipoprotein(a) is a peculiar class of lipoprotein particles formed by the assembly of LDL particles and a highly carbohydrate-rich protein, apo(a), linked by a disulfide bond to the apo B-100 component of LDL.1 Apo(a) is highly variable in size owing to length polymorphism in the apo(a) gene2 and has a high degree of structural homology with plasminogen, a key protein of the coagulation process.3 Apo(a) is formed by a protease domain, one copy of kringle 5 domain (both exhibiting over 85% homology with the corresponding regions of plasminogen), and multiple copies of the plasminogen-like K4 domain.1 On the basis of amino acid differences, apo(a) K4 can be divided into 10 distinct K4 types, 9 of which are present in a single copy, while K4 type 2 varies in number from 3 to over 40 repeats.2 The varying number of K4 type 2 is the major determinant of apo(a) size polymorphism and gives origin to the large number of apo(a) isoforms detected in human plasma.4 Plasma concentrations of Lp(a) vary widely among individuals and are highly heritable, with ≈90% of the variability attributable to sequences at or closely linked to the apo(a) locus.5 Additionally, there is an inverse relationship between the number of K4 type 2 repeats in the apo(a) gene and plasma Lp(a) concentrations.6
The majority of the numerous retrospective case-control studies have found an association between high Lp(a) values and CAD (for review see Reference 7). However, retrospective studies cannot establish that the Lp(a) level was high before the development of CHD and are subject to numerous biases. The few prospective studies so far reported have provided contradictory results, with four of them finding an association between CHD and high Lp(a) values8 9 10 11 and three not.12 13 14 While it is difficult to find complete explanations for such contradictory results, problems related to the length and temperature of storage of the samples and limitations of the immunoassays used to measure Lp(a) should be taken into consideration. We have recently documented that due to the variable number of apo(a) K4 type 2 repeats, Lp(a) values are underestimated or overestimated, depending on the apo(a) size in the samples, when the analytical methods are based on antibodies recognizing this variable part of the molecule.15 We have also demonstrated that Lp(a) can be accurately measured if a monoclonal antibody that does not recognize K4 type 2 is used in the assay.15
In none of the reported prospective studies were apo(a) size isoforms evaluated to determine whether apo(a) polymorphism, in addition to Lp(a) values, may play a role in CHD. Additionally, no confirmation of lack of symptoms suggestive of preclinical CHD in control groups was reported. With the availability of state-of-the-art analytical methods, the Stanford Five-City Project provided an opportunity to rigorously evaluate the contribution of Lp(a) concentration and apo(a) size to CHD risk in a prospective case-control study of a general population of men and women in which control subjects were confirmed to be free of a clinical history and symptoms suggestive of prevalent clinical CHD and CVD.
Design and Subjects
The Stanford Five-City Project, which began in 1978, is a long-term field trial evaluating the effects of a community-wide health education program on CVD risk factors, morbidity, and mortality. To assess risk-factor change, representative cross-sectional population surveys were conducted every 2 years from 1979 to 1986 and again in 1989 to 1990. All residents of randomly selected households who were aged 12 to 74 years were eligible for the surveys; response rates ranged from 56% to 69%. Further details on the study design, field methodology, and results have been reported previously.16 17
To monitor morbidity and mortality, a comprehensive community surveillance system was established to identify and validate all hospitalized fatal and nonfatal myocardial infarctions and strokes, as well as fatal coronary disease and stroke not resulting in hospitalization, in the 1979 through 1992 period.18 Cases were validated by nurse and physician review of clinical records, using a diagnostic algorithm including chest pain, electrocardiograms, cardiac enzymes, and autopsy findings. Individual identification information is maintained confidentially in the surveillance database to allow linkage of different events within persons. The present study uses a nested case-control design by combining the survey and surveillance databases. The series of cross-sectional surveys was not designed to be a cohort study of end points, but we were able to identify cases of acute myocardial infarction among the total survey population, ages 25 to 74 (n=9140), by matching the list of names and birth dates with the surveillance database. Potential matches were confirmed by examination of addresses and, when available, social security numbers. Since the surveillance database ended at age 74, we also matched all survey participants to California State death certificate records for all ages, looking for deaths listed with CVD (ICD-9 codes 410 through 414) as the underlying cause.
Through this process, 215 cases of acute fatal or nonfatal myocardial infarction or fatal CAD were identified (198 from surveillance and 17 from the death records) among the participants of the surveyed population who had available plasma samples in storage. For this investigation of Lp(a), only incident cases were included. To eliminate prevalent CAD, we excluded cases who were either surveyed after the date of their first event in the surveillance database or whose medical records from additional surveillance data indicated a history of coronary disease or stroke before the survey date. For events without a clear medical history in this database and for the fatal events not appearing in the surveillance database (those from state death certificates), we used a question on the survey about prior history of a heart attack. This process left 119 incident cases with available plasma samples.
A single control subject was matched to each case for age (within 5 years), sex, ethnicity, residence in a treatment or control city, and time of survey. Potential control subjects were contacted by telephone by experienced survey staff who administered a questionnaire designed to identify persons suffering from both diagnosed and undiagnosed heart disease and CVD. Potential control subjects who gave a history of angina-like chest pain, positive treadmill test or angiography, myocardial infarction, stroke, transient ischemic attack, or other CVD were excluded from the study. During this process, individuals with a history of angioplasty or coronary artery bypass surgery were identified and reclassified as probable cases. For 15 such persons, matching control subjects were identified and contacted and the pair was included. The final study population thus included 134 case-control pairs, 90 male and 44 female. The ethnic composition of the 134 case-control pairs was as follows: 119 white (81 male, 38 female), 8 Hispanic (4 male, 4 female), 1 black (male), and 6 of other race (2 female, 4 male).
Sixty-one cases met the criteria for definite myocardial infarction from autopsy or from the presence of at least two of the following criteria: chest pain of at least 20 minutes' duration, two cardiac enzyme measurements exceeding twice the upper limit of normal, and evolving Q waves on serial electrocardiograms. Probable CAD or myocardial infarction occurred in 73 cases and was defined as either ICD-9 codes 410 through 414, the presence of one of the above criteria, or identification through the phone screening described above.
For all participants at each survey, venous blood samples were drawn into vacuum tubes containing EDTA. Prior fasting was not required. Blood was drawn in the sitting position, and other recommendations for blood collection specified in Lipid Research Clinics Program Manual of Laboratory Operations, Lipid and Lipoprotein Analysis19 were followed. Plasma was separated from cells within 30 minutes and stored at 4°C to 7°C before and during shipment to Stanford, which occurred twice a week. An aliquot of each sample was then stored at −70°C until Lp(a) concentrations and apo(a) size isoforms were measured in 1994 and 1995. Storage times did not differ significantly between cases and control subjects; P=.95 for men and P=.89 for women.
Determination of Lp(a) Values
On a weekly basis, 30 samples were thawed and filtered through a 20-μm filter. An aliquot of each sample, blinded for case or control status, was sent in wet ice by overnight express to the University of Washington, where analyses were immediately performed. A small pilot study, consisting of three samples collected at different time points (5, 10, and 15 years before this study) from 20 participants in the cohort study, was conducted before the main study. Both sexes and a variety of age groups were included. Lp(a) analyses were performed to evaluate whether our method for measuring Lp(a) was affected by the length of storage of the samples. Lp(a) values were determined by an enzyme-linked immunosorbent assay performed as previously described.15 The detection monoclonal antibody in this assay (MAb a-40) is specifically directed to an epitope expressed in apo(a) K4 type 9, and this method has been demonstrated to accurately measure Lp(a) independently of apo(a) size polymorphism. The Lp(a) values are expressed in nanomoles per liter of Lp(a) protein.
Determination of Apo(a) Isoforms
The apo(a) size isoforms were determined by a high-resolution SDS–agarose gel electrophoresis followed by immunoblotting as previously reported.4 20 We have recently determined the relationship between the number of K4 domains determined by pulsed-field gel electrophoresis21 and the mobility of the isoforms in SDS–agarose gel electrophoresis4 and found that the log of the K4 number was highly correlated with the mobility of the isoforms in agarose gel.22 Therefore, the apo(a) size isoforms are designated by the relative number of K4 domains.
Fresh plasma was analyzed for lipid and lipoprotein levels within 7 days of blood drawing. From 1979 to 1986, total and HDL cholesterol concentrations were determined with the Lipid Research Clinics' methodology on the Technicon AutoAnalyzer II (Technicon Instruments Corporation) and the heparin-manganese precipitation procedure for HDL separation.19 For the final survey in 1989 to 1990, total and HDL cholesterol concentrations were determined enzymatically by using an Abbott ABA 200 (Abbott Laboratories). HDL separation was performed by a dextran sulfate/Mg2+ precipitation method.23 Throughout the study, the laboratory remained in a standardized status within the Centers for Disease Control/National Heart, Lung, and Blood Institute Standardization Program.
Trained health professionals interviewed participants regarding their demographic characteristics, health knowledge, cardiovascular risk–related attitudes and behavior, and medical history. Participants brought prescription drug vials to the survey centers, and dose and frequency were recorded from the labels by the survey team. Smokers were defined as those individuals who reported ever smoking cigarettes on a regular basis and who had smoked even one cigarette in the past 48 hours. For this study, current and ex-smokers were combined to give a group of ever-smokers.
Blood pressure was measured three times after the participant had been seated quietly for 2 minutes, using a semiautomated device, the Sphygmetrics SR-2 automatic blood pressure recorder (Sphygmetrics, Inc).24 The average of the second and third readings was used for analysis.
For this study the null hypothesis was that baseline Lp(a) levels and apo(a) polymorphs would not differ between cases and control subjects. Two secondary alternative hypotheses were also tested: that Lp(a) level would better discriminate cases from control subjects at high levels of non-HDL cholesterol and at low levels of HDL cholesterol.
The statistical significance of differences in CHD risk factors by case-control status was investigated using paired t tests for continuous variables and McNemar's test for categorical variables. Because of their positively skewed distribution, Lp(a) levels are presented as the median value, and the case-control differences were compared using the Wilcoxon signed rank test. Lp(a) levels were logarithmically transformed before inclusion in multiple regression models. Distributions by case-control status of Lp(a) level and apo(a) size isoform, the latter as reflected by the number of apo(a) K4 repeats, were also compared using the Wilcoxon Mann-Whitney test against shift alternatives, which is of primary interest in this study. For analysis of apo(a) size isoforms in heterozygous subjects, when there was a predominantly expressed isoform, as determined by the intensity of the band in the immunoblotting, that K4 number was used in the analysis; when the two bands were of apparent equal intensity, we used the mean of the two K4 numbers. Alternative approaches for determining case-control differences in apo(a) size were also used. Statistical evaluation by paired t test was performed by either taking into consideration for heterozygous individuals the smaller of the two isoforms or the larger. Additionally, we used an approach in which, for heterozygotes, both isoforms were evaluated. The Mann-Whitney test was used in this instance rather than the paired t test because some cases had two isoforms, while their respective control subjects had only one isoform and vice versa. Conditional logistic regression modeling was used to adjust for covariates (see Table 1⇓) and to explore for interactions of the covariates with Lp(a) level.
No statistically significant differences in Lp(a) values were found in plasma samples from 20 individuals collected 15, 10, and 5 years before this study and stored at −70°C along with all the samples of the Stanford Five-City Project participants. In each of the 20 samples, the difference in Lp(a) values at the three different time points was within the expected biological variation of Lp(a) as previously reported,25 and no overall significant trend was evident. Additionally, the mean Lp(a) value of these frozen samples (54.4 nmol/L) was very similar to that obtained in 22 fresh samples from free-living subjects (48.9 nmol/L), analyzed for comparison purposes. Taken together, these results indicate that the integrity of Lp(a) data in our study was not compromised by potential problems related to sample handling and storage.
Comparison of CHD risk factors by case-control status and sex are presented in Table 1⇑. Paired t tests were used for all variables except where noted. Because the plasma samples were not drawn after a fast, we could not calculate the LDL cholesterol level; therefore, we present the difference between the total and HDL cholesterol levels as “non-HDL cholesterol.” All of the standard risk factors are in the expected direction and most attain statistical significance despite the limited sample size. The number of ever-smokers was higher in cases than in control subjects for both men and women, but the difference was not statistically significant (McNemar's test). Among the 44 women case-control pairs, 35 pairs were not estrogen users, 1 pair both took estrogen, and 8 pairs were discordant; in 5, the case took estrogen and in 3, the control. In the control group, median Lp(a) values for men and women (21.2 nmol/L and 24.3 nmol/L, respectively) were very similar to those obtained in over 2000 white individuals (men 19.4 nmol/L; women, 21.9 nmol/L; S.M. Marcovina et al, unpublished data). The median Lp(a) level was higher in cases than control subjects for both sexes, but the difference was statistically significant (Wilcoxon signed rank test) only in men in whom the Lp(a) concentration in cases was almost double that in control subjects (41.8 nmol/L versus 21.2 nmol/L). Median Lp(a) level was 34% higher in women cases than control subjects (32.5 nmol/L and 24.3 nmol/L, respectively), but this difference did not reach statistical significance. Considering that estrogen has been reported to affect Lp(a) levels, we evaluated case-control Lp(a) differences for only the 35 pairs of women who did not report estrogen use. The median Lp(a) value was 33.8 nmol/L in cases and 25.2 nmol/L in control subjects, P=.74. Thus, the case-control difference in Lp(a) level is almost identical in non–estrogen users to that in the entire sample.
Fig 1⇓ shows the positively skewed distribution of Lp(a) levels in cases and control subjects for men and women. In men, Lp(a) distribution was less skewed toward low levels in cases than in control subjects, and very few subjects in the control group had Lp(a) values >180 nmol/L. The Wilcoxon Mann-Whitney test for equality of distribution by case-control status was significantly different (P<.01), paralleling the results of the Wilcoxon signed rank test presented in Table 1⇑. In women, again confirming the results presented in Table 1⇑, there was not a significant difference in Lp(a) distribution between cases and control subjects.
Determination of apo(a) size isoforms showed that the large majority of subjects had two different isoforms and the percent of heterozygotes was identical in cases and control subjects (male, 82% versus 81% and female, 81% versus 81%). To evaluate the potential contribution of apo(a) size to CVD, case-control differences in apo(a) isoforms were analyzed. On the basis of the assumption that in heterozygous individuals the predominantly expressed isoform, as evaluated by the intensity of the band in immunoblotting, is the major contributor, that isoform was used for statistical analysis. By this approach, apo(a) size was smaller in male cases than in male control subjects, but the difference, as presented in Table 1⇑, was borderline significant (P=.07, paired t test).
The frequency distribution of apo(a) size isoforms, expressed in terms of the relative number of K4 repeats, by case-control status and sex is shown in Fig 2⇓. Male cases had a higher frequency of small apo(a) size isoforms, with fewer than 23 K4 repeats, whereas male control subjects had an increased frequency of intermediate and large apo(a) isoforms. This difference, evaluated by the Wilcoxon Mann-Whitney test, which is highly sensitive to an expected shift in distribution, was statistically significant (P=.05). To fully evaluate in men the association between small apo(a) size isoforms and CHD risk, we alternatively performed paired t test analyses by first taking into consideration, in heterozygous individuals, only the smaller of the two apo(a) isoforms. By this approach, the apo(a) sizes were significantly smaller in cases than in control subjects (P=.04). When the evaluation was performed considering only the larger of the two isoforms, again cases had significantly smaller apo(a) sizes than control subjects (P=.009). Additionally, when all the isoforms were included in the analysis, apo(a) isoforms in cases were shifted toward smaller sizes relative to control subjects (P=.04, Mann-Whitney test). As evident from Table 1⇑ and Fig 2⇓, no significant difference was found in apo(a) size between women cases and control subjects, and this result was regardless of the different approaches used to evaluate apo(a) isoform differences.
To evaluate the association between Lp(a) and case-control status for potential confounding by other measured risk factors, we calculated a series of models using conditional logistic regression. In each model the dependent variable was case-control status and the independent variables included log Lp(a) level as a continuous variable, one other risk factor, and an interaction term between Lp(a) level and the other factor. We chose this method rather than calculating a model with all potential confounders because of the small number of pairs and the potentially large number of interactions. We were particularly interested in the interactions, because Lp(a) may be a more important risk factor at different levels of other lipoproteins. We included those variables that approached statistical significance in the bivariate analysis in Table 1⇑. There was a strong confounder between smoking status and estrogen use in the women, which, along with the very small number of users, precluded this variable from analysis.
Crude and adjusted ORs for a three-quintile difference in log Lp(a) are presented in Table 2⇓. The unadjusted OR shows that a man with an Lp(a) level in the top quintile has 2.36 times the likelihood of being a case than a man with an Lp(a) level in the bottom quintile. As expected from the bivariate analyses, the crude OR is significant in men but not women. In men, when adjusted sequentially for the other lipid and lipoprotein variables, the association of Lp(a) with CHD does not change; plasma total, HDL, and non-HDL cholesterol are all significant predictors of CHD in men, but none of the interaction terms enter the model. Thus, Lp(a) level in men was a significant independent risk factor, and there was no evidence in this small data set that its effect was different at different levels of HDL or non-HDL cholesterol. Likewise, the association between Lp(a) level and case-control status was not affected by apo(a) size, which was not a significant predictor of CHD once Lp(a) concentration was entered. In women, only total and non-HDL cholesterol were significantly associated with the odds of being a case, and there was no evidence of an interaction between Lp(a) and these variables. Systolic blood pressure and cigarette smoking status were also analyzed but were not significant predictors of CHD in either sex, so the association with Lp(a) was also independent of these factors.
In addition to independence, the presence of a graded relationship between a putative risk factor and the outcome strengthens the likelihood that the relationship is causal. There is no method for exploring such a dose-response relationship in the present data while maintaining the case-control pairs. Thus, as an exploratory analysis, we broke the pairing and analyzed the odds of being a case by quintile of Lp(a) level in men. The quintile cut points were 6.3, 20.7, 37.5, and 112.5 nmol/L. Using the first quintile as the comparison group, the odds (and 95% CI) of being a case in each successive quintile (2 to 5) were, respectively, 0.77 (0.3-2.0); 0.89 (0.34-2.32); 2.00 (0.77-5.16); and 2.34 (0.90-6.08). Thus, the odds are higher in the highest two quintiles, but this pattern seems more compatible with a threshold risk value. Indeed, if we compare quintiles 4 and 5 with quintiles 1 and 2, the OR is 2.47 (1.26-4.83), which is statistically significant (P=.008).
If there is a threshold at which Lp(a) is a risk factor, then an obvious question is the level of Lp(a) that defines the threshold. To find the Lp(a) level that best discriminates between cases and control subjects, we employed signal-detection methods.26 In men, the most efficient threshold (maximizing both sensitivity and specificity) was 35 nmol/L; the OR for being a case above this threshold compared with below it was 2.84 (1.53-5.27; P<.001). As would be expected, no level of Lp(a) was found to discriminate cases and control subjects among the women.
In this prospective study of the associations between apo(a) polymorphs, Lp(a) levels, and CHD, we found Lp(a) levels to be higher in a group of subjects who later developed CHD than in matched subjects without CHD. However, when gender-specific statistical analyses were performed, no significant difference in the distribution of Lp(a) levels and apo(a) size isoforms between cases and control subjects was found in women. A striking difference in Lp(a) concentration was found in men, in whom cases had 97% higher Lp(a) levels than control subjects (41.8 nmol/L and 21.2 nmol/L, respectively, P<.01). Additionally, male cases had an increased frequency of small apo(a) size isoforms, and the difference in isoform distribution between cases and control subjects was statistically significant independent of the different approaches used to analyze apo(a) isoform data.
Our finding that in men Lp(a) concentration is predictive of CAD events is in agreement with the results reported in four prospective studies.8 9 10 11 However, three other nested case-control studies have provided contrasting results.12 13 14 In the Helsinki Heart Study, in which hypercholesterolemic participants were randomized to receive gemfibrozil or placebo, Lp(a) levels did not differ between 138 subjects with CHD and 130 control subjects in a group of 40- to 55-year-old men who were followed up for 5 years.12 A prospective study of 7424 Finnish men and women from the general population of 40 to 64 years of age free of atherosclerotic disease at baseline found no significant difference between Lp(a) levels in 134 male and 131 female case-control pairs.14 No difference was found in the Physicians' Health Study in Lp(a) levels between 296 men who had suffered an acute myocardial infarction during a mean follow-up period of 5 years and paired control subjects matched for smoking status and age.13 This study also included a randomized trial of aspirin and beta carotene, and these compounds may have affected the thrombogenic and atherogenic potential of Lp(a). Although treatment group was controlled for in the analysis, it is possible that an interaction between Lp(a) level and treatment group could have influenced the results. While it is very difficult to evaluate all the possible variables that could play a role in explaining the conflicting Lp(a) results obtained by the prospective studies, it is worthwhile to mention that our study is the first in which apo(a) isoforms have been determined and the first in which Lp(a) values have been accurately determined by a method insensitive to apo(a) size heterogeneity.
In the present study, the association of Lp(a) level and CHD status in men was independent of other risk factors and there was no evidence for a modification of this risk at different levels of HDL or non-HDL cholesterol. In exploratory analysis, the association in men appeared to emerge most strongly at Lp(a) levels >35 nmol/L. In women, the case-control difference in Lp(a) level was not statistically significant regardless of HDL and non-HDL cholesterol level.
The Framingham Heart Study27 has produced the only prospective study of elevated Lp(a) levels and CVD in women. However, Lp(a) values were not immunochemically determined, but a surrogate measure, the presence of a band representing sinking pre-β lipoprotein on paper electrophoresis of fresh plasma samples, was used. Band presence was a strong, independent predictor of myocardial infarction, intermittent claudication, and CVD in 3103 women observed for a median of 12 years. When the present manuscript was accepted for publication, the results of the prospective study of the Framingham offspring male cohort were reported.28 In that study, elevated Lp(a), determined by the same electrophoretic approach previously used for the female cohort,27 was an independent risk factor for development of premature CHD in men.
Our study demonstrates little evidence for an association between Lp(a) level or apo(a) size and CHD in women. The 95% CI for the crude OR (Table 2⇑) reflects the power of this study to detect a true difference in Lp(a) level in women. The most likely OR (1.25) is not very large for a three-quintile difference, but the CI (0.48-3.25) does include the level observed in men (2.36). The variance of log Lp(a) level is similar in men and women in this study, so the null hypothesis could be accepted with more assurance in a study in which at least 100 pairs of women are evaluated.
The strengths of this study include its prospective design, the storage of the sample at −70°C with documented lack of effect of the length of storage on Lp(a) measurements, the use of an Lp(a) assay that is independent of apo(a) size variability, the determination of apo(a) size isoforms, and the setting of a general population of mainly white ethnicity in five Northern California cities. The weaknesses include the relatively small number of case-control pairs and the reliance on history for determining disease-free status at baseline. The first weakness makes a type II error more likely, and thus the findings in women must be considered with caution and confirmed. The relatively small number of case-control pairs has also limited our ability to fully evaluate the possibility that the association of high Lp(a) values with small apo(a) size isoforms better defines the pathological role of Lp(a) than Lp(a) concentration alone. The fact that apo(a) size did not add to the predictive power of Lp(a) concentration should not be interpreted to mean that apo(a) size is necessarily unimportant. The results of statistical models of epidemiological data do not necessarily reflect the true underlying metabolic mechanisms, especially when the independent variables correlate, as do apo(a) size and Lp(a) concentration. Considering the high number of apo(a) isoforms and the low frequency of the small size polymorphs in the general population, a considerably larger number of subjects would need to be studied to enable us to derive definitive conclusions. However, despite these limiting factors, the results of this study strongly support the conclusion that Lp(a) concentration is an independent marker for CHD risk, at least in white men.
The presence of an elevated level of Lp(a) may dictate more aggressive efforts at lowering the conventional risk factors. It has been recently reported that when substantial LDL cholesterol reduction was obtained in men with CHD, persistent high levels of Lp(a) were no longer atherogenic or clinically threatening.29 Treatment directed at lowering the Lp(a) level would not be appropriate until the effectiveness of such treatment is established in trials or the mechanism for the increased risk is better understood. Lp(a) concentration may reflect underlying subclinical disease rather than necessarily participating in the causal pathway. Basic and clinical studies directed at elucidating potential causal mechanisms, and the possible male-female difference, are clearly required.
Selected Abbreviations and Acronyms
|CAD||=||coronary artery disease|
|CHD||=||coronary heart disease|
This work was supported by National Institutes of Health (Bethesda, Md) grants HL30086 (to Dr Marcovina) and HL21906 (to Drs Fortmann and Wild). Dr Wild was supported in part by a grant from the Dean, Stanford University School of Medicine. The authors gratefully acknowledge the technical assistance of Tom Meyers and Jean Mernaugh in the determination of Lp(a) values and apo(a) size isoforms, Ann Varady in data analysis, and Marlene Hunter in lipoprotein evaluation.
Lackner C, Cohen JC, Hobbs HH. Molecular definition of the extreme size polymorphism in apolipoprotein(a). Hum Mol Genet.. 1993;2:933-940.
Boerwinkle E, Leffert CC, Lin J, Lackner C, Chiesa G, Hobbs HH. Apolipoprotein(a) gene accounts for greater than 90% of the variation in plasma lipoprotein(a) concentrations. J Clin Invest.. 1992;90:52-60.
Utermann G. The mysteries of lipoprotein(a). Science.. 1989;246:904-910.
Marcovina SM, Levine DM, Lippi G. Lipoprotein(a): structure, measurement, and clinical significance. In: Rifai N, Warnick R, eds. Laboratory Measurement of Lipids, Lipoproteins, and Apolipoproteins. Washington, DC: AACC Press; 1994:235-263.
Rosengren A, Wilhelmsen L, Eriksson E, Risberg B, Wedel H. Lipoprotein(a) and coronary heart disease: a prospective case-control study in a general population sample of middle aged men. Br Med J.. 1990;301:1248-1251.
Cremer P, Nagel D, Labrot B, Mann H, Muche R, Elster H, Seidel D. Lipoprotein Lp(a) as predictor of myocardial infarction in comparison to fibrinogen, LDL cholesterol and other risk factors: results from the prospective Gottingen Risk, Incidence and Prevalence Study (GRIPS). Eur J Clin Invest.. 1994;24:444-453.
Jauhiainen M, Koskinen P, Ehnholm C, Frick HM, Manttari M, Manninen V, Huttunen JK. Lipoprotein(a) and coronary heart disease risk: a nested case-control study of the Helsinki Heart Study participants. Atherosclerosis.. 1991;85:59-67.
Marcovina SM, Albers JJ, Gabel B, Koschinsky ML, Gaur VP. Effect of the number of apolipoprotein(a) kringle 4 domains on immunochemical measurements of lipoprotein(a). Clin Chem.. 1995;41:246-255.
Farquhar JW, Fortmann SP, Maccoby N, Haskell WL, Williams PT, Flora JA, Taylor CB, Brown BW Jr, Solomon DS, Hulley SB. The Stanford Five City Project: design and methods. Am J Epidemiol.. 1985;122:323-334.
Fortmann SP, Haskell WL, Williams PT, Varady AN, Hulley SB, Farquhar JW. Community surveillance of cardiovascular diseases in the Stanford Five City Project. Am J Epidemiol.. 1986;123:656-669.
The Lipid Research Clinics Program Manual of Laboratory Operations, Lipid and Lipoprotein Analysis. 2nd ed. Washington, DC: US Department of Health and Human Services; 1982. (Originally published as US Dept of Health, Education, and Welfare publication NIH 75-628.)
Lackner C, Boerwinkle E, Leffert CC, Rahmig T, Hobbs HH. Molecular basis of apolipoprotein(a) isoform size heterogeneity as revealed by pulsed-field gel electrophoresis. J Clin Invest.. 1991;41:246-255.
Marcovina SM, Hobbs HH, Albers JJ. Relation between number of apolipoprotein(a) kringle 4 repeats and mobility of isoforms in agarose gel: bases for a standardized isoform nomenclature. Clin Chem.. 1996;42:436-439.
Warnick GR, Benderson J, Albers J. Dextran sulfate–Mg++ precipitation procedure for quantitation of high density lipoprotein cholesterol. Clin. Chem. 1982;28:1379-1388.
Fortmann SP, Marcuson R, Bitter PH, Haskell WL. A comparison of the Sphygmetrics SR-2 automatic blood pressure recorder to the mercury sphygmomanometer in population studies. Am J Epidemiol.. 1981;114:836-844.
Marcovina SM, Gaur VP, Albers JJ. Biological variability of cholesterol, triglyceride, LDL and HDL cholesterol, lipoprotein(a), apolipoproteins A-I and B. Clin Chem.. 1994;40:574-578.
Kraemer HC. Evaluating Medical Tests: Objective and Quantitative Guidelines. Newbury Park, Calif: Sage Publications, Inc; 1992.
Boston AG, Gagnon DR, Cupples A, Wilson PWF, Jenner JL, Ordovas JM, Schaefer EJ, Castelli WP. A prospective investigation of elevated lipoprotein(a) detected by electrophoresis and cardiovascular disease in women: the Framingham Heart Study. Circulation.. 1994;90:1688-1695.