Lipoprotein(a) as a Determinant of Coronary Heart Disease in Young Women
Background Lipoprotein(a) [Lp(a)] appears to be a risk factor for coronary heart disease (CHD) in men. The role of Lp(a) in women, however, is less clear.
Methods and Results We examined the ability of Lp(a) to predict CHD in a population-based case-control study of women 65 years of age or younger who lived in the greater Stockholm area. Subjects were all patients hospitalized for an acute CHD event between February 1991 and February 1994. Control subjects were randomly selected from the city census and were matched to patients by age and catchment area. Lp(a) was measured 3 months after hospitalization by use of an immunoturbidometric method (Incstar) calibrated to the Northwest Lipid Research Laboratories (coefficient of variation was <9%). Of the 292 consecutive patients, 110 (37%) were hospitalized for an acute myocardial infarction, and 182 were hospitalized (63%) for angina pectoris. The mean age for both patients and control subjects was 56±7 years. Of participants, 74 patients (25%) and 84 control subjects (29%) were premenopausal. The distributions of Lp(a) were highly skewed in both patients and control subjects, with a range from 0.001 to 1.14 g/L. Age-adjusted odds ratio for CHD in the highest versus the lowest quartile of Lp(a) was 2.3 (95% confidence interval [CI], 1.4 to 3.7). After adjustment for age, smoking, education, body mass index, systolic blood pressure, total cholesterol, triglycerides, and HDL, the odds ratio was 2.9 (95% CI, 1.6 to 5.0). The odds ratios were similar when myocardial infarction and angina patients were compared with their respective control subjects. The odds ratios were 5.1 (95% CI, 1.4 to 18.4) and 2.4 (95% CI, 1.3 to 4.5) in premenopausal and postmenopausal women, respectively.
Conclusions These results suggest that Lp(a) is a determinant of CHD in both premenopausal and postmenopausal women.
The role of Lp(a) in the pathogenesis of CHD has been subject to recent debate.1 In men, the epidemiological evidence of Lp(a) as a risk factor for clinical CHD events is contradictory; in women, it is virtually absent. Several clinical and epidemiological studies have shown an association between elevated levels of Lp(a) and the occurrence of acute CHD events in men.2 3 4 5 Most convincingly, results from the Lipid Research Clinic demonstrated Lp(a) to be an independent predictor of future nonfatal myocardial infarction and CHD death in hyperlipemic men 35 to 59 years of age.2 The empirical evidence in women is scarce, and population-based studies of Lp(a) in young women are lacking. In a report from the Framingham Heart Study based on 3103 women followed for a median of 12 years, the presence of sinking prebeta lipoprotein on electrophoresis, a proxy measure for Lp(a), was a strong predictor of fatal and nonfatal CHD.6
Lp(a) is made up of apolipoprotein B-100 (the apolipoprotein of LDL cholesterol) disulfide linked to apolipoprotein(a), a large and polymorphous glycoprotein with structural homology to the fibrinolytic proenzyme plasminogen. This structural homology has led to the hypothesis that Lp(a) may represent an important link between atherosclerosis and intravascular thrombosis. Indeed, several in vitro studies have demonstrated that Lp(a) competes with plasminogen for binding sites on endothelial cell surfaces.7 8 Furthermore, Lp(a) has been shown to inhibit activation of plasminogen by tissue-type plasminogen activator.9 In addition to these effects, Lp(a) may promote atherosclerosis when oxidized Lp(a) particles are engulfed by macrophages, causing their transformation into foam cells within the arterial wall.10 11 The accumulation of Lp(a) in atherosclerotic plaque has been demonstrated repeatedly in postmortem studies and more recently in fresh human arterial wall tissue.12
We have studied Lp(a) in the Stockholm Female Coronary Risk Study, a population-based case-control study of women 65 years of age or younger who during a 3-year period were hospitalized in the greater Stockholm area for an acute CHD event. The women with CHD were compared with age-matched healthy women from the census register of the same catchment area.
Recruitment of Patients and Control Subjects
All patients who were admitted between February 1991 and February 1994 for an acute CHD event (either acute MI or stable or unstable angina pectoris) and resided in the greater Stockholm area were asked to participate in the study. Stockholm, the capital of Sweden, has a population of 1.2 million. The study group was restricted to patients 65 years of age or younger who spoke Swedish. The Swedish healthcare system provides care to all residents, regardless of income, socioeconomic, or insurance status. Thus, we were certain to reach virtually all patients who needed and sought hospital care for an acute CHD event during this time period.
Recruitment was carried out through a network of contact nurses at the 10 existing coronary care units who reported on a weekly basis all hospitalized cases of suspected acute MI and stable or unstable angina pectoris in female patients 65 years of age or younger. In Stockholm, the criteria for admission to intensive coronary care units are the same at all 10 cardiology clinics. They have been developed and agreed on by a special clinical coordination group. Patients were included in the study if their hospital records indicated any of the following criteria: (1) definite or suspected MI based on the World Health Organization criteria of typical chest pain, typical enzyme patterns, and diagnostic ECG changes13 (ECG changes were classified by use of the Minnesota code14 ); (2) unstable angina pectoris defined as newly debuted severe angina pectoris that had deteriorated during the last 4 weeks before admission, with an increase in pain intensity and pain duration or with pain at rest or very low physical exertion15 ; or (3) spasmangina, defined as angina pectoris at rest with pathological ST-segment changes on ECG and with normal coronary arteries on acute clinical coronary angiography.
During the 3-year study period, 335 women with CHD were identified in the coronary care units of the 10 cardiology clinics. All patients were contacted first by mail and then by phone.
Of the identified patients, 43 (13%) could not be included in the study. Of these patients, 5 died during the 3 months between hospitalization and examination, 13 were too sick to come to the research center, 2 could not participate because of transportation difficulties, and 2 declined because they were recruited for other studies. Another 21 patients declined to participate for other reasons, including inability to speak Swedish fluently. Using hospital administrative registers in the same catchment area, we compared a subsample of the study recruitment group with register data on intensive coronary care patients. The proportion of patients missed in our study group was 12%.
Control subjects were selected from the census register of greater Stockholm. This register is based on the person identification number of the residents in Stockholm. This 10-digit identification number is unique for each individual and is assigned to each resident either at birth or on immigration to Sweden. It is based on birth date and sex. Therefore, identification of closely age-matched control subjects was possible. For each patient, a healthy woman born on the same day or another day as close as possible who lived in the same hospital catchment area as the patient was chosen. “Healthy” was defined as being free of symptoms of heart disease and without hospitalization for any illness during the prior 5-year period. Control subjects were contacted in the same way as patients. A letter explaining the objectives and the focus of the study and inviting them to participate was mailed to them. Those who did not call the clinic spontaneously were then contacted by phone. Of those eligible, 17% declined to participate, mainly because of difficulties in arranging time off from work to participate in the study.
All patients were examined between 3 and 6 months after hospitalization. Matched control subjects were examined during a corresponding time period. For each subject, the study was carried out during 2 consecutive days. Patients continued their usual medications, which were verified on arrival at the research clinic. One patient was taking nicotinic acid, the only hypolipemic agent likely to affect Lp(a) levels. Her Lp(a) level was 0.233 g/L.
A questionnaire was mailed to subjects before their visit to the research clinic. The questionnaire included questions about educational level and smoking history. In addition, lifestyle, behavioral, and psychosocial measures were included. Each questionnaire was checked by the research nurse for problems and/or missing data. For this report, educational level was divided into three categories: mandatory level (corresponding to 9 years of school education), high school or partial college education, and completion of college. Smoking status was categorized as follows: never smoked, smoked previously, or currently smoking.
The first day of the study included a detailed cardiological examination, resting and exercise ECGs, and placement of a 24-hour Holter ECG monitor. The second day included extensive interview and questionnaire assessments of lifestyle and behavioral characteristics, as well as anthropometric measures and full lipid and routine laboratory profiles.
Subjects arrived at the research clinic between 8 and 10 am of the second day, having fasted since midnight. Blood samples were drawn with subjects in the supine position after 5 minutes' rest. Height was measured in centimeters; weight was measured in kilograms. BMI was obtained as height divided by weight squared. Systolic and diastolic blood pressures were measured with subjects in the supine position after 5 minutes' rest. Phases I and V of the Korotkoff sound were used.
Menopausal status was assessed with a gynecological interview by the research nurse. Postmenopausal status was defined as having had no menses for at least 6 months. A history of gynecological surgery was obtained. A patient was classified as having had surgical menopause if she had undergone a bilateral oophorectomy. A complete history regarding HRT was also obtained. Women who had begun HRT before menopause were considered menopausal if they were >50 years of age.
Venous blood samples were drawn from the right arm of each patient and control subject into serum-separated tubes, which were centrifuged for 10 minutes at 3000g. Plasma (4 mL) was obtained and frozen to −70°C. They were sent in batches to the processing laboratory (CALAB) once per month. Tubes were identified by number only, and laboratory personnel were blinded as to case or control status. Each batch contained samples from both patients and control subjects in random order.
Lp(a) was analyzed by use of an immunoturbidometric method (INCSTAR Corp). Calibrators were referenced to the Northwest Lipid Research Laboratories' reference for Lp(a) (coefficient of variation was <8.8%).16 Total cholesterol was determined with CHOD-PAP; triglycerides, with GPD-PAP enzymatic methods with reagents from Boehringer Mannheim (Germany). HDLs were determined on the basis of the isolation of LDL and VLDL from serum by precipitation. The cholesterol content of the supernatant, ie, HDL cholesterol, was measured enzymatically.17 All measurements were carried out in the same laboratory (CALAB) with an automated multichannel analyzer.18
Differences in baseline characteristics between patients and control subjects were evaluated with Wilcoxon signed rank test and χ2 tests for continuous and discrete variables, respectively. Because of the highly skewed distribution of Lp(a), comparisons of the distribution of Lp(a) between patients and control subjects were made by use of the Wilcoxon signed rank test and the Kolgomorov-Smirnov test. Odds ratios were estimated as a measure of relative risk with logistic regression models. In all logistic models, the study sample was divided into quartiles of Lp(a) that were based on the distribution among the control subjects. Because patients were matched to control subjects by age, age was included as a covariate in all models. Multivariable logistic regression models controlling for age, smoking, education, BMI, systolic blood pressure, total cholesterol, triglycerides, and HDL were performed. A test for linear trend was used to assess the association between Lp(a) level and CHD risk.19 A final set of multivariate models was produced in which the natural logarithm of Lp(a) was entered into the logistic model to compare the results obtained with the quartile-based analyses described above.
The study sample included 292 patients and as many age-matched control subjects. Of the 292 patients, 110 (37%) were hospitalized for acute MI, and 182 (63%) were hospitalized for angina pectoris. Table 1⇓ gives the baseline characteristics of CHD patients and control subjects. The average age of both the patients and control subjects was 56±7 years. The youngest patient was 30 years old; 20% of the patients were <50 years of age, and 40% were >60 years.
Smoking, defined as current or previous smoking, was present in 198 patients (68%) and 157 control subjects (54%; P<.001). Low socioeconomic status, as estimated by not having attained more than mandatory education, was found in 181 patients (62%) and 152 control subjects (52%; P=.02).
Postmenopausal status, defined as being free of menses for at least 6 months, was found in 218 patients (75%) and 207 control subjects (71%). Surgical menopause was found in 21 patients (7%) and 16 control subjects (5%). HRT was present in 36 patients (12%) and 40 control subjects (14%). Among women on HRT, 6 patients and 6 control subjects had had surgical menopause. The remaining women were on HRT for perimenopausal symptoms.
Obesity was more prominent in patients than in control subjects (P<.001), and systolic blood pressure was higher, although not significantly. A history of hypertension, verified by treatment with medications or a physician's report of high blood pressure (≥160/90 mm Hg), was found in 144 patients (50%) and 31 control subjects (11%; P<.001). Patients had significantly higher levels of total cholesterol (P<.001) and triglycerides (P<.001) and lower levels of HDL cholesterol (P<.001). The Figure⇓ shows the distribution of Lp(a) levels in patients and control subjects. The distributions of Lp(a) values for both groups were highly skewed, ranging from 0.001 to 1.14g/L. The mean Lp(a) value in patients (0.281±0.238 g/L) was significantly higher than in control subjects (0.2150.203 g/L; P=.002). Median Lp(a) values were 0.200 g/L for patients and 0.145 g/L for control subjects. Correlations between Lp(a) and other clinical measures listed in Table 1⇑ were all weak and nonsignificant (data not shown).
Table 2⇓ gives the odds ratios for being a CHD patient in each of the Lp(a) quartiles. With the lowest quartile as the reference category, the age-adjusted odds ratio increased with increasing Lp(a) levels to reach 2.3 (95% CI, 1.4 to 3.7) in the highest quartile (P<.001 for trend). Further adjustment for smoking, education, BMI, systolic blood pressure, cholesterol, triglycerides, and HDL slightly increased the odds ratio to 2.9 (95% CI, 1.6 to 5.0; P<.001 for trend). Table 3⇓ shows that among both premenopausal and postmenopausal women, the odds ratios increased with increasing levels of Lp(a). The multivariable adjusted odds ratios for women in the highest versus lowest quartile of Lp(a) were 5.1 (95% CI, 1.4 to 18.4) and 2.4 (95% CI, 1.3 to 4.5) for premenopausal and postmenopausal women, respectively. Furthermore, for both MI and angina patients, the odds ratios also increased with increasing levels of Lp(a). The multivariable adjusted odds ratios for women in the highest versus lowest quartile of Lp(a) were 2.7 (95% CI, 1.2 to 5.7) and 2.9 (95% CI, 1.5 to 5.5) for MI and angina patients compared with their respective control subjects.
Similar results were obtained from analyses that treated Lp(a) as a continuous variable. The multivariable adjusted odds ratio comparing the median level of Lp(a) in the highest and lowest quartiles (0.02 to 0.46 g/L) resulted in an odds ratio of 2.2 (95% CI, 1.8 to 3.7).
In this study, we found that increasing levels of Lp(a) were associated with increasing risk of CHD in women. These results were consistent in both premenopausal and postmenopausal women, all of whom were 65 years of age or younger. The risk of hospitalization for an acute CHD event was 2.9 times higher for women in the highest quartile of Lp(a), corresponding to a level >0.30 g/L, compared with women in the lowest quartile. These results persisted in multivariable logistic regression models that controlled for age, smoking, education, BMI, systolic blood pressure, cholesterol, triglycerides, and HDL, indicating that Lp(a) contributed significantly to the prediction of CHD, independent of the potentially confounding covariates.
These findings in Swedish women are consistent with the recent report of a female cohort (mean age, 45 years) of the Framingham Heart Study. The Framingham study used the presence of a sinking prebeta lipoprotein band on electrophoresis as a surrogate measure for Lp(a). In a subsample, this measure was found to have low sensitivity (51%) but high specificity (95%) for elevated levels of Lp(a), as assessed by an ELISA. It is interesting to note that the Framingham odds ratio for MI cases6 (2.4; 95% CI, 1.5 to 3.8) is very similar to the odds ratio of the present study.
Lp(a) has been studied more frequently in men than in women. Among 3806 hyperlipidemic men of the Lipid Research Clinic Trial, CHD patients were found to have 20% to 22% higher Lp(a) levels than control subjects who were matched for age and other variables. The odds ratio for the highest compared with the lowest quintile of Lp(a) was 2.1 (95% CI, 1.2 to 3.6).2 Furthermore, in 776 50-year-old Swedish men randomly obtained from the population of Gothenburg, the risk during 6-year follow-up of fatal and nonfatal CHD in the highest Lp(a) quintile was about twice as high as for the lowest quintile.3 Likewise, in 1332 Icelandic men followed for 8.6 years, Lp(a) was an independent risk factor for fatal and nonfatal CHD, with an odds ratio of 1.2 per 1-SD increase in Lp(a).5
In contrast, in the Helsinki Heart Study, elevated levels of Lp(a) did not significantly predict CHD in 4081 hyperlipidemic men. However, there was a trend toward an increased CHD risk, with an odds ratio for the highest versus lowest Lp(a) tertile of 1.3, and the CI included effects as large as those seen in positive studies (95% CI, 0.8 to 2.0).20
In another study, 14 916 male physicians, 40 to 84 years of age, were followed for 5 years. Lp(a) was completely unrelated to fatal or nonfatal MI. Both range and distribution of Lp(a) were remarkably similar in patients and nested control subjects.21 In a cross-sectional analysis of 1202 men and 1512 women >60 years of age in the Dubbo study in Australia, Simons et al22 found no significant increase in Lp(a) in patients with prevalent CHD. Mean levels were generally higher in women than in men, and the odds ratios of CHD reached 1.4 and 1.3, respectively, in the fourth and fifth quintiles of the female Lp(a) distribution. These odds ratios were higher than those in men, but the differences were not statistically significant.
In a case-control study based on the population of Jerusalem, increased Lp(a) levels were associated with first MI in men but not in women.4 However, blood samples were stored at −20°C for several months, and Lp(a) levels were generally low, suggesting a relative degradation of the lipoprotein over time. In addition, the number of female subjects in the Jerusalem study was probably too small (n=47) to draw conclusions about risk.
The first study to show an association of Lp(a) with significant CAD as assessed by coronary angiography was performed in 86 women and 216 men. Presence of significant CAD was related to increased levels of Lp(a) in women of all ages but only in men <55 years of age.23 Furthermore, in two large series of male and female patients who underwent coronary angiography, Lp(a) was found to predict significant CAD and to be a better predictor in women than in men.24 25 In 1054 Belgian patients, one third of whom were women, the odds ratios of significant CAD in the highest versus the lowest Lp(a) quintiles were 2.4 for women and 1.6 for men. These were obtained after controlling for age and other lipid variables.24 In a recent report by Terres et al,26 Lp(a) was related to rapid progression of CAD over 66 days, regardless of the patient's sex. The proportion of female patients was not specified. In another angiographic follow-up study, Lp(a) predicted restenosis after percutaneous transluminal coronary angiography in 240 consecutive patients. Sex differences were not reported.27
Thus, in younger women, the Framingham and the Stockholm studies suggest a strong association of Lp(a) with CHD risk, with odds ratios >2. One cross-sectional study of women >60 years of age found an increased odds ratio but a nonsignificant risk estimate. Finally, another case-control study of female CHD patients found no association. The latter study, however, was probably too small to be conclusive.
The present study has several potential limitations. Like all case-control studies, there is the possibility of case selection bias. To minimize this bias, we used a population-based approach and attempted to enroll all consecutive patients in the catchment area. Matched control subjects also were randomly sampled from the same population from which the patients were obtained. Furthermore, unlike many other study populations, this group of Swedish women had an extremely high participation rate (87% for patients; 83% for control subjects).
There is also the possibility that Lp(a) levels may have been influenced by the occurrence of acute coronary events among the patients. Slunga et al28 have shown that Lp(a) levels are transiently elevated after an acute MI. However, the levels invariably return to baseline within 30 days29 and therefore reflect long-term Lp(a) levels. Thus, to minimize this bias, we evaluated patients 3 to 6 months after the acute event.
Another problem is the possibility of survivor bias because we were able to obtain data only from the women who survived for 3 months after an acute coronary event. In our study, only 5 of the 335 eligible women (1%) died between discharge from the hospital and examination. However, women who died suddenly before reaching the hospital were excluded. Although we do not know the exact figure for the Stockholm area, an average of 20% to 30% of MI patients are reported to die before reaching a hospital.30 Because roughly one third of our patients were hospitalized for MI, about 6% to 10% of the target population may not have been reached. If Lp(a) levels were strongly related to prognosis, then it is possible that the distribution of Lp(a) among the patients could be distorted. Because higher Lp(a) levels are associated with rapid progression of atherosclerosis,26 it is possible that this may lead to a poorer prognosis. The effect of such a bias, however, would be to produce an underestimation of the true odds ratio.
Although controversy exists over the ideal method to measure Lp(a), the immunoturbidometric method used in this study had a low coefficient of variation and was calibrated to the Northwestern Lipid Research Laboratories. Furthermore, to minimize extraneous variability, we used uniform blood drawing and specimen handling techniques. All specimens were centrifuged immediately and stored at −70°C for <1 month before the Lp(a) assays were performed. Previous validation studies have documented that storage of serum at −70°C shows no degradation in Lp(a) over a 6-month period.31 To reduce bias further, patient and control specimens were processed in the same batches, and laboratory personnel were blinded as to status. Thus, any bias resulting from measurement error of Lp(a) would tend to produce an underestimation of the true odds ratio. Furthermore, the distribution of Lp(a) in our study is similar to that in other female and male white populations.2 21 32
In summary, the results suggest an effect of Lp(a) on CHD risk in younger women. However, longitudinal studies of Lp(a) in women are needed before we can recommend that this risk factor be used as a screening tool. Because Lp(a) is genetically determined, we also recommend further studies to examine the relationship between family history of CHD and Lp(a) levels. Increased knowledge of the role of Lp(a) as a risk factor for CHD in women would be of great benefit. High-risk women would be more easily identified and could possibly be treated for other existing risk factors.
Selected Abbreviations and Acronyms
|BMI||=||body mass index|
|CAD||=||coronary artery disease|
|CHD||=||coronary heart disease|
|HRT||=||hormone replacement therapy|
This work was supported by grant HL-45785 from the US NIH, grant B93-19X-10407 from the Swedish Medical Research Council, and a grant from the Swedish Labour Market Insurance Company. We thank Diane Walkoff, BS, and CALAB, Stockholm, for excellent technical assistance.
- Received July 1, 1996.
- Revision received September 9, 1996.
- Accepted September 9, 1996.
- Copyright © 1997 by American Heart Association
Rosengren A, Wihelmsen L, Eriksson E, Risberg B, Wedel H. Lipoprotein(a) and coronary heart disease: a prospective case-control study in the general population sample of middle aged men. BMJ. 1990;301:1248-1251.
Bostom AG, Gagnon DR, Cupples LA, 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. Circulation. 1994;90:1688-1695.
Harpel PC, Gordon BR, Parker TS. Plasmin catalyses binding of lipoprotein(a) to immobilized fibrinogen and fibrin. Proc Natl Acad Sci U S A. 1989;86:3847-3851.
Myocardial Infarction Community Registers. Results of WHO international collaborative study. WHO Regional Office for Europe, Copenhagen, Sweden; 1976.
Gillom RF, Fortman SP, Princas RJ, Kottke TE. International diagnostic criteria for acute myocardial infarction and stroke. Prog Cardiol. 1984;108:150-157.
Braunwald E. Unstable angina: a classification. Circulation. 1989;80:410-414.
Riepponen P, Marniemi J, Rautaoja T. Immunoturbidimetric determination of apolipoprotein A-1 and B in serum. Scand J Clin Lab Invest. 1987:47;739.
Rothman KJ. Modern Epidemiology. Boston, Mass: Little, Brown & /001/Co; 1986.
Dahle´n GH, Guyton JR, Attar M, Farmer JA, Kautz JA, Gotto AM Jr. Association of levels of lipoprotein Lp(a), plasma lipids, and other lipoproteins with coronary artery disease documented by angiography. Circulation. 1986;74:758-765.
Labeur C, De Bacquer D, De Backer G, Vincke J, Muyidermans L, Vanderkerhove Y, Van der Stichele E, Rousseneu M. Plasma lipoprotein(a) values and severity of coronary artery disease in a large population of patients undergoing coronary angiography. Clin Chem. 1992;38:2261-2266.
Terres W, Tatsis E, Pfalzer B, Beil U, Beisiegel U, Hamm ChW. Rapid angiographic progression of coronary artery disease in patients with elevated lipoprotein(a). Circulation. 1995;91:948-950.
Desmarais RL, Sarembock IJ, Ayers CR, Vernon SM, Powers ER, Gimple LW. Elevated serum lipoprotein(a) is a risk factor for clinical recurrence after coronary balloon angioplasty. Circulation. 1995;91:1403-1409.
American Heart Association. Heart and Stroke Facts: 1995 Statistical Supplement. Dallas, Tex: American Heart Association; 1994.
Jenner JL, Ordovas JM, Lamon-Fava S, Schaefer MM, Wilson PW, Castelli WP, Schaefer EJ. Effects of age, sex and menopausal status on Lp(a) levels: Framingham Off-spring Study. Circulation. 1993;87:1135-1141.