|
|
| |||||||||||||||||||||||||||||||||||||||
Background Whether increasing parity or gravidity is a risk factor for coronary heart disease has been debated, but the question remains unresolved.
Methods We tested the association between the number of pregnancies and a variety of cardiovascular end points in two groups of women who had completed childbearing. One group comprised 2357 women who were followed for 28 years through the Framingham Heart Study, and the other 2533 women followed for at least 12 years through the first National Health and Nutrition Examination Survey National Epidemiologic Follow-up Study (NHEFS).
Results The rates of coronary heart disease were higher among multigravid women than among women who had never been pregnant, in both the Framingham Heart Study and the NHEFS, but in both studies, the higher rates were statistically significant only in women with six or more pregnancies. For the women in the Framingham Study, the rate ratio adjusted for age and educational level in the group with six or more pregnancies (as compared with women who had never been pregnant) was 1.6 (95 percent confidence interval, 1.1 to 2.2). For the women in the NHEFS, the same adjusted rate ratio was 1.5 (95 percent confidence interval, 1.1 to 1.9). Adjustments for other known cardiovascular risk factors, including weight, did not markedly alter this risk. The rate of total cardiovascular disease was also significantly higher among multigravid women in the Framingham Study than in the women who had never been pregnant.
Conclusions In two prospective American studies, having six or more pregnancies was associated with a small but consistent increase in the risk of coronary heart disease and cardiovascular disease. Whether gravidity itself or some other unmeasured factor accounts for the increase in risk that we observed requires further investigation.
The results of several investigations of the relation between reproductive history and coronary heart disease have been inconsistent; some case-control studies have shown positive associations for childbearing12,13,14,15 or age at first birth,16,17 and others have not shown such effects18,19,20,21. Similarly, three reports from two prospective studies have been conflicting22,23,24. Analyses from the Nurses' Health Study reported a slightly increased rate of coronary heart disease in parous as compared with nulliparous women, and two previous analyses from the Framingham Heart Study included findings of a slightly decreased rate and no trend. The limitations of the previous studies include the analysis of relatively small numbers of incident events, particularly in multiparous women, and the lack of adjustment for the effects of measured cardiovascular risk factors.
To explore the relation between women's reproductive histories and the occurrence of cardiovascular morbidity and mortality later in life, we studied two groups whose experience was prospectively ascertained: the complete cohort of women enrolled in the Framingham Heart Study through 28 years of follow-up, and the cohort of women 45 to 74 years of age who were enrolled in the first National Health and Nutrition Examination Survey (NHANES I) National Epidemiologic Follow-up Study (NHEFS), a nationally representative cohort of women, and followed for at least 12 years.
Methods
Study Population
Framingham Heart Study
The procedures for the recruitment and examination of subjects for the Framingham Heart Study have been described elsewhere25,26,27. We used the fourth biennial examination as a base line, because the majority of women had by then completed childbearing. Of the 2532 women who participated in the study, we excluded 47 who had not completed childbearing, 112 who had prevalent cardiovascular disease, and 16 who had missing information about pregnancy, leaving an overall sample of 2357 women at base line. A 28-year follow-up (to examination 18) was accomplished by biennial interviews and physician examinations. During this time, 15 women were lost to follow-up.
NHEFS
Base-line data for the NHEFS were taken from NHANES I, which was conducted by the National Center for Health Statistics from 1971 through 1975. This cross-sectional survey included interviews and physical examinations of a national probability sample of the U.S. civilian, noninstitutionalized population28,29. Participants ranging from 25 to 74 years of age were traced as part of the NHEFS30. Vital status was determined, and information was gathered from subject or proxy reports, medical and hospital records, and death-certificate data. For this paper, we used follow-up data collected through 1987.
We limited our sample to postmenopausal white women. Of the 3424 postmenopausal white women ranging from 45 to 74 years of age who were examined in NHANES I, we excluded 709 women who had prevalent coronary heart disease at base line, 9 who had missing information about pregnancy, 53 about whom there were no data on vital status, and 120 for whom information was missing on one or more covariates. The final sample included 2533 women.
Parity and Gravidity
Framingham Heart Study
The reported number of pregnancies for each woman was coded either as the actual number of pregnancies (for zero to seven pregnancies) or as eight (for eight or more pregnancies) at the first examination, and as either zero or one at each subsequent biennial examination. The number of pregnancies at base line was defined as the sum of the pregnancies reported at the first examination plus any additional pregnancies at the second through the fourth examinations.
NHEFS
Data on the total number of live births and the total number of pregnancies were available at base line. The results of analyses of parity and gravidity were similar; for consistency with the results from the Framingham Study, only the analyses for gravidity are presented here.
In both the Framingham and NHEFS data, women with more pregnancies had higher age-adjusted rates of cardiovascular disease than women with no or few pregnancies. For the purposes of presentation, the number of pregnancies was divided into four categories of zero, one to two, three to five, and six or more, to create more stable estimates.
Other Variables
Framingham Heart Study
Physiologic variables included systolic blood pressure, plasma glucose levels (measured by the Somogyi-Nelson method31), serum cholesterol values (measured by the Abell-Kendall method32), weight, height, and left-ventricular-hypertrophy status (according to standard electrocardiographic criteria),27 which was defined as present, possible, or not present. Body-mass index, calculated as the weight in kilograms divided by the square of the height in meters,33 and the presence of glucose intolerance34 were derived from these direct measures. Weight cycling was defined by a method similar to that used by Lissner et al.,35 as the coefficient of variation for the estimate of body-mass index at the age of 25 and the body-mass index at examinations 1 through 4. Fasting plasma lipoprotein cholesterol values were determined by a modification of the Lipid Research Clinics technique at examination 10, 11, or 1236. Other variables included educational status at examination 1 (categorized into nine groups ranging from women with no formal education to those with post-college study) and the number of cigarettes smoked per day. The use of estrogen-replacement therapy was measured at examination 2. In analyses using this variable, only postmenopausal women were included.
NHEFS
At base line, measured variables included systolic blood pressure and serum cholesterol levels (according to a modification of the Abell-Kendall method)28,29,37. Information was also obtained regarding the number of years of education (less than 9 years, 9 to 12 years, or more than 12 years), weight, height, and smoking history38 (current smokers, former smokers, or those who never smoked). Body-mass index was calculated. A history of diabetes was recorded if women had been told by a physician that they had diabetes. The use of estrogen-replacement therapy was recalled retrospectively at follow-up. Plasma glucose levels, electrocardiographic data, weight-cycling information, and plasma lipoprotein levels were not available.
Cardiovascular End Points
Framingham Heart Study
Cardiovascular events were defined with the use of well-established criteria, based on periodic physician review of examination notes, hospital and physician records, and death certificates25,26. For this study, the category of total cardiovascular disease included any of the following four end points: coronary heart disease (including angina pectoris, diagnosed myocardial infarction, unrecognized myocardial infarction, and coronary death or sudden death), cerebrovascular disease (including stroke or transient ischemic attack), congestive heart failure (either two of the major criteria or one of the major and two of the minor criteria described by Stokes et al.,27 including factors such as paroxysmal nocturnal dyspnea, distention of the neck veins when the patient is sitting, basal rales, roentgenographic evidence of cardiomegaly, and pulmonary edema), and intermittent claudication. Only the first occurrence of an end point was used in separate analyses for each end point; analyses for total cardiovascular disease used the first occurrence of any end point for a given woman. Since a woman could be recorded as having more than one type of end point, the sum of cases for the four end points was greater than the total number of cases of cardiovascular disease.
NHEFS
Incident cases of coronary heart disease were defined by the finding of a diagnosis of coronary heart disease, including angina pectoris (codes 410 through 414 of the International Classification of Diseases, 9th revision [ICD-9]), on hospital discharge records or any mention of coronary heart disease (ICD-9 codes 410 through 414) on a death certificate. Over 80 percent of cases were defined on the basis of data from hospital records.
Statistical Analysis
For descriptive purposes, the mean values of covariates for each gravidity category were adjusted for age by analysis of covariance. Statistical tests for trend were performed with the extension of the Mantel-Haenszel chi-square test39. The P values reported are two-tailed.
Cox proportional-hazards models were used to calculate rate ratios and 95 percent confidence intervals relating the number of pregnancies to the incidence of disease, with the number of pregnancies used as a categorical variable (0, 1 or 2, 3 to 5, or 6 or more)40. For the analyses of the data from the Framingham Study, multivariate Cox proportional-hazards models included age, body-mass index, plasma cholesterol level, systolic blood pressure, the number of cigarettes smoked per day, and educational level as continuous variables and the presence or absence of left ventricular hypertrophy and glucose intolerance as categorical variables. For the analyses of the NHEFS data, multivariate Cox proportional-hazards models included age, body-mass index, serum cholesterol level, and systolic blood pressure as continuous variables and smoking status, the presence of diabetes (as reported by the subjects), and educational status as categorical variables. Smoking status (current smokers, former smokers, or those who never smoked) rather than the number of cigarettes smoked per day was used because the former measure has been found to be more reliable than the latter in the NHEFS data38. The sampling weights were not used for the NHEFS analysis, as recommended and discussed by Korn and Graubard41. Base-line descriptive features were entered into the models, and continuous follow-up was used. Termination of follow-up occurred at the time of death or on loss to follow-up.
For some of the NHEFS analyses, women were stratified a priori according to age into two groups -- those 45 to 64 years of age at base line (n = 1240) and those 65 to 74 years of age at base line (n = 1293) -- because of the oversampling of older women in the sample. The women who were 45 to 64 years of age were more comparable to the women in the Framingham Study with respect to base-line age (35 to 68 years) and the crude incidence of coronary heart disease. The rates of coronary heart disease per 1000 woman-years of follow-up with zero, one to two, three to five, and six or more pregnancies were 9.0, 10.0, 10.2, and 17.9, respectively, for the Framingham Study; 9.7, 8.5, 11.6, and 16.5, respectively, for women in the NHEFS who were 45 to 64 years old; and 18.1, 18.7, 20.4, and 27.1, respectively, for all the women in the NHEFS.
Results
Descriptive Features
Framingham Heart Study
The 2357 women studied at the fourth biennial examination ranged in age from 35 to 68 years. The mean number of pregnancies was 2.3, and the median was 2. Women with greater numbers of pregnancies had significantly lower mean serum cholesterol values and educational levels, but a higher mean body-mass index than women with fewer pregnancies (Table 1).
|
The 2533 women in the NHEFS were 45 to 74 years of age at base line. They had both a mean and a median of 3.0 pregnancies. As in the case of the women in the Framingham Study, increased numbers of pregnancies were associated with significantly lower levels of education and higher body-mass indexes (Table 1). Total cholesterol levels at base line were somewhat lower in the women with six or more pregnancies, but followed no clear trend with respect to the other gravidity categories. Multigravid women also more commonly reported a history of diabetes than less prolific women. Whereas in the Framingham Study, the women with six or more pregnancies were also the oldest, in the NHEFS, women who were nulligravid were the oldest.
Gravidity and the Incidence of Cardiovascular Disease
Framingham Heart Study
As compared with women who had never been pregnant, the rates of coronary heart disease adjusted for age and education level were substantially higher in women with six or more pregnancies (rate ratio, 1.6; 95 percent confidence interval, 1.1 to 2.2) (Table 2). Adjustments for age, systolic blood pressure, number of cigarettes smoked, the presence of glucose intolerance, the presence of left ventricular hypertrophy, total cholesterol level, educational level, and body-mass index did not substantially alter this risk in women in the highest gravidity category (rate ratio, 1.7; 95 percent confidence interval, 1.2 to 2.4). These 161 women had 57 coronary heart disease events. At the highest level of gravidity, there was also an increase in the multivariate adjusted rate ratios of each of the other end points studied: cerebrovascular disease, congestive heart failure, intermittent claudication, and total cardiovascular disease, although this increase was only significant for total cardiovascular disease (adjusted rate ratio, 1.5; 95 percent confidence interval, 1.1 to 2.0) (Table 3).
|
|
As in the Framingham Study, there was an increase in the rate of coronary heart disease adjusted for age and educational level in multigravid women (rate ratio for six or more pregnancies, 1.5; 95 percent confidence interval, 1.1 to 1.9) (Table 2). This increase was small and remained statistically significant (P = 0.04) after an adjustment for age, smoking status (current or former), the presence of a history of diabetes, educational level, body-mass index, total cholesterol level, and systolic blood pressure (rate ratio, 1.3; 95 percent confidence interval, 1.0 to 1.8).
The relation between the number of pregnancies and coronary heart disease was also evaluated separately in two age categories. There was no apparent effect of a large number of pregnancies in women 65 to 74 years of age, whereas the effect was apparent in women 45 to 64 years of age. These results are, however, not statistically incompatible with a common effect of a large number of pregnancies in both groups; we found no statistically significant interaction between age and the effect of multigravidity on coronary heart disease (P = 0.33).
Other Analyses
A number of other analyses were performed. First, the elimination of nulliparous women or those who had never married from the sample had little effect on the reported rate ratios. Second, an adjustment for additional cardiovascular risk factors had little effect on the estimated rates of coronary heart disease that were associated with multigravidity. For example, in the women in the Framingham Study, the addition of estrogen-replacement use and weight cycling to multivariate models resulted in a rate ratio for women with six or more pregnancies of 1.5 (95 percent confidence interval, 1.0 to 2.1). Additional adjustments for the high-density lipoprotein cholesterol level and for the low-density lipoprotein level produced rate ratios in women with six or more pregnancies of 1.6 (95 percent confidence interval, 0.8 to 2.9) and 1.8 (95 percent confidence interval, 1.0 to 3.5), respectively. Because analyses including estrogen-replacement use were restricted to postmenopausal women, and because lipoproteins were not measured until examinations 10 to 12, a large number of women were eliminated from the cohort in each of these analyses, thereby reducing the power to detect associations. These adjustments also had inconsequential effects on the rate ratios for the number of pregnancies in relation to other cardiovascular end points. In women in the NHEFS, an additional adjustment for postmenopausal estrogen use minimally altered the rate ratios (rate ratio in all women with six or more pregnancies, 1.5; 95 percent confidence interval, 1.2 to 2.0).
Third, we reanalyzed the data after eliminating angina from the definition of coronary heart disease. Approximately half of the first cases of coronary heart disease among the women in the Framingham Study were due to angina, whereas only 14 percent of the cases in the NHEFS were due to angina. The elimination of these cases slightly diminished the strength of the association in the Framingham Study (rate ratio for women with six or more pregnancies, 1.4; 95 percent confidence interval, 0.9 to 2.2) but had no effect on the association in the NHEFS (rate ratio for women with six or more pregnancies, 1.3; 95 percent confidence interval, 1.0 to 1.7).
Fourth, we explored whether multigravid women employed outside the home were at particularly high risk of complications from coronary heart disease, as was found by Haynes and Feinleib24. Women who reported working outside the home in a full-time, part-time, or semiretired capacity and those who were self-employed made up the outside-employment group. Women who were housewives, unemployed, or retired made up the working-at-home group. The rate ratios for coronary heart disease in women with six or more pregnancies were similar in the outside-employment and working-at-home groups: 1.6 (95 percent confidence interval, 0.9 to 2.8) and 2.0 (95 percent confidence interval, 1.2 to 3.5), respectively.
Finally, we evaluated whether the effect of six or more pregnancies could be explained by differential trends in body-mass index, smoking status, blood pressure, cholesterol levels, glucose intolerance, and left-ventricular-hypertrophy status during the follow-up period. Between examinations 4 and 10, in each gravidity category, similar changes were noted in the age-adjusted mean levels of these risk factors, with the exception of the serum cholesterol level, which was increased in multigravid as compared with nulligravid women (13.3 vs. 6.9 mg per deciliter). Adding individual changes in risk-factor levels to multivariate models and starting the follow-up at examination 10 had little effect on the elevated rate ratios for women with six or more pregnancies (e.g., rate ratio for coronary heart disease, 2.2; 95 percent confidence interval, 1.3 to 3.5).
Discussion
These data show a consistent association between the number of pregnancies and subsequent cardiovascular disease in two separate studies, the Framingham Study and the NHEFS. In both studies, the independent effect was seen primarily in women with six or more pregnancies. In both studies, an adjustment for a variety of cardiovascular risk factors had little effect on the estimates of rate ratios, and the rate ratios were similar in both studies despite the different methods of measuring other risk factors. Because the strength of the observed association is slight, further exploration of this relation will be important to exclude the possibility that unmeasured confounding factors, rather than the number of pregnancies, account for the elevated rates of cardiovascular disease in multigravid women.
The current study differs from previous analyses in several ways. First, both the data sets examined provided more power to detect small associations, particularly in multigravid women, than did previous studies. Indeed, the 95 percent confidence interval for the association between parity and coronary heart disease reported in the Nurses' Health Study (0.7 to 1.4)22 includes the NHEFS point estimate. Second, we controlled for objectively measured cardiovascular disease risk factors at base line. The Nurses' Health Study analyzed self-reports of risk factors, and one previous report from the Framingham Study23 did not assess the effect of known risk factors. Third, in contrast to the Nurses' Health Study,22 angina was included as a component of the coronary heart disease end point in the Framingham Study and the NHEFS, and thus disease was detected earlier. It might be postulated that the misclassification of angina introduces bias into our results if angina is diagnosed more often in multigravid women than in nulligravid women. The exclusion of angina as an end point slightly reduced the rate ratio for coronary heart disease in the women in the Framingham Study but not in those in the NHEFS. It is unclear whether the loss of statistical significance in relating multigravidity and coronary heart disease when angina was excluded as an end point is a result of an inadequate sample to demonstrate the association or reflects the absence of a true association in the women in the Framingham Study.
How might the association between multigravidity and cardiovascular disease be mediated? Pregnancy may affect the risk of cardiovascular disease through insulin resistance, a major cardiovascular risk factor in perimenopausal women42. Insulin resistance is greater in multigravid women than in women with fewer pregnancies, perhaps as a result of post-pregnancy alterations in the distribution of body fat43,44,45. Our data did not allow an adequate evaluation of this hypothesis, because insulin resistance and the distribution of body fat in the perimenopausal period were not measured. Of the factors available for analysis, body-mass index, waist-to-hip ratios, the presence of glucose intolerance, and weight cycling did not alter the association between the number of pregnancies and cardiovascular disease in the women in the Framingham Study. Also, cholesterol and lipoprotein levels did not seem to mediate the observed association, despite previous reports that multigravid women have lower high-density lipoprotein cholesterol levels years after childbearing8,9.
Our findings may be surprising in view of data suggesting that larger, more supportive social networks of friends and relatives reduce cardiovascular mortality in women46. However, the specific role of large numbers of children in the social network is not known47. In one study, informal networks of nonrelatives were stronger among older women with fewer children. Thus, the absence of children may compel women to form equally important informal networks.
If verified, our findings raise questions about the mechanism by which an association between multigravidity and cardiovascular risk may be mediated.
Supported in part by a training grant (5T32CA09529) from the National Institutes of Health.
Source Information
From the Clinical Epidemiology Unit and the Emergency Department, Department of Medicine (R.B.N.), and the Department of Psychiatry (A.J.S.), University of Pennsylvania, Philadelphia; the National Center for Health Statistics, Centers for Disease Control and Prevention, Hyattsville, Md. (T.H., K.M.F.); the Department of Health Research and Policy, Stanford University, San Francisco (J.L.K.); and the Department of Mathematics, Statistics Unit, Boston University, Boston (J.C., A.B., R.B.D.).
Address reprint requests to Dr. Ness at the University of Pittsburgh Graduate School of Public Health, Department of Epidemiology, Rm. A527, Crabtree Hall, Pittsburgh, PA 15261.
References
| |||||||||||||||||||||||||||||||||||||||
Related Letters:
Number of Pregnancies and Risk of Cardiovascular Disease
Grinblatt J. A., Dekker J. M., Schouten E. G., Palmer J. R., Rosenberg L., Shapiro S., Ness R. B., Harris T., Cobb J.
Extract |
Full Text
N Engl J Med 1993;
329:1893-1895, Dec 16, 1993.
Correspondence
This article has been cited by other articles:
HOME | SUBSCRIBE | SEARCH | CURRENT ISSUE | PAST ISSUES | COLLECTIONS | PRIVACY | TERMS OF USE | HELP | beta.nejm.org Comments and questions? Please contact us. The New England Journal of Medicine is owned, published, and copyrighted © 2009 Massachusetts Medical Society. All rights reserved. |
Previous
