Birth Weight and Adult Hypertension and Obesity in Women
Background Low birth weight has been associated with an increased risk of hypertension, and high birth weight has been associated with increased adult body mass index. Published studies on adults have included only a small number of women.
Methods and Results We studied 71 100 women in the Nurses' Health Study I (NHS I) who were 30 to 55 years of age in 1976 and 92 940 women in the Nurses' Health Study II (NHS II) who were 25 to 42 years of age in 1989. Information on birth weight, blood pressure, physician-diagnosed hypertension, and other relevant variables was collected by biennial mailed questionnaire. Ninety-five percent of the women were white. Compared with women in the middle category of birth weight (NHS I, 7.1 to 8.5 lb; NHS II, 7.0 to 8.4 lb), the age-adjusted odds ratio of hypertension in NHS I women with birth weights <5.0 lb was 1.39 (95% CI, 1.29 to 1.50); in NHS II, for birth weights <5.5 lb, the age-adjusted odds ratio was 1.43 (95% CI, 1.31 to 1.56). There was no material change in the estimates after adjustment for other risk factors. In addition, compared with women in NHS I who weighed 7.1 to 8.5 lb at birth, those who weighed >10 lb had an age-adjusted odds ratio of 1.62 (95% CI, 1.38 to 1.90) of being in the highest (>29.2 kg/m2) versus the lowest (<21.9 kg/m2) quintile of body mass index in midlife. Similar results were seen in the NHS II cohort.
Conclusions Early life exposures affecting birth weight may be important in the development of hypertension and obesity in adults.
Early life exposures may be important in the eventual development of chronic diseases in adult life. For example, low birth weight has been associated with an increased risk of cardiovascular disease,1 diabetes mellitus,2 hypertension,3 and chronic obstructive pulmonary disease.4 Obesity, another chronic condition in adults, may also partially originate in infancy.5 6 7 8 Identifying early life risk factors for these chronic diseases, which are major causes of morbidity and mortality in the United States and other developed countries, may lead to a better understanding of their origins and improved preventive strategies.
The prevalence of hypertension, defined as systolic blood pressure ≥140 mm Hg and/or diastolic blood pressure ≥90 mm Hg, is ≈20% in the general US population and increases to 67% in those ≥65 years of age.9 The cause of the elevated blood pressure in the vast majority of individuals is unknown; this condition is referred to as essential hypertension.10 Previously identified risk factors for hypertension include diet, sex, ethnicity, and obesity.11 Recently, it was suggested that birth weight is inversely related to blood pressure in infants,12 young children,13 and adults,3 13 but others have found no association in adolescents.14 15 The limited published data on adults have come predominantly from Britain and include only a small number of women.3 16
The average weight of adults in the United States has been increasing. Approximately one in three persons is defined as obese.17 Obesity is associated with an increased risk of diabetes mellitus18 and coronary artery disease19 and decreased longevity.20 Childhood obesity has been associated with adult obesity, but only two studies have examined the relation between birth weight and adult obesity.8 16
To examine the relation between birth weight and adult hypertension and obesity, we studied >160 000 participants in the Nurses' Health Study I (NHS I) and II (NHS II).
In 1976, 121 700 female registered nurses between 30 and 55 years of age residing in 1 of 11 US states completed and returned the initial questionnaire and constitute the NHS I cohort.21 In 1989, a second group of 116 686 female registered nurses from 15 states, 25 to 42 years of age, completed and returned the initial questionnaire providing information on lifestyle practices and medical history; they constitute the NHS II cohort.22 Both cohorts have been followed by means of biennial mailed questionnaires that inquire about lifestyle practices, other exposures of interest, and the incidence of physician-diagnosed disease. The follow-up rate for each cohort exceeds 90%.
Assessment of Birth Weight
The 1992 NHS I questionnaire included a series of questions regarding birth weight and early life exposures. The categories of birth weight responses (in lb) were <5.0, 5.0 to 5.5, 5.6 to 7.0, 7.1 to 8.5, 8.6 to 10.0, >10.0, and unknown. Information from the NHS II cohort was obtained in 1991, and because of space constraints on the questionnaire, six categories of birth weight responses (in lb) were specified: <5.5, 5.5 to 6.9, 7.0 to 8.4, 8.5 to 9.9, ≥10.0, and unknown.
To assess the validity of the reported birth weights, the actual birth weights of 220 randomly selected NHS II participants were obtained from state birth records.23 The mean values (in lb) for the five birth weight categories calculated with the state birth records were 4.8, 6.3, 7.6, 8.9, and 10.3. In addition, 70.0% of the NHS II participants reported the same birth weight category as was obtained from state birth records. The Spearman correlation between self-reported birth weight and weights recorded on state birth records was .74 (P<.001).
Assessment of Body Mass Index
Body mass index (BMI) was calculated as weight in kilograms divided by the square of height in meters. Height was reported by NHS I subjects in 1976 and by NHS II subjects in 1989. Current weight is asked on each biennial questionnaire, and the most recent weights available were used to calculate BMI. Weight at 18 years of age was reported by NHS I cohort members in 1980 and by NHS II participants in 1989. In a validation study of current weight in NHS I, self-reported weights were highly correlated with actual measurements (r=.96).24 Mean self-reported weights were 3.3 lb lower than actual measurements.24 A validation study of recalled weight at 18 years of age in NHS II compared recalled weight with records from physical examinations conducted at college or nursing school entrance. The correlation between recalled and measured past weight was .87 (P<.001); the correlation for BMI at age 18 was .84 (P<.001).25
Assessment of Blood Pressure and Hypertension
The 1980 NHS I and 1989 NHS II questionnaires inquired about the subjects' blood pressures during the preceding 2 years. The categories of systolic blood pressure responses (in mm Hg) were <105, 105 to 114, 115 to 124, 125 to 134, 135 to 144, 145 to 154, 155 to 164, 165 to 174, 175+, and unknown or not checked within the past 2 years. The categories of diastolic pressure responses (in mm Hg) were <65, 65 to 74, 75 to 84, 85 to 89, 90 to 94, 95 to 104, 105+, and unknown or not checked within the past 2 years. The baseline and biennial follow-up questionnaires asked about physician-diagnosed hypertension. A study participant was considered to have hypertension if it was reported on any questionnaire. The diagnosis of hypertension is reported reliably in the NHS I cohort.26 Self-reported blood pressure and hypertension are strong predictors of coronary heart disease in this group.26
Assessment of Other Factors
Women in both cohorts were asked about family history, including maternal and paternal histories of hypertension and diabetes mellitus. On the most recent questionnaire, we asked whether the participant was born prematurely, defined as >2 weeks before the due date. In addition, subjects were asked if they were part of a multiple birth.
On the 1988 NHS I questionnaire (see the “Appendix”), nine diagrams of different female body figures were drawn ranging from very thin to very obese27 (see the “Appendix”). Cohort members were asked to choose the diagram that best depicted their mothers' figures at 50 years of age.
The characteristics of the cohort were age adjusted according to 5-year age groups by direct standardization to the overall respective cohort. Multiple logistic regression analysis was used to compute the odds ratios of hypertension in the individual birth weight groups, with the middle birth weight group considered the reference category (7.1 to 8.5 lb in NHS I and 7.0 to 8.4 lb in NHS II), while potentially confounding variables were adjusted for simultaneously.28 The reference categories were chosen a priori because these categories contained the largest number of subjects and therefore would provide more stable estimates of the relative risks. The variables considered in these models were birth weight categories (including a category for missing), age (in 5-year categories), adult BMI (10 categories), maternal history of hypertension (yes or no), parental history of hypertension (yes or no), and maternal history of diabetes mellitus (yes or no). Effect modification by prematurity, maternal history of diabetes, or maternal or paternal history of hypertension was investigated by examination of separate logistic regression models within each stratum. For all odds ratios, we calculated 95% CIs.
The magnitude of the error in the measurement of true birth weight using the categories of self-reported birth weight was estimated by use of data from the birth weight validation study in NHS II. The point estimates and CIs from the logistic regression models were corrected for this measurement error. This statistical correction of the point estimates better estimates the association that would have been observed if no measurement error had been present.29 30 Because the birth weight values from the birth certificates were continuous, the categories from the questionnaire were converted to continuous values by use of assigned values of 4.75 for the lowest category, 10.5 for the highest, and midpoints for other categories. After controlling for the other variables included in the multivariate analysis, the partial correlation coefficient for the actual and reported birth weights was .78 (P<.001).
Multiple linear regression analysis was performed to estimate the contribution of individual variables to systolic and diastolic blood pressures. Systolic and diastolic pressures, in separate models, were the dependent variables; the independent variables were birth weight (continuous, by use of the midpoints of the categories), age (continuous), BMI (10 categories), and parental history of hypertension. For all linear regression estimates, we calculated 95% CIs.
Whites were the most common racial group, comprising 96% of the NHS I cohort and 94% of the NHS II population. The number of nonwhite respondents with reported birth weight data was too small to perform a subgroup analysis by race.
Tables 1 and 2⇓⇓ give the characteristics of the women according to birth weight categories for NHS I and NHS II, respectively. In NHS I, 11.0% of the women with reported birth weights were in the two lowest birth weight categories (≤5.5 lb). In NHS II, 8.0% reported a birth weight of <5.5 lb. The smaller proportion of low birth weights in NHS II is consistent with national secular trends.31 The proportion of women with missing or unknown birth weight values was 20.4% in NHS I and 8.0% in NHS II. Overall, the women with missing birth weight data were similar to those in the middle birth weight category group for the characteristics listed (Tables 1 and 2⇓⇓).
The mean BMI was 25.8 kg/m2 in NHS I in 1990 and 24.6 kg/m2 in the younger NHS II cohort in 1991 (Tables 1 and 2⇑⇑). The distribution of adult BMI varied by birth weight category (Figs 1 and 2⇓⇓). To examine the association between birth weight and obesity, we calculated the odds ratio of being in the highest quintile of BMI compared with the lowest quintile using the middle birth weight category as the reference group. Compared with NHS I women in the reference birth weight category (7.1 to 8.5 lb), women with birth weights of 8.6 to 10.0 lb had an age-adjusted odds ratio for being in the highest (>29.2 kg/m2) versus the lowest (<21.9 kg/m2) quintile of adult BMI of 1.19 (95% CI, 1.10 to 1.29); the comparable number for women with birth weights >10.0 lb was 1.62 (95% CI, 1.38 to 1.90). The odds ratios did not change after controlling for the nine categories of mother's figure at 50 years of age. The similar comparisons in NHS II women (>27.9 versus <20.7 kg/m2) yielded age-adjusted odds ratios of 1.30 (95% CI, 1.22 to 1.40) and 1.95 (95% CI, 1.60 to 2.38), respectively. The same pattern, although slightly more pronounced, was observed in both cohorts for BMI at 18 years of age (data not shown).
The odds of being in the highest compared with the lowest quintile of BMI for women in NHS I varied by the reported mother's figure at 50 years of age (Fig 3⇓). For this comparison, the reference group was women with birth weights of 7.1 to 8.5 lb and whose mothers had thin figures at 50 years of age (see the “Appendix,” group A). The odds of being in the highest quintile of BMI increased with increasing maternal obesity. There also appeared to be a U-shaped relation between birth weight and BMI in all groups of maternal figures, with a lower prevalence in the 5.0- to 5.5-lb and 5.6- to 7.0-lb birth weight categories. In addition, among women with mothers with the middle body figures (group B), women in the highest birth weight category compared with those in the reference birth weight category were more likely to be in the highest than in the lowest quintile of adult BMI (odds ratio, 2.12; 95% CI, 1.55 to 2.88). No significant differences in the odds ratios for being in the highest category of adult BMI were seen among women with the leanest or heaviest mothers.
Blood Pressure and Hypertension
The mean systolic and diastolic blood pressures reported on the 1990 NHS I questionnaire were 126.1 and 77.9 mm Hg, respectively. The mean reported systolic and diastolic blood pressures for NHS II in 1991 were 113.7 and 71.5 mm Hg. The mean age-adjusted systolic and diastolic blood pressures varied across birth weight categories in both cohorts (Fig 4⇓).
The prevalence of a history of hypertension was 33.6% in 1992 in NHS I and 6.6% in 1991 in NHS II (Tables 1 and 2⇑⇑), ranging from 2.7% among nurses 25 to 29 years of age to 46.1% among nurses 65 to 69 years of age. The percentage of NHS I subjects having their blood pressures checked between 1986 and 1988 as reported on the 1988 questionnaire was 85.3% and was similar across all birth weight categories (data not shown).
The age-adjusted odds ratio for hypertension was higher in the lowest birth weight category compared with the reference birth weight category (NHS I: odds ratio, 1.39; 95% CI, 1.29 to 1.50; NHS II: odds ratio, 1.43; 95% CI, 1.31 to 1.56; Table 3⇓). Because of the importance of other known risk factors for hypertension, such as parental history of hypertension and BMI, a multivariate analysis was performed to examine the relation between birth weight and hypertension; however, the estimates for the lowest birth weight category changed minimally (NHS I: odds ratio, 1.42; 95% CI, 1.31 to 1.54; NHS II: odds ratio, 1.40; 95% CI, 1.28 to 1.54; Table 3⇓). Although the age-adjusted odds ratio for hypertension also was elevated in the highest birth weight category in NHS II, this was due mostly to the association of high birth weight with higher adult BMI. After adjustment for adult BMI in the multivariate analysis, we found no material increase in the odds ratio for hypertension among the highest birth weight categories (Table 3⇓). Separate logistic regression analyses also were performed according to history of prematurity, maternal and paternal histories of hypertension, and maternal history of diabetes. The observed associations between birth weight and hypertension were not modified by any of these four factors (data not shown).
We corrected the odds ratios and CIs for the effects of measurement error of birth weight in the NHS II cohort. The multivariate odds ratios rose slightly and remained highly significant (Table 3⇑).
Birth weight as a continuous variable also was used in the linear regression models to estimate the magnitude of the effect of birth weight on systolic and diastolic blood pressures. After age, BMI, and parental history of hypertension were controlled for, the 1990 NHS I systolic blood pressure decreased by 0.43±0.04 mm Hg (mean±SD) for each 1-lb increase in birth weight, and diastolic blood pressure decreased by 0.21±0.03 mm Hg. The same analysis in the NHS II cohort revealed that the 1989 systolic blood pressure decreased by 0.09±0.03 mm Hg for each 1-lb increase in birth weight, and diastolic blood pressure decreased by 0.09±0.02 mm Hg. Because the reported blood pressure values included subjects with treated hypertension, we repeated the analysis after assigning a blood pressure of 150/95 mm Hg to all individuals with a history of hypertension. In the NHS I cohort, for each 1-lb increase in birth weight, the systolic and diastolic blood pressures decreased by 0.63±0.05 and 0.41±0.03 mm Hg, respectively. In NHS II, systolic and diastolic pressures were decreased by 0.16±0.03 and 0.13±0.02 mm Hg, respectively, per 1-lb increase in birth weight.
These data support the hypothesis that low birth weight is independently associated with an increased risk of hypertension as an adult. In addition, a high birth weight is associated with a higher adult BMI.
Although birth weight, adult height and weight, and the diagnosis of hypertension were self-reported, errors in reporting alone are unlikely to explain our findings. Errors in the reporting of birth weight probably would be random, and this random misclassification would tend to underestimate the association between birth weight and hypertension. This expected underestimation was confirmed by the results of the odds ratios corrected for error in the measurement of birth weight. Biased recall is unlikely because the association between birth weight and hypertension has only recently been reported. Detection bias also is unlikely because the proportion of participants who had their blood pressures checked in 1988 was similar among the birth weight categories in NHS I, and similar results were seen among those who had their blood pressures measured.
Most of the published literature on the relation between birth weight and blood pressure in adults has been reported by Barker and colleagues13 in Britain. In these field studies, birth weight was obtained from birth records, and recent blood pressure was directly measured by the investigators. The most comprehensive report combined three adult cohorts, including 3240 men and women 36 years of age, 459 persons 46 to 54 years of age, and 1231 individuals 59 to 71 years of age.3 In all three groups, systolic blood pressure was inversely related to birth weight, and the magnitude of the blood pressure increase rose with increasing age. After current BMI was controlled for, systolic blood pressure in the oldest group decreased by 5.2 mm Hg per 1-kg (2.4 mm Hg per 1-lb) increase in birth weight.
The results from our study are consistent with the British data, but the magnitude of the effect of birth weight on systolic blood pressure was much smaller. The accuracy of the nurses' self-reported birth weights, blood pressure values, and history of hypertension was validated previously. Interestingly, the mean systolic blood pressures in our two cohorts were substantially lower than those reported from Britain.3 Among British women 36 years of age, the mean systolic blood pressure was 118 compared with 114 mm Hg for NHS II (mean age, 36.5 years). Among British women 51 to 54 years of age, the mean systolic blood pressure was 149 compared with 126 mm Hg in NHS I (mean age, 56 years). The British study and our study included values from subjects with treated hypertension. It is unclear why the blood pressure values were so much higher in Britain than in the United States, but one reason may be the differences in treatment of elevated blood pressure in the two countries. In addition, the higher blood pressures may have been due in part to differences in BMI. In a recent report from the group in Britain,16 the mean BMI in British women was 27.0 kg/m2 compared with 25.8 kg/m2 in NHS I.
The magnitude of the association between birth weight and hypertension may appear to be inconsistent with the small decrease in the mean systolic blood pressure per 1-lb birth weight. The absolute differences in the age-adjusted frequency of hypertension between the lowest and middle birth weight categories were 7.4% in NHS I and 2.4% in NHS II. Thus, the additional fraction of hypertensive women may have had only a slight impact on the mean systolic blood pressure for the low birth weight category.
The mechanism for the apparent increase in blood pressure associated with low birth weight is unknown. One or more events in utero (eg, changes in fetal blood flow or hormonal variations) may result in an abnormality of the vasculature,32 aberrant autonomic nervous system or endocrine regulation, or disruption of nephrogenesis.33 Once initiated, these events, either alone or in combination with other exposures, may lead to the generation and maintenance of a higher blood pressure.
Obesity in adults is associated with adolescent and childhood obesity, but only limited data are available regarding birth weight and adult BMI.8 16 Fall et al16 reported that BMI rose with increasing birth weight, but this increase was not linear, similar to the results for NHS I and NHS II. Risk factors for higher birth weight include maternal diabetes mellitus, maternal obesity, excessive maternal weight gain, and prolonged gestation.34 After the maternal figures at 50 years of age are stratified, the data suggest that genetic factors and factors associated with birth weight itself are determinants of adult obesity. The likelihood of a woman having a higher adult BMI increased with increasing maternal obesity. The likelihood also increased with higher birth weight just within the middle group of maternal obesity, suggesting modification of the effect of maternal obesity on adult BMI by birth weight in this group of women. The genetic influences on the development of obesity are well known; however, the mechanism for the potential independent association between birth weight and obesity is unknown. The adverse health impact of obesity as early as adolescence was described previously,35 emphasizing the importance of identifying early life risk factors for obesity.
Our findings are most directly generalizable to women ≥25 years of age. However, other studies suggest a similar association in men.3 13 Because infants with low birth weights make up <10% of the general population, low birth weight probably is not a major cause of hypertension in the United States. However, in subgroups with a higher frequency of low-birth-weight infants, it may be a more important cause.31 Similarly, 97% of obese nurses were not high-birth-weight babies; thus, high birth weight is not a major cause of obesity in the United States. Nevertheless, these associations strongly suggest that early life exposures play a role in the development of certain chronic diseases in adults.
The 1988 NHS I questionnaire included the following question: Which diagram best depicts the approximate outline of your mother at 50 years of age (see Fig 5⇓)? For the analysis, diagrams 1 through 3 were combined and are referred to as group A; diagrams 4 through 6 are group B; and diagrams 7 through 9 are group C.
This work was supported by research grants (DK-45362, DK-46200, CA-40356, and CA-50385) from the NIH. We are indebted to the study participants for their continuing cooperation; to Dr Sharon Curhan for helpful comments; and to Gary Chase, Karen Corsano, Barbara Egan, Lori Ward, and Stefanie Bechtel for their ongoing support.
- Received December 5, 1995.
- Revision received March 17, 1996.
- Accepted March 26, 1996.
- Copyright © 1996 by American Heart Association
Osmond C, Barker DJP, Winter PD, Fall CHD, Simmonds SJ. Early growth and death from cardiovascular disease in women. BMJ. 1993;307:1519-1524.
Hales CN, Barker DJP, Clark PMS, Cox LJ, Fall C, Osmond C, Winter PD. Fetal and infant growth and impaired glucose tolerance at age 64. BMJ. 1991;303:1019-1022.
Law CM, de Swiet M, Osmond C, Fayers PM, Barker DJP, Cruddas AM, Fall CHD. Initiation of hypertension in utero and its amplification throughout life. BMJ. 1993;306:24-27.
Barker DJP, Godfrey KM, Fall C, Osmond C, Winter PD, Shaheen SO. Relation of birth weight and childhood respiratory infection to adult lung function and death from chronic obstructive airways disease. BMJ. 1991;303:671-675.
Taitz LS. Infantile overnutrition among artificially fed infants in the Sheffield region. Br Med J. 1971;1:315-316.
Shukla A, Forsyth HA, Anderson CM, Marwah SM. Infantile overnutrition in the first year of life: a field study in Dudley, Worcestershire. Br Med J. 1972;4:507-515.
Prokopec M, Bellesle F. Adiposity in Czech children followed from 1 month of age to adulthood: analysis of individual BMI patterns. Ann Hum Biology. 1993;20:517-525.
Kaplan NM. Hypertension in the population at large. Clinical Hypertension. 5th ed. Baltimore, Md: Williams & Wilkins, 1990;chap 1:1-25.
Kaplan NM. Primary hypertension: pathogenesis. Clinical Hypertension. 5th ed. Baltimore, Md: Williams & Wilkins, 1990;chap 3:54-111.
Launer LJ, Hofman A, Grobbee DE. Relation between birth weight and blood pressure: longitudinal study of infants and children. BMJ. 1993;307:1451-1454.
Barker DJP, Osmond C, Golding J, Kuh D, Wadsworth MEJ. Growth in utero, blood pressure in childhood and adult life, and mortality from cardiovascular disease. BMJ. 1989;298:564-567.
Matthes JWA, Lewis PA, Davies DP, Bethel JA. Relation between birth weight at term and systolic blood pressure in adolescence. BMJ. 1994;308:1074-1077.
Seidman DS, Laor A, Gale R, Stevenson DK, Mashiach S, Danon YL. Birth weight, current body, weight and blood pressure in late adolescence. BMJ. 1991;302:1235-1237.
Fall CHD, Osmond C, Barker DJP, Clark PMS, Hales CN, Stirling Y, Meade TW. Fetal and infant growth and cardiovascular risk factors in women. BMJ. 1995;310:428-432.
Mann GV. The influence of obesity on health. N Engl J Med. 1974;291:178-185, 226-232.
Colditz GA. The Nurses' Health Study: a cohort of US women followed since 1976. J Am Med Wom Assoc. 1995;50:40-44.
Troy LM, Michels KB, Hunter DJ, Spiegelman D, Manson JE, Colditz GA, Stampfer MJ, Willett WC. Self-reported birthweight and history of having been breast-fed among younger women: an assessment of validity. Int J Epidemiol. 1996;25:122-127.
Willett W, Stampfer MJ, Bain C, Lipnick R, Speizer FE, Rosner B, Cramer D, Hennekens CH. Cigarette smoking, relative weight and menopause. Am J Epidemiol. 1983;117:651-658.
Colditz GA, Martin P, Stampfer MJ, Willett WC, Sampson L, Rosner B, Hennekens CH, Speizer FE. Validation of questionnaire information on risk factors and disease outcomes in a prospective cohort study of women. Am J Epidemiol. 1986;123:894-900.
Stunkard AJ, Sorensen T, Schulsinger F.Use of the Danish adoption register for the study of obesity and thinness. In: Kety SS, Rowland LP, Sidman RL, Matthysse SW, eds. Genetics of Neurological and Psychiatric Disorders. New York, NY: Raven Press Publishers; 1983:115-120.
Kleinbaum DG, Kupper LL, Muller KE. Applied Regression Analysis and Other Multivariable Methods. Boston, Mass: PWS-KENT Publishing; 1988.
Willett WC. Nutritional Epidemiology. New York, NY: Oxford University Press; 1990.
Rosner B, Spiegelman D, Willett WC. Correction of logistic regression relative risk estimates and confidence intervals for measurement error: the case of multiple covariates measured with error. Am J Epidemiol. 1990;132:734-745.
Barker DJP, Bull AR, Osmond C, Simmonds SJ. Fetal and placental size and risk of hypertension in adult life. BMJ. 1990;301:259-262.
Sokol RJ, Brindley BA, Dombrowski MP. Practical diagnosis and management of abnormal labor. In: Scott JR, DiSaia PJ, Hammond CB, Spellacy WN, eds. Danforth's Obstetrics and Gynecology. 7th ed. Philadelphia, Pa: JB Lippincott Co; 1994:521-561.