Maternal Size at Birth and the Development of Hypertension During Pregnancy: A Test of the Barker Hypothesis

Background  Whether individuals who were small at birth are at increased risk of developing cardiovascular disease (the Barker hypothesis) is a topic of great controversy. Although an increased risk has been suggested by several reports, the reports have been criticized for being based on ill-defined populations, for the large numbers of subjects who were unavailable for follow-up, and for inadequate control of socioeconomic status.

Objective  To determine whether a woman's weight and gestational age at birth predict the development of hypertension during her subsequent pregnancies.

Design  Prospective observational study.

Subjects  Women born in Copenhagen, Denmark, as subjects in the Danish Perinatal Study (1959-1961) were traced through the Danish Population Register. Information was obtained on their pregnancies from 1974 to 1989.

Main Outcome Measures  Onset of hypertension in pregnancy, defined by the presence of a systolic blood pressure of 140 mm Hg or greater or a diastolic blood pressure of 90 mm Hg or greater on 2 visits at or after 140 days' gestation.

Results  Hypertension developed in 11.3% of the pregnant women who were small for gestational age at birth, compared with 7.2% of the pregnant women who were not small for gestational age at birth (odds ratio [OR], 1.7; 95% confidence interval [CI], 1.1-2.6), and in 9.4% of the pregnancies in women who were preterm at birth, compared with 7.6% of pregnancies in women who were not preterm at birth (OR, 1.3; 95% CI, 0.8-2.0). After adjustment for adult body mass index, smoking, birth order, and hypertension in the subjects' own mothers, the ORs for small-for-gestational-age women and preterm women to develop hypertension during pregnancy were 1.8 (95% CI, 1.1-2.8) and 1.5 (95% CI, 0.96-2.5), respectively.

Conclusion  These results support the Barker hypothesis, while addressing many of the methodological criticisms of previous investigations.

THE DEGREE to which chronic diseases during adulthood originate before birth is under active investigation. A large number of studies, primarily by members of the Environmental Epidemiology Unit of the University of Southampton, Southampton, England, have reported that both men and women who were small at birth are at increased risk of developing a variety of chronic diseases during adulthood. In particular, small size at birth has been associated with an increased risk of cardiovascular disease, dyslipidemia, hypertension, and type 2 diabetes mellitus during adulthood.^1-3 Similar findings have been noted by other investigators as well,^4,5 and a detailed review was published in 1996.⁶ These studies have been criticized on methodological grounds, primarily for being based on ill-defined starting populations, with large numbers of subjects unavailable for follow-up, and for inadequate control of socioeconomic status.⁶ The present report describes the association between size at birth and hypertension during pregnancy, while addressing many of these criticisms.

The data are from a study of the association of birth outcomes in consecutive generations in Denmark, methods of which have been reported.^7,8 The women were originally members of the Danish Perinatal Cohort Study, which included all deliveries occurring at the State University Hospital (Rigshospitalet) in Copenhagen from September 1959 to December 1961. The investigators used a tightly controlled series of data collection procedures, which have been described previously.⁹

In May 1987 and again in May 1989, all women in this cohort were searched for in the Danish Population Register to determine who had given birth. The Population Register was established in 1924 and since 1969 has included the names and personal identification numbers of the parents and children of each Danish resident. Maintaining a current address in the Register is required to receive the many social benefits accorded Danish residents, and the Population Register includes essentially all births in the country. From among the women in the Perinatal Cohort who had given birth, we selected 158 who were themselves preterm at birth, 160 who were small for gestational age (SGA, <10 percentile for Danish females) at term, 37 who were both preterm and SGA, and 943 who were born at term and had grown appropriately. These selections were made to study the intergenerational correlations of fetal growth and duration of pregnancy, as has been described previously.⁷ The study was approved by the institutional review board of the University of Copenhagen.

We attempted to interview all selected women and to abstract the medical records of all their live births and of their stillbirths of at least 28 weeks' gestation; 1097 of the 1298 selected women were interviewed. Record abstraction was carried out independently of the interviews, and the records of all women were abstracted regardless of whether they were interviewed. Information on confounding factors was obtained from medical records. If this information was missing from the record, it was taken from the interview questionnaire. Information on hypertensive disorders of pregnancy was obtained only from medical records.

Women with a history of hypertension antedating the pregnancy or who had 2 systolic blood pressure values of 140 mm Hg or greater and/or 2 diastolic blood pressure values of 90 mm Hg or greater measured at less than 140 days' gestation were deemed to have chronic hypertension and were excluded. The de novo onset of hypertension in pregnancy (hereafter referred to simply as hypertension) was diagnosed when a systolic blood pressure of 140 mm Hg or greater and/or a diastolic blood pressure of 90 mm Hg or greater was noted on 2 separate visits at or after 140 days' gestation. Proteinuria was diagnosed when there were 2 or more urine protein dipsticks of at least 2+ (100 mg/dL) at or after 140 days' gestation or a 24-hour urine sample containing more than 0.3 g of protein. Preeclampsia required the diagnosis of both hypertension and proteinuria. However, proteinuria was inconsistently recorded in the medical charts, and only 25 pregnancies (1.2%) were complicated by preeclampsia. Therefore, the analysis was restricted to hypertension regardless of proteinuria. Intrapartum blood pressure measurements were not abstracted.

Neither obstetrical records nor interviews could be obtained for 10 women. The 1288 remaining women had 2103 births, 27 of which had missing records. After excluding 30 twin births, 31 births in women with chronic hypertension, and 7 pregnancies with insufficient data in the record to determine whether there had been a hypertensive disorder, there were 2008 singleton pregnancies among 1261 women from 1974 to 1989.

The primary study outcome was the presence of hypertensive disorders during the subjects' pregnancies. The women may have had more than 1 pregnancy, causing the study outcomes to be nonindependent. A commercially available set of statistical programs¹⁰ designed for the analysis of data where the observations are not independent of each other (eg, cluster surveys or multiple births in 1 woman) was used to correct the P values and confidence limits for the correlation of hypertension in successive pregnancies in the same woman. All 95% confidence intervals for comparisons involving the women's pregnancies were derived from the SEs of linear (for continuous outcomes) or logistic (for dichotomous outcomes) regression procedures. Analyses of the mothers' own characteristics at birth used χ² and Student t tests for categorical and continuous variables, respectively.

Hypertension complicated 157 (7.8%) of the 2008 pregnancies. Maternal characteristics associated with the occurrence of hypertension are presented in Table 1. Hypertension was more common among first births and among women who did not smoke during the third trimester. The mean prepregnant body mass index (BMI or Quetelet index: calculated as the weight in kilograms divided by the square of the height in meters: weight (kg)/[height (m)]²) of women who developed hypertension (22.8 kg/m²) was significantly greater than that of women who were normotensive (21.1 kg/m², P<.001). The increase was attributable to an increased risk of hypertension only among the most obese women (Table 1). An almost identical pattern was observed for prepregnant weight. Hypertension was not associated with the woman's height, education, whether she worked, or the occupational status of those who did work. Women who developed hypertension were younger when the pregnancy began (22.3 years) than were normotensive women (22.9 years, P=.01), but the difference was no longer statistically significant after adjustment for parity (P=.39).

Hypertension occurred in 9.4% of the pregnant women who were themselves preterm at birth, compared with 7.6% of the pregnant women born at term (odds ratio [OR], 1.3; 95% confidence interval [CI], 0.8-2.0). Hypertension complicated 11.3% of the pregnancies in women who were themselves SGA at birth and 7.2% of the pregnancies in those who were not SGA at birth (OR, 1.7; 95% CI, 1.1-2.6). Women who weighed less than the fifth percentile for gestational age had a similar risk of developing hypertension (11.1%) as women who weighed from the fifth to the tenth percentile (11.5%).

As judged by the characteristics in Table 1, women who were SGA at birth would be expected to be at lower risk than non-SGA women of developing hypertension during pregnancy.Women who were SGA at birth were also more likely to smoke (64% vs 55%, P=.02). They were of lower mean prepregnant BMI (20.8 kg/m² vs 21.3 kg/m², P=.05), less likely to be in the top quartile of BMI (20% vs 25.4%), and more likely to be in the bottom quartile of BMI (30% vs 25%; P for trend, .04). Compared with non-SGA women, SGA women were equally represented among first and later births. Preterm women did not differ on any of these characteristics from women who were not preterm. There was no evidence that the elevation of hypertension risk among SGA women was greater among the subset who were obese as adults (OR for SGA among women in the top quartile of adult BMI, 1.4; P=.46), nor was maternal ponderal index at birth associated with the later development of hypertension.

An important distinction is whether being small at birth is per se a risk factor for hypertension or whether it is a marker for a familial tendency both to be small at birth and to develop hypertension during pregnancy. To address this question, the 1959-1961 records of the Danish Perinatal Study were accessed. Although no formal definitions are available in that study, hypertensive disorders were classified as hypertension without preeclampsia (16.3% of births), preeclampsia (3.4%), eclampsia (0), and hypertension in the immediate postpartum period (10.7%). Approximately 50% of women with the latter diagnosis also had elevated blood pressure levels before delivery; 22.8% of the subjects' mothers had 1 or more of these conditions during pregnancy from 1959-1961. The relatively high frequency of hypertension reflects the Rigshospitalet's status as the major referral center in eastern Denmark.

The characteristics associated with hypertensive disorders in the subjects' mothers' pregnancies during 1959 to 1961 were similar to those observed in the subjects' own pregnancies (Table 2). Antepartum hypertension was slightly but not significantly less common in first births (15% vs 17%, P=.30). However, preeclampsia (5% vs 2%, P=.003) and postpartum hypertension (13% vs 8%, P=.009) were more common in first births. Hypertensive diagnoses of any kind were slightly but not significantly more common in first births (24% vs 22%, P=0.40). Women who smoked during the third trimester of their pregnancies during the 1959 to 1961 period were at reduced risk of developing antepartum hypertension (13% vs 20%, P<.001), preeclampsia (2% vs 5%, P=.03), postpartum hypertension (8% vs 13%, P=.004), and any hypertensive diagnosis (17% vs 29%, P<.001). Increased BMI was significantly associated with antepartum hypertension (22.5 kg/m² vs 21.5 kg/m², P=.002) and with any hypertensive disorder (22.4 kg/m² vs 21.5 kg/m², P<.001). Similar but not statistically significant trends were observed for preeclampsia (22.3 kg/m² vs 21.7 kg/m², P=.40) and postpartum hypertension (22.1 kg/m² vs 21.7 kg/m², P=0.30). Prepregnant weight exhibited a similar pattern.

Socioeconomic status was available for 1116 (88%) of the study women when they were 1 year of age in 1960-1962; 9% were of high (Registrar General's class I), 49% were of middle (class II-III), and 43% were of lower (class IV-V) social class. There was no trend for hypertension in 1959-1961 to be associated with lower social class or with missing social class (Table 2). All 4 hypertensive diagnoses were more common in women whose mothers were of middle class compared with both upper and lower social classes. This association was statistically significant for every diagnosis except preeclampsia (antepartum hypertension, P=.04; postpartum hypertension, P=.03; preeclampsia, P=.11; and any hypertensive diagnosis, P=.005).

The association between various characteristics of the subjects' mothers' 1959-1961 pregnancies and hypertension in the subjects' own pregnancies is presented in Table 3. All the hypertensive diagnoses in 1959-1961 elevated the woman's subsequent risk of developing hypertension during pregnancy in the 1974 to 1989 period to a comparable degree, with ORs ranging from 1.3 to 1.8; the associations with postpartum hypertension (P=.02) and any hypertensive diagnosis (P=.02) were statistically significant. Neither the prepregnant BMI nor the weight of the subjects' mothers in 1959-1961 was associated with hypertension in the subjects' pregnancies in 1974-1989. Compared with normotensive women, women who experienced hypertension were slightly more likely to be firstborn (8.8% vs 6.6%, P=.10). The women's social class at 1 year of age was not associated with their subsequent development of hypertension in 1974-1989.

After hypertensive disorders of pregnancy in each subject's mother, the subject's own birth order, and the subject's third-trimester smoking, BMI, and parity were controlled for, women who were SGA at birth had an OR of 1.8 (95% CI, 1.1-2.8) for the development of hypertension (Table 4). Women who were preterm at birth had an OR of 1.5 (95% CI, 0.96-2.5) for the development of hypertension. Further adjustment for maternal socioeconomic status at the age of 1 year and at the time of pregnancy, either singly or together, changed these ORs by less than 5% and did not influence the statistical significance levels. Substitution of prepregnant weight for BMI resulted in adjusted ORs of 2.0 (95% CI, 1.2-3.3) for women who were SGA and 1.5 (95% CI, 0.94-2.5) for women who were preterm at birth for the development of hypertension during pregnancy.

Women with presumed chronic hypertension were excluded from this study. This exclusion affected 2.8% of the pregnancies in SGA women vs 1.3% of the pregnancies in non-SGA women (P=.10) and 2.2% of the pregnancies in preterm women vs 1.4% of the pregnancies in nonpreterm women (P=.30). Therefore, had women with presumed chronic hypertension been included in this study, the results would have been largely unchanged.

Women who were SGA at birth were found to be at increased risk of developing hypertension during pregnancy, and the risk was not explained by the presence of hypertensive disorders during their mothers' pregnancies. Women who were preterm but not SGA at birth were also at slightly increased risk of developing hypertension, although the increase was not statistically significant (P=.07). Compared with fetuses in normal pregnancies, fetuses that are born preterm are on average smaller at each gestational age,¹¹ which may account for the modest increase in hypertension observed in women who were preterm at birth. While the association between size at birth and the development of hypertension during pregnancy is of interest, these results may have implications beyond pregnancy, since hypertension arising during pregnancy is a marker for increased risk of developing chronic hypertension years later.^12,13

These results support the Barker hypothesis that reduced fetal growth is associated with an increased risk of developing a variety of risk factors for cardiovascular disease, including hypertension, during adulthood.¹⁴ The present study addresses many of the shortcomings of previous studies: a defined population was selected; a high rate of follow-up was achieved; several measures of socioeconomic status, both during childhood and adulthood, were considered; and specific criteria were used to diagnose hypertension from detailed medical records. Although hypertension during pregnancy was the primary outcome, there was evidence that SGA women were at increased risk for chronic hypertension as well.

Several shortcomings of this study should be noted. We did not collect data to diagnose hypertension arising during the intrapartum period, the time that a substantial fraction of hypertensive disorders of pregnancy are first noted.¹⁵ That hypertension was less common in the present study compared with 2 recent cohorts from the United States^16,17 might be attributable to this difference in surveillance, as well as to a lower BMI and a higher prevalence of smoking in the Danish cohort compared with the American cohorts. As long as the relative distribution of time at onset of hypertension is similar between women who are SGA at birth and other women, underascertainment should not influence the results. We are unaware of any data that address this issue.

Although explicit criteria were used to diagnose hypertension in the 1974-1989 pregnancies, the criteria used to diagnose hypertension in the 1959-1961 pregnancies were not stated. However, the increased occurrence of hypertension among women with a higher BMI and the reduced occurrence among women who smoked were virtually identical in both generations. They were also very similar to results observed in the 2 American cohorts.^16,17 This suggests that the criteria used to diagnose hypertensive disorders of pregnancy were not likely to have differed substantially in 1959-1961 than in more recent times.

The socioeconomic status of the study population needs to be considered for its effects both on the composition of the cohort and as a potential confounding factor. As was noted previously,⁷ the Rigshospitalet was the largest academic medical center in Denmark, and under the centralized hospital system, women from eastern Denmark who were experiencing complications of pregnancy or who were expected to have a difficult delivery were often referred there for care. Young, unwed mothers also were sent routinely to the Rigshospitalet for delivery. Therefore, the Danish Perinatal Cohort contained a larger fraction of low-birth-weight offspring and was of lower socioeconomic status than the Danish population in general. While it is unlikely that these referral practices would select women (ie, the grandmothers in this study) who were simultaneously at risk of giving birth to an infant that was both SGA and at future risk of hypertension during her own pregnancy, this limitation should be noted.

The information we collected on socioeconomic status was limited to the Registrar General's classification of the women's parents when the women were 1 year old and a somewhat more modern classification of the women's occupations at the time of their pregnancies. As was noted in a recent review of the "Barker hypothesis,"⁶ additional measures of material wealth, such as income, car ownership, or home ownership, are associated with reductions in coronary artery disease mortality even after more traditional measures of social class are controlled for. While none of our measures of socioeconomic status was associated with the risk of hypertension during the index pregnancies, it is possible that unmeasured socioeconomic factors, and not maternal size at birth per se, account for our findings.

Our definition of preexisting hypertension (ie, history of hypertension antedating the pregnancy or 2 systolic blood pressure values of 140 mm Hg or greater and/or diastolic blood pressure values of 90 mm Hg or greater before 20 weeks' gestation) may have been excessively restrictive given that among both hypertensive and normotensive women, blood pressure levels normally decrease in the first half of pregnancy. Indeed, 2 recent clinical trials of preeclampsia prevention attempted to exclude chronically hypertensive women by requiring that gestational blood pressure values be less than 135/85 mm Hg.^16,18 When we lowered the definition of preexisting hypertension to 2 blood pressure values of 130/80 mm Hg or greater, the prevalence of presumed chronic hypertension increased from 1.5% to 10.5%. Nevertheless, the results were largely unchanged. The adjusted OR for SGA women to develop hypertension remained at 1.8 (95% CI, 1.1-3.0), but the adjusted OR for preterm women to develop hypertension increased to 1.7 (95% CI, 1.1-2.9) and became statistically significant.

There are several possible biological mechanisms for these findings. Brenner et al¹⁹ have hypothesized that a congenitally reduced number of nephrons, with the resultant hyperfiltration nephropathy, is an underlying defect in essential hypertension. Growth-restricted human fetuses have been shown to have a reduced number of nephrons compared with larger fetuses.²⁰ Women who were small at birth may have a reduced number of nephrons and are therefore at risk of becoming hypertensive during pregnancy as a result of the physiological increase in glomerular filtration and resultant hyperfiltration nephropathy.

While some cases of de novo hypertension during pregnancy may represent the unmasking of latent chronic hypertension, the increased risk of the condition with first births,²¹ as well as with changes in paternity,²² suggests an immunologic mechanism. The genetics of a putative immunologic mechanism for hypertension during pregnancy were reviewed recently by Zhang et al,²¹ who noted that at this time data are insufficient to draw conclusions. Hereditary hypercoagulable states, such as occur with the factor V Leiden mutation, have been reported to predispose to both reduced fetal growth and preeclampsia.²³

An additional biological mechanism might be insulin resistance. The chronic diseases of adulthood that have been associated with reduced size at birth—hypertension, dyslipidemia, and glucose intolerance—are attributable in part to insulin resistance.²⁴ Since insulin is the primary growth-promoting hormone in the fetus,²⁵ an individual who is relatively insulin resistant will be SGA at birth and will be at increased risk of developing hypertension, dyslipidemia, and glucose intolerance during adulthood. Infants who were SGA at birth have been shown to have impaired sensitivity to insulin when evaluated during childhood.²⁶ If the Barker hypothesis is not attributable to artifacts of study design and is attributable to insulin resistance, then it is important to determine whether insulin resistance is the primary defect or whether it arises as part of a fetal adaptive response to intrauterine malnutrition.²⁷ The former may not be preventable, but the latter might be ameliorated by improved maternal nutrition before and during pregnancy and perhaps by prevention of pregnancy complications associated with reduced fetal growth. The present results support the Barker hypothesis. Additional research should address the mechanism of the Barker hypothesis. If insulin resistance plays a role, research should distinguish between primary and adaptive fetal insulin resistance and their respective roles in the pathogenesis of chronic disease during adulthood.

Reprints: Mark A. Klebanoff, MD, MPH, Division of Epidemiology, Statistics, and Prevention Research, National Institute of Child Health and Human Development, National Institutes of Health, 6100 Bldg, Room 7B03, Bethesda, MD 20892-7510.

Accepted for publication December 1, 1998.

This project was supported in part by contract NO1-HD-7-2902 from the National Institutes of Health, Bethesda, Md.