J. Perinat. Med. 2015; 43(6): 695–701
Tamika Royal-Thomas*, Daniel McGee, Debajyoti Sinha, Clive Osmond and Terrence Forrester
Association of maternal blood pressure in pregnancy with blood pressure of their offspring through adolescence Abstract: This article looks at the association of maternal blood pressure with the blood pressure of the offspring from birth to childhood. The Barker hypothesis states that maternal and “in utero” attributes during pregnancy affect a child’s cardiovascular health throughout life. We present an analysis of a unique dataset that consists of three distinct developmental processes: maternal cardiovascular health during pregnancy; fetal development; and child’s cardiovascular health from birth to 14 years. This study explored whether a mother’s blood pressure reading in pregnancy predicts fetal development and determines if this in turn is related to the future cardiovascular health of the child. This article uses data that have been collected prospectively from a Jamaican cohort which involves the following three developmental processes: (1) maternal cardiovascular health during pregnancy which is the blood pressure and anthropometric measurements at seven time-points on the mother during pregnancy; (2) fetal development which consists of ultrasound measurements of the fetus taken at six time-points during pregnancy; and (3) child’s cardiovascular health which consists of the child’s blood pressure measurements at 24 time-points from birth to 14 years. The inter-relationship of these three processes was examined using linear mixed effects models. Our analyses indicated that attributes later in childhood development, such as child’s weight, child’s baseline systolic blood pressure (SBP), age and sex, predict the future cardiovascular health of children. The results also indicated that maternal attributes in pregnancy, such as mother’s baseline SBP and SBP change, predicted significantly child’s SBP over time.
*Corresponding author: Dr. Tamika Royal-Thomas, Tropical Medicine Research Institute, University of the West Indies, Mona Campus, Kingston 7, Jamaica, Phone: +876-927-1884, E-mail: [email protected] Daniel McGee and Debajyoti Sinha: Statistics Department, Florida State University, Tallahassee, FL, USA Clive Osmond: MRC Lifecourse Epidemiology Unit, University of Southampton, Southampton General Hospital, Southampton, UK Terrence Forrester: UWI Solutions for Developing Countries, University of the West Indies, Mona Campus, Kingston 7, Jamaica
Keywords: Adolescence; blood pressure; cardiovascular disease (CVD); coronary heart disease (CHD); fetal development; in utero; infant; maternal health; pregnancy; systolic. DOI 10.1515/jpm-2014-0038 Received February 4, 2014. Accepted July 22, 2014. Previously published online August 28, 2014.
Introduction The purpose of this study is to examine the association of maternal blood pressure with the blood pressure of the offspring from birth to childhood. The Barker hypothesis states that adverse influences early in development, especially during intrauterine life, can result in permanent changes in physiology and metabolism, which lead to increased risk in adulthood [2, 3, 5, 6, 8]. Barker came up with this theory when he noticed that regions in England that had the highest rates of infant mortality in the early 1900s also had the highest rates of mortality from coronary heart disease (CHD) decades later [1, 5, 15]. As infant mortality was related to low birth weight, these observations led to the idea that infants who had low birth weight and survived infancy and childhood may be at increased risk for CHD [5]. The Barker hypothesis asserts that human fetuses have to adapt to a limited supply of nutrients and in doing so they can permanently change their structure and metabolism [1, 3, 8, 15]. Osmond and Barker [15] indicated that these fetal changes may be the origin of a number of diseases later in life which includes CHD, hypertension, and non-insulin dependent diabetes. Barker et al. looked at the relationship between small head circumference and thinness at birth to death from CHD in adult life [3]. Thame et al. demonstrated that blood pressure is associated with placental volume and birth weight [18]. Other studies have demonstrated factors that are associated with cardiovascular disease (CVD) or onset of hypertension later in life such as social class [10, 16] and low birth weight accompanied by accelerated childhood growth [12, 20]. There has been extensive work done on the hypothesis that adult CVD is determined by what happens “in utero”
Brought to you by | University of California - San Diego Authenticated Download Date | 2/20/16 11:55 PM
696 Royal-Thomas et al., Maternal blood pressure in pregnancy and offspring blood pressure through adolescence and during the early days of an infant’s childhood. Several studies have described the relationship between birth weight and blood pressure [2, 3, 6, 13, 15, 18], but very few studies include longitudinal data. Furthermore, the methods of data analysis were mainly descriptive or did not fully utilize the available data. This article has utilized data that have been collected prospectively from a Jamaican cohort involving three longitudinal processes. The Jamaican dataset is called the vulnerable windows cohort (VWC) and is unique in that it has the potential to address whether and how much the prebirth measurements influence future cardiovascular risk. A more complete description of the data and the longitudinal processes is presented in the Methods section. The VWC is a complex dataset which gives a unique opportunity to examine epidemiological associations on a more detailed level. The main aims of the study were to examine if a child’s fetal development predicts his cardiovascular health later in life, and does the maternal cardiovascular health during pregnancy predict the child’s fetal development, and in turn, do both processes predict the child’s cardiovascular health throughout life. The results from the study indicated that attributes such as mother’s baseline systolic blood pressure (SBP) and SBP change in pregnancy have a positive association with child’s SBP over time while controlling child’s weight, age, and sex.
Methods Data Some 712 women attending the antenatal clinic of the University Hospital of the West Indies, Jamaica, were invited to participate in a prospective study called the VWC. The recruitment period lasted from November 1992 to July 1994. The women were included if they were between 15 and 40 years of age, were sure of the date of their last menstrual period (as confirmed by a 14-week ultrasound), had singleton pregnancies, and did not have systemic illnesses (e.g., pregnancy-induced hypertension, diabetes mellitus, etc.) or genetic abnormalities. After excluding 82 pregnancy losses/stillbirths, five twin gestations, and 56 women who refused to participate in the study, 569 women and their offspring were enrolled. The mothers’ ages, weights, and heights were recorded together with their parity on their first visit to the antenatal clinic. All women were offered transportation to and from the hospital to enhance participation in the study. Their socioeconomic status was scored using the variables of household amenities, crowding in the home, parents’ educational level, and parents’ occupation [4, 17]. Maternal hemodynamic, anthropometric, biochemical, and hematological statuses were assessed throughout the pregnancy. At each visit, maternal blood pressure was measured with an oscillometric sphygmomanometer (Dinamap TM monitor Model 8100,
Critikon Inc., Tampa, FL, USA). Three individuals, Dr. Minerva Thame (MT), a nurse and a medical technologist, made all measurements. Two of the three observers made the ultrasound measurements (MT and the technologist). All three were trained to apply the questionnaires and make the measurements. At the start, and at 3 monthly intervals for the duration of the study, inter- and intraobserver measurement variabilities were assessed, and training and recertification prescribed for any observer whose scores were not acceptable. The children were measured at birth, at 6 weeks, every 3 months–2 years, and then every 6 months thereafter at a special follow-up clinic. At each clinic attendance, blood pressure was measured with an oscillometric device (Dinamap TM Monitor Model 8100, Critikon Inc., Tampa, FL, USA) and manually. Two measurements were taken using the Dinamap which were averaged and two manual readings were taken which were also averaged. The averaged Dinamap readings were used in the statistical analyses [17]. Fetal and placental growth were assessed by maternal abdominal ultrasound and birth anthropometric measurements were made using standardized techniques. The method used to measure placental volume required that the entire placenta be seen on the screen. After 20 weeks’ gestation, many placentas are too large for this, so placental volume was measured at only the first three visits. The average of three repeats was used for each measurement. Placental volume was measured by identifying and recording on videotape, the long axis of the placenta. A continuous recording of the image of the placenta orthogonal to the axis was made by sweeping the probe along the axis at constant velocity. This axis was divided into six sections of equal length; the five interior cross-sectional areas were measured and integrated to estimate the placental volume [17]. At birth, detailed anthropometric measurements of the newborn infant were obtained using standardized methods and included weight, crown-heel length, chest and head circumferences, and placental weight. Gestational age was estimated from the date of the last menstrual period. The offspring were measured at birth, at 6 weeks, every 3 months–2 years, and then every 6 months thereafter at a special follow-up clinic. As described by Thame et al. [18], at each clinic attendance blood pressure was measured with an oscillometric device (Dinamap TM Monitor, Model 8100, Critikon Inc. Tampa, FL, USA) and manually. Birth weight was measured with a Health o Meter® 457 scale (Pelstar LLC, Bridgeview, IL, USA); height with a length board (Holtain Ltd., Crymych, Dyfed, UK) for children ≤ 2 years and thereafter with a stadiometer (Holtain Ltd., Crymych, Dyfed, UK). Four skinfold thicknesses were measured [4]. Mothers and children were seen twice a year since then for anthropometric, blood pressure, and pulse rate measurements. The present data analysis is confined to children who were seen at all scheduled visits between birth and age 14 years, which started out with 487 children and decreased over time. A detailed description of the measurements is provided by Thame et al. [17, 18]. The VWC data consist of three developmental processes. The first developmental process is the maternal cardiovascular health during pregnancy data which describes the mother’s SBP, diastolic blood pressure, pulse rate, and anthropometric measurements such as height, weight, skinfold measurements among other variables at seven time-points (8–10, 14, 17, 20, 25, 30, 35 weeks) during pregnancy. The second developmental process is the fetal development data that have been collected on the infants while they were in the womb. Six measurements of femoral length, abdominal circumference, biparietal diameter, and head circumference were estimated at 14, 17, 20, 25, 30 and 35 weeks by ultrasound, and placental volume
Brought to you by | University of California - San Diego Authenticated Download Date | 2/20/16 11:55 PM
Royal-Thomas et al., Maternal blood pressure in pregnancy and offspring blood pressure through adolescence 697
was estimated at three time-points (14, 17, and 20 weeks) by ultrasound (see Table 1). The third developmental process consists of the child’s cardiovascular health data taken from birth to current age (see Table 2 for the descriptive summary of birth data). This process consisted of approximately 34 measurements of anthropometric data such as height, weight, head circumference, waist circumference, hip circumference, skinfold measurements, and other anthropometric variables and 24 measurements of SBP, diastolic blood pressure,
and pulse rate measurements which were taken from 2 years to current age.
Statistical analyses There are various traditional statistical methods that have been used for the analysis of longitudinal data, i.e., data collected over
Table 1 In utero descriptive measurements during pregnancy. Variable
n
Mean
Standard deviation
Minimum
Maximum
Placental volume at 14 weeks (mL) Placental volume at 17 weeks (mL) Placental volume at 20 weeks (mL) Biparietal diameter at 14 weeks (mm) Biparietal diameter at 17 weeks (mm) Biparietal diameter at 20 weeks (mm) Biparietal diameter at 25 weeks (mm) Biparietal diameter at 30 weeks (mm) Biparietal diameter at 35 weeks (mm) Head circumference at 14 weeks (mm) Head circumference at 17 weeks (mm) Head circumference at 20 weeks (mm) Head circumference at 25 weeks (mm) Head circumference at 30 weeks (mm) Head circumference at 35 weeks (mm) Femoral length at 14 weeks (mm) Femoral length at 17 weeks (mm) Femoral length at 20 weeks (mm) Femoral length at 25 weeks (mm) Femoral length at 30 weeks (mm) Femoral length at 35 weeks (mm) Abdominal circumference at 14 weeks (mm) Abdominal circumference at 17 weeks (mm) Abdominal circumference at 20 weeks (mm) Abdominal circumference at 25 weeks (mm) Abdominal circumference at 30 weeks (mm) Abdominal circumference at 35 weeks (mm)
408 451 435 417 456 445 412 412 402 416 455 448 412 412 402 413 455 447 413 412 401 409 455 448 412 412 402
119.51 240.07 354.72 28.86 38.68 48.49 63.87 77.61 87.05 98.92 136.17 173.31 229.14 276.93 309.31 15.18 24.36 33.44 46.58 58.31 68.79 86.05 119.02 152.69 206.46 263.09 314.79
57.89 75.82 82.76 3.65 3.76 3.72 4.01 3.71 3.43 14.12 14.00 13.89 13.61 12.71 11.68 3.36 3.57 3.33 3.32 3.21 3.21 13.03 12.57 13.73 14.82 17.26 18.61
29 54 167 14 25 34 47 63 74 41 85 121 165 223 263 6 11 22 31 45 56 47 72 113 152 209 252
390 503 763 49 60 67 79 89 98 176 204 235 281 312 339 30 37 48 59 69 78 149 185 222 264 320 386
Table 2 Descriptive summary of children’s birth measurements. Variable
n
Mean
Standard deviation
Minimum
Maximum
Gestational age (days) Apgar score at 1 min Apgar score at 5 min Birth weight (kg) Head circumference (cm) Crown-heel length (cm) Crown-rump length (cm) Mid upper arm circumference (cm) Chest circumference (cm) Abdominal circumference at xyphisternum (cm) Placental weight (g)
481 474 474 487 484 484 481 475 479 476 480
275.13 8.06 9.29 3.14 34.33 49.45 33.04 10.36 32.42 32.39 576.71
13.08 1.78 1.06 0.51 1.69 2.94 2.24 1.01 2.17 2.01 128.48
193 0 1 1.23 27 36 23.5 5.3 23.5 24 226
299 10 10 4.69 44.5 57 48.5 13.5 41.5 37 1200
Brought to you by | University of California - San Diego Authenticated Download Date | 2/20/16 11:55 PM
698 Royal-Thomas et al., Maternal blood pressure in pregnancy and offspring blood pressure through adolescence time which are seen in the VWC developmental data. These methods include the random effects models and the generalized estimating equations models [7, 11, 14, 19]. This study uses linear mixed model (LMM) [9, 11].
Variable selection In the LMM, the main outcome variable was the child’s cardiovascular health data which were represented by the child’s longitudinal SBP from age 2 to 14 years. Regression models were fitted in a step-wise/hierarchical manner, which facilitated firstly fitting the mother’s cardiovascular health data along with her socioecomic status and weight gain during pregnancy; and secondly fitting the fetal developmental data followed by the birth measurements and adjusting for child’s cardiovascular health data with multiple regression. See Table 3 for a description of variables used in the selection process. We controlled for the child’s sex, age, and weight in the models. The slopes or change over time of some variables were computed to examine it’s association with child’s SBP later in life. The Akaike Information Criteria (AIC) was used in model selection and models with the smallest AIC were chosen as the optimal
models. In modeling the longitudinal data we applied the LMM with a random intercept and autocorrelation with lag 1 (AR(1)), which adjusted for serial correlation. In the LMM, continuous variables were standardized by transforming them to a z-score, which helped in creating standardized effect sizes for comparison of the variables that were measured on different scales. Some 12,553 measurements were taken of the children from birth to 14 years. The LMM notation that was used for the data is as follows:
Yij = β0 i + β 1 Xij 1 + β 2 Xij 2 + .........+ β p Xijp + εij ,
(1)
where εij = Wij+Zij is the within-subject error that composes the AR(1) correlation and measurement errors, respectively. β0i is the random intercept and Xijk are covariates for k = 1, …, p covariates with fixed effects βk. The final models were chosen as the optimal models with the smallest AIC. The LMM for our final two models are as follows: CBPij = β0 i + β 1 MBP0 i .1 + β 2 ∆MBPi .2 + β 3CBP0 i .3 + β4 Sexi .4 + β 5 Ageij 5
+ β6WTij 6 + εij ,
(2)
where CBPij represents the child’s longitudinal SBP for j = 1, …, 24 time-points for i = 487 children. MBP0i.1 is the mother’s baseline SBP
Table 3 Description of the variables used for selection in the LMM. Data
Independent variables
Description
Mother’s cardiovascular data
1. Mother’s baseline SBP 2. Mother’s SBP change
Mother’s first SBP reading during pregnancy Change in mother’s SBP over time computed by a simple regression model Variable from raw data Mother’s weight gain from booking to 35 weeks (kg) Change in child’s placental volume over time computed by a simple regression model Change in child’s biparietal diameter over time computed by a simple regression model Change in child’s femoral length over time computed by a simple regression model Change in child’s head circumference over time computed by a simple regression model Change in child’s abdominal circumference over time computed by a simple regression model Variable from raw data Variable from raw data Variable from raw data Variable from raw data Variable from raw data Variable from raw data Variable from raw data Variable from raw data Variable from raw data Variable from raw data Variable from raw data Child’s first SBP after birth Variable from raw data Variable from raw data Variable from raw data
Fetal developmental data
Child’s cardiovascular data at birth
3. Mother’s socioeconomic status 4. Mother’s weight gain in pregnancy 1. Placental volume slope
2. Biparietal diameter slope
3. Femoral length slope
4. Head circumference
5. Abdominal circumference
Child’s cardiovascular data from birth to 14 years
1. Gestational age (days) 2. Apgar score at 1 min 3. Apgar score at 5 min 4. Birth weight (kg) 5. Head circumference (cm) 6. Crown-heel length (cm) 7. Crown-rump length (cm) 8. Mid upper arm circumference (cm) 9. Chest circumference (cm) 10. Abdominal circumference (cm) 11. Placental weight (g) 1. Child’s baseline SBP 2. Weight 3. Age 4. Sex
Brought to you by | University of California - San Diego Authenticated Download Date | 2/20/16 11:55 PM
Royal-Thomas et al., Maternal blood pressure in pregnancy and offspring blood pressure through adolescence 699
and ΔMBPi.2 is the mother’s SBP slope estimates during pregnancy. The child’s baseline SBP at age 2 years is CBP0i.3 and Sexi.4 represents the sex of the child where 1 was for males and 2 was for females. The age of the child was Ageij5 and WTij6 is the longitudinal weight measurement of the child up to age 14 years. The error term εij = Wij+Zij is the within-subject error that composes the AR(1) correlation and measurement errors, respectively.
CBPij = β0 i + β 1 MBP0 i .1 + β 2 BWTi .2 + β 3 Sexi .3 + β4WTij 4 + εij .
(3)
Model 3 has the same notations as model 2 with the addition of BWTi.2, which represents child’s birth weight. There was collinearity among child’s baseline SBP, mother’s SBP slope, and mother’s baseline SBP measurements; therefore, these variables were fitted separately with birth weight in three different models. The best of these three models was chosen and that was Model 3 which had the lowest AIC.
Results The children’s measurements from birth to present indicated that there was an increase over time in the children’s height, weight, chest circumference, abdominal circumference, hip circumference, and lower segment. Children’s blood pressure demonstrated a slight increase over time, both manual and machine measurements for systolic and diastolic were consistent in this trend (see Figure 1). Children’s SBP was checked for normality and non-constancy of variance (heteroskedascity). From the boxplot in Figure 1, we can see that there is variability,
Figure 1 Loess curves and boxplots of children’s blood pressure measurements.
but this is consistent over time and the Loess smooth plot shows an increase in SBP over time in the same Figure 1. Table 4 shows the summary statistics for the mother’s data. Table 1 displays the summary statistics for the fetal development data and Table 2 shows the summary statistics for the children’s birth data. Table 5 shows the LMM estimates, which are all significantly positively predicting for child’s SBP. There was a strong correlation among child’s baseline SBP, mother’s SBP slope, and mother’s baseline SBP measurements. In order to remove this multicollinearity, these variables were fitted separately with birth weight in three different models. The best of these three models was chosen which is shown in Table 6. All the estimates for the covariates are positive except for birth weight which is negative and not statistically significant. Child’s weight and maternal SBP had the highest regression estimates of 0.374 and 0.132, respectively (see Table 6).
Discussion and conclusions Table 5 shows all significantly positive coefficients predicting for child’s SBP. The results indicate that mother’s SBP is significantly associated with child’s SBP while adjusting for the other variables. This demonstrates that a mother with a high baseline SBP and an increased trend in SBP predict for an increase in the child’s SBP over time. It seems that males are at an increased risk for higher SBP vs. females. We also see that with an increase in child’s weight, the child’s SBP increases while adjusting for the other variables. The results for Table 6 indicate that mother’s baseline SBP is significant in predicting child’s SBP over time, which again indicates that mothers with high SBP at the beginning of pregnancy will predict for children with higher SBP and vice versa. Table 6 also highlighted a most important finding which was that birth weight was not significant in it’s association with child’s SBP once maternal SBP, child’s sex, age, and weight were in the model. This indicates that maternal cardiovascular health and child’s cardiovascular health factors may be more important than birth weight in predicting a
Table 4 Mother’s descriptive measurements during pregnancy. Variable
n
Mean
Standard deviation
Minimum
Maximum
Systolic blood pressure (mm Hg) Weight (kg) Socioeconomic score
481 486 486
107.46 69.40 19.29
8.35 13.22 4.37
86.71 37.91 3.00
146.90 124.71 32.00
Brought to you by | University of California - San Diego Authenticated Download Date | 2/20/16 11:55 PM
700 Royal-Thomas et al., Maternal blood pressure in pregnancy and offspring blood pressure through adolescence Table 5 Maternal and children’s factors predicting child’s SBP. Predictors
Estimate (β)
Standard error
P-value
Intercept Mother’s baseline SBP Mother’s SBP change Child’s baseline SBP Age Sex (males) Child’s weight
–0.630 0.087 0.074 0.261 0.063 0.076 0.368
0.044 0.021 0.021 0.021 0.006 0.041 0.025
Tamika Royal-Thomas*, Daniel McGee, Debajyoti Sinha, Clive Osmond and Terrence Forrester
Association of maternal blood pressure in pregnancy with blood pressure of their offspring through adolescence Abstract: This article looks at the association of maternal blood pressure with the blood pressure of the offspring from birth to childhood. The Barker hypothesis states that maternal and “in utero” attributes during pregnancy affect a child’s cardiovascular health throughout life. We present an analysis of a unique dataset that consists of three distinct developmental processes: maternal cardiovascular health during pregnancy; fetal development; and child’s cardiovascular health from birth to 14 years. This study explored whether a mother’s blood pressure reading in pregnancy predicts fetal development and determines if this in turn is related to the future cardiovascular health of the child. This article uses data that have been collected prospectively from a Jamaican cohort which involves the following three developmental processes: (1) maternal cardiovascular health during pregnancy which is the blood pressure and anthropometric measurements at seven time-points on the mother during pregnancy; (2) fetal development which consists of ultrasound measurements of the fetus taken at six time-points during pregnancy; and (3) child’s cardiovascular health which consists of the child’s blood pressure measurements at 24 time-points from birth to 14 years. The inter-relationship of these three processes was examined using linear mixed effects models. Our analyses indicated that attributes later in childhood development, such as child’s weight, child’s baseline systolic blood pressure (SBP), age and sex, predict the future cardiovascular health of children. The results also indicated that maternal attributes in pregnancy, such as mother’s baseline SBP and SBP change, predicted significantly child’s SBP over time.
*Corresponding author: Dr. Tamika Royal-Thomas, Tropical Medicine Research Institute, University of the West Indies, Mona Campus, Kingston 7, Jamaica, Phone: +876-927-1884, E-mail: [email protected] Daniel McGee and Debajyoti Sinha: Statistics Department, Florida State University, Tallahassee, FL, USA Clive Osmond: MRC Lifecourse Epidemiology Unit, University of Southampton, Southampton General Hospital, Southampton, UK Terrence Forrester: UWI Solutions for Developing Countries, University of the West Indies, Mona Campus, Kingston 7, Jamaica
Keywords: Adolescence; blood pressure; cardiovascular disease (CVD); coronary heart disease (CHD); fetal development; in utero; infant; maternal health; pregnancy; systolic. DOI 10.1515/jpm-2014-0038 Received February 4, 2014. Accepted July 22, 2014. Previously published online August 28, 2014.
Introduction The purpose of this study is to examine the association of maternal blood pressure with the blood pressure of the offspring from birth to childhood. The Barker hypothesis states that adverse influences early in development, especially during intrauterine life, can result in permanent changes in physiology and metabolism, which lead to increased risk in adulthood [2, 3, 5, 6, 8]. Barker came up with this theory when he noticed that regions in England that had the highest rates of infant mortality in the early 1900s also had the highest rates of mortality from coronary heart disease (CHD) decades later [1, 5, 15]. As infant mortality was related to low birth weight, these observations led to the idea that infants who had low birth weight and survived infancy and childhood may be at increased risk for CHD [5]. The Barker hypothesis asserts that human fetuses have to adapt to a limited supply of nutrients and in doing so they can permanently change their structure and metabolism [1, 3, 8, 15]. Osmond and Barker [15] indicated that these fetal changes may be the origin of a number of diseases later in life which includes CHD, hypertension, and non-insulin dependent diabetes. Barker et al. looked at the relationship between small head circumference and thinness at birth to death from CHD in adult life [3]. Thame et al. demonstrated that blood pressure is associated with placental volume and birth weight [18]. Other studies have demonstrated factors that are associated with cardiovascular disease (CVD) or onset of hypertension later in life such as social class [10, 16] and low birth weight accompanied by accelerated childhood growth [12, 20]. There has been extensive work done on the hypothesis that adult CVD is determined by what happens “in utero”
Brought to you by | University of California - San Diego Authenticated Download Date | 2/20/16 11:55 PM
696 Royal-Thomas et al., Maternal blood pressure in pregnancy and offspring blood pressure through adolescence and during the early days of an infant’s childhood. Several studies have described the relationship between birth weight and blood pressure [2, 3, 6, 13, 15, 18], but very few studies include longitudinal data. Furthermore, the methods of data analysis were mainly descriptive or did not fully utilize the available data. This article has utilized data that have been collected prospectively from a Jamaican cohort involving three longitudinal processes. The Jamaican dataset is called the vulnerable windows cohort (VWC) and is unique in that it has the potential to address whether and how much the prebirth measurements influence future cardiovascular risk. A more complete description of the data and the longitudinal processes is presented in the Methods section. The VWC is a complex dataset which gives a unique opportunity to examine epidemiological associations on a more detailed level. The main aims of the study were to examine if a child’s fetal development predicts his cardiovascular health later in life, and does the maternal cardiovascular health during pregnancy predict the child’s fetal development, and in turn, do both processes predict the child’s cardiovascular health throughout life. The results from the study indicated that attributes such as mother’s baseline systolic blood pressure (SBP) and SBP change in pregnancy have a positive association with child’s SBP over time while controlling child’s weight, age, and sex.
Methods Data Some 712 women attending the antenatal clinic of the University Hospital of the West Indies, Jamaica, were invited to participate in a prospective study called the VWC. The recruitment period lasted from November 1992 to July 1994. The women were included if they were between 15 and 40 years of age, were sure of the date of their last menstrual period (as confirmed by a 14-week ultrasound), had singleton pregnancies, and did not have systemic illnesses (e.g., pregnancy-induced hypertension, diabetes mellitus, etc.) or genetic abnormalities. After excluding 82 pregnancy losses/stillbirths, five twin gestations, and 56 women who refused to participate in the study, 569 women and their offspring were enrolled. The mothers’ ages, weights, and heights were recorded together with their parity on their first visit to the antenatal clinic. All women were offered transportation to and from the hospital to enhance participation in the study. Their socioeconomic status was scored using the variables of household amenities, crowding in the home, parents’ educational level, and parents’ occupation [4, 17]. Maternal hemodynamic, anthropometric, biochemical, and hematological statuses were assessed throughout the pregnancy. At each visit, maternal blood pressure was measured with an oscillometric sphygmomanometer (Dinamap TM monitor Model 8100,
Critikon Inc., Tampa, FL, USA). Three individuals, Dr. Minerva Thame (MT), a nurse and a medical technologist, made all measurements. Two of the three observers made the ultrasound measurements (MT and the technologist). All three were trained to apply the questionnaires and make the measurements. At the start, and at 3 monthly intervals for the duration of the study, inter- and intraobserver measurement variabilities were assessed, and training and recertification prescribed for any observer whose scores were not acceptable. The children were measured at birth, at 6 weeks, every 3 months–2 years, and then every 6 months thereafter at a special follow-up clinic. At each clinic attendance, blood pressure was measured with an oscillometric device (Dinamap TM Monitor Model 8100, Critikon Inc., Tampa, FL, USA) and manually. Two measurements were taken using the Dinamap which were averaged and two manual readings were taken which were also averaged. The averaged Dinamap readings were used in the statistical analyses [17]. Fetal and placental growth were assessed by maternal abdominal ultrasound and birth anthropometric measurements were made using standardized techniques. The method used to measure placental volume required that the entire placenta be seen on the screen. After 20 weeks’ gestation, many placentas are too large for this, so placental volume was measured at only the first three visits. The average of three repeats was used for each measurement. Placental volume was measured by identifying and recording on videotape, the long axis of the placenta. A continuous recording of the image of the placenta orthogonal to the axis was made by sweeping the probe along the axis at constant velocity. This axis was divided into six sections of equal length; the five interior cross-sectional areas were measured and integrated to estimate the placental volume [17]. At birth, detailed anthropometric measurements of the newborn infant were obtained using standardized methods and included weight, crown-heel length, chest and head circumferences, and placental weight. Gestational age was estimated from the date of the last menstrual period. The offspring were measured at birth, at 6 weeks, every 3 months–2 years, and then every 6 months thereafter at a special follow-up clinic. As described by Thame et al. [18], at each clinic attendance blood pressure was measured with an oscillometric device (Dinamap TM Monitor, Model 8100, Critikon Inc. Tampa, FL, USA) and manually. Birth weight was measured with a Health o Meter® 457 scale (Pelstar LLC, Bridgeview, IL, USA); height with a length board (Holtain Ltd., Crymych, Dyfed, UK) for children ≤ 2 years and thereafter with a stadiometer (Holtain Ltd., Crymych, Dyfed, UK). Four skinfold thicknesses were measured [4]. Mothers and children were seen twice a year since then for anthropometric, blood pressure, and pulse rate measurements. The present data analysis is confined to children who were seen at all scheduled visits between birth and age 14 years, which started out with 487 children and decreased over time. A detailed description of the measurements is provided by Thame et al. [17, 18]. The VWC data consist of three developmental processes. The first developmental process is the maternal cardiovascular health during pregnancy data which describes the mother’s SBP, diastolic blood pressure, pulse rate, and anthropometric measurements such as height, weight, skinfold measurements among other variables at seven time-points (8–10, 14, 17, 20, 25, 30, 35 weeks) during pregnancy. The second developmental process is the fetal development data that have been collected on the infants while they were in the womb. Six measurements of femoral length, abdominal circumference, biparietal diameter, and head circumference were estimated at 14, 17, 20, 25, 30 and 35 weeks by ultrasound, and placental volume
Brought to you by | University of California - San Diego Authenticated Download Date | 2/20/16 11:55 PM
Royal-Thomas et al., Maternal blood pressure in pregnancy and offspring blood pressure through adolescence 697
was estimated at three time-points (14, 17, and 20 weeks) by ultrasound (see Table 1). The third developmental process consists of the child’s cardiovascular health data taken from birth to current age (see Table 2 for the descriptive summary of birth data). This process consisted of approximately 34 measurements of anthropometric data such as height, weight, head circumference, waist circumference, hip circumference, skinfold measurements, and other anthropometric variables and 24 measurements of SBP, diastolic blood pressure,
and pulse rate measurements which were taken from 2 years to current age.
Statistical analyses There are various traditional statistical methods that have been used for the analysis of longitudinal data, i.e., data collected over
Table 1 In utero descriptive measurements during pregnancy. Variable
n
Mean
Standard deviation
Minimum
Maximum
Placental volume at 14 weeks (mL) Placental volume at 17 weeks (mL) Placental volume at 20 weeks (mL) Biparietal diameter at 14 weeks (mm) Biparietal diameter at 17 weeks (mm) Biparietal diameter at 20 weeks (mm) Biparietal diameter at 25 weeks (mm) Biparietal diameter at 30 weeks (mm) Biparietal diameter at 35 weeks (mm) Head circumference at 14 weeks (mm) Head circumference at 17 weeks (mm) Head circumference at 20 weeks (mm) Head circumference at 25 weeks (mm) Head circumference at 30 weeks (mm) Head circumference at 35 weeks (mm) Femoral length at 14 weeks (mm) Femoral length at 17 weeks (mm) Femoral length at 20 weeks (mm) Femoral length at 25 weeks (mm) Femoral length at 30 weeks (mm) Femoral length at 35 weeks (mm) Abdominal circumference at 14 weeks (mm) Abdominal circumference at 17 weeks (mm) Abdominal circumference at 20 weeks (mm) Abdominal circumference at 25 weeks (mm) Abdominal circumference at 30 weeks (mm) Abdominal circumference at 35 weeks (mm)
408 451 435 417 456 445 412 412 402 416 455 448 412 412 402 413 455 447 413 412 401 409 455 448 412 412 402
119.51 240.07 354.72 28.86 38.68 48.49 63.87 77.61 87.05 98.92 136.17 173.31 229.14 276.93 309.31 15.18 24.36 33.44 46.58 58.31 68.79 86.05 119.02 152.69 206.46 263.09 314.79
57.89 75.82 82.76 3.65 3.76 3.72 4.01 3.71 3.43 14.12 14.00 13.89 13.61 12.71 11.68 3.36 3.57 3.33 3.32 3.21 3.21 13.03 12.57 13.73 14.82 17.26 18.61
29 54 167 14 25 34 47 63 74 41 85 121 165 223 263 6 11 22 31 45 56 47 72 113 152 209 252
390 503 763 49 60 67 79 89 98 176 204 235 281 312 339 30 37 48 59 69 78 149 185 222 264 320 386
Table 2 Descriptive summary of children’s birth measurements. Variable
n
Mean
Standard deviation
Minimum
Maximum
Gestational age (days) Apgar score at 1 min Apgar score at 5 min Birth weight (kg) Head circumference (cm) Crown-heel length (cm) Crown-rump length (cm) Mid upper arm circumference (cm) Chest circumference (cm) Abdominal circumference at xyphisternum (cm) Placental weight (g)
481 474 474 487 484 484 481 475 479 476 480
275.13 8.06 9.29 3.14 34.33 49.45 33.04 10.36 32.42 32.39 576.71
13.08 1.78 1.06 0.51 1.69 2.94 2.24 1.01 2.17 2.01 128.48
193 0 1 1.23 27 36 23.5 5.3 23.5 24 226
299 10 10 4.69 44.5 57 48.5 13.5 41.5 37 1200
Brought to you by | University of California - San Diego Authenticated Download Date | 2/20/16 11:55 PM
698 Royal-Thomas et al., Maternal blood pressure in pregnancy and offspring blood pressure through adolescence time which are seen in the VWC developmental data. These methods include the random effects models and the generalized estimating equations models [7, 11, 14, 19]. This study uses linear mixed model (LMM) [9, 11].
Variable selection In the LMM, the main outcome variable was the child’s cardiovascular health data which were represented by the child’s longitudinal SBP from age 2 to 14 years. Regression models were fitted in a step-wise/hierarchical manner, which facilitated firstly fitting the mother’s cardiovascular health data along with her socioecomic status and weight gain during pregnancy; and secondly fitting the fetal developmental data followed by the birth measurements and adjusting for child’s cardiovascular health data with multiple regression. See Table 3 for a description of variables used in the selection process. We controlled for the child’s sex, age, and weight in the models. The slopes or change over time of some variables were computed to examine it’s association with child’s SBP later in life. The Akaike Information Criteria (AIC) was used in model selection and models with the smallest AIC were chosen as the optimal
models. In modeling the longitudinal data we applied the LMM with a random intercept and autocorrelation with lag 1 (AR(1)), which adjusted for serial correlation. In the LMM, continuous variables were standardized by transforming them to a z-score, which helped in creating standardized effect sizes for comparison of the variables that were measured on different scales. Some 12,553 measurements were taken of the children from birth to 14 years. The LMM notation that was used for the data is as follows:
Yij = β0 i + β 1 Xij 1 + β 2 Xij 2 + .........+ β p Xijp + εij ,
(1)
where εij = Wij+Zij is the within-subject error that composes the AR(1) correlation and measurement errors, respectively. β0i is the random intercept and Xijk are covariates for k = 1, …, p covariates with fixed effects βk. The final models were chosen as the optimal models with the smallest AIC. The LMM for our final two models are as follows: CBPij = β0 i + β 1 MBP0 i .1 + β 2 ∆MBPi .2 + β 3CBP0 i .3 + β4 Sexi .4 + β 5 Ageij 5
+ β6WTij 6 + εij ,
(2)
where CBPij represents the child’s longitudinal SBP for j = 1, …, 24 time-points for i = 487 children. MBP0i.1 is the mother’s baseline SBP
Table 3 Description of the variables used for selection in the LMM. Data
Independent variables
Description
Mother’s cardiovascular data
1. Mother’s baseline SBP 2. Mother’s SBP change
Mother’s first SBP reading during pregnancy Change in mother’s SBP over time computed by a simple regression model Variable from raw data Mother’s weight gain from booking to 35 weeks (kg) Change in child’s placental volume over time computed by a simple regression model Change in child’s biparietal diameter over time computed by a simple regression model Change in child’s femoral length over time computed by a simple regression model Change in child’s head circumference over time computed by a simple regression model Change in child’s abdominal circumference over time computed by a simple regression model Variable from raw data Variable from raw data Variable from raw data Variable from raw data Variable from raw data Variable from raw data Variable from raw data Variable from raw data Variable from raw data Variable from raw data Variable from raw data Child’s first SBP after birth Variable from raw data Variable from raw data Variable from raw data
Fetal developmental data
Child’s cardiovascular data at birth
3. Mother’s socioeconomic status 4. Mother’s weight gain in pregnancy 1. Placental volume slope
2. Biparietal diameter slope
3. Femoral length slope
4. Head circumference
5. Abdominal circumference
Child’s cardiovascular data from birth to 14 years
1. Gestational age (days) 2. Apgar score at 1 min 3. Apgar score at 5 min 4. Birth weight (kg) 5. Head circumference (cm) 6. Crown-heel length (cm) 7. Crown-rump length (cm) 8. Mid upper arm circumference (cm) 9. Chest circumference (cm) 10. Abdominal circumference (cm) 11. Placental weight (g) 1. Child’s baseline SBP 2. Weight 3. Age 4. Sex
Brought to you by | University of California - San Diego Authenticated Download Date | 2/20/16 11:55 PM
Royal-Thomas et al., Maternal blood pressure in pregnancy and offspring blood pressure through adolescence 699
and ΔMBPi.2 is the mother’s SBP slope estimates during pregnancy. The child’s baseline SBP at age 2 years is CBP0i.3 and Sexi.4 represents the sex of the child where 1 was for males and 2 was for females. The age of the child was Ageij5 and WTij6 is the longitudinal weight measurement of the child up to age 14 years. The error term εij = Wij+Zij is the within-subject error that composes the AR(1) correlation and measurement errors, respectively.
CBPij = β0 i + β 1 MBP0 i .1 + β 2 BWTi .2 + β 3 Sexi .3 + β4WTij 4 + εij .
(3)
Model 3 has the same notations as model 2 with the addition of BWTi.2, which represents child’s birth weight. There was collinearity among child’s baseline SBP, mother’s SBP slope, and mother’s baseline SBP measurements; therefore, these variables were fitted separately with birth weight in three different models. The best of these three models was chosen and that was Model 3 which had the lowest AIC.
Results The children’s measurements from birth to present indicated that there was an increase over time in the children’s height, weight, chest circumference, abdominal circumference, hip circumference, and lower segment. Children’s blood pressure demonstrated a slight increase over time, both manual and machine measurements for systolic and diastolic were consistent in this trend (see Figure 1). Children’s SBP was checked for normality and non-constancy of variance (heteroskedascity). From the boxplot in Figure 1, we can see that there is variability,
Figure 1 Loess curves and boxplots of children’s blood pressure measurements.
but this is consistent over time and the Loess smooth plot shows an increase in SBP over time in the same Figure 1. Table 4 shows the summary statistics for the mother’s data. Table 1 displays the summary statistics for the fetal development data and Table 2 shows the summary statistics for the children’s birth data. Table 5 shows the LMM estimates, which are all significantly positively predicting for child’s SBP. There was a strong correlation among child’s baseline SBP, mother’s SBP slope, and mother’s baseline SBP measurements. In order to remove this multicollinearity, these variables were fitted separately with birth weight in three different models. The best of these three models was chosen which is shown in Table 6. All the estimates for the covariates are positive except for birth weight which is negative and not statistically significant. Child’s weight and maternal SBP had the highest regression estimates of 0.374 and 0.132, respectively (see Table 6).
Discussion and conclusions Table 5 shows all significantly positive coefficients predicting for child’s SBP. The results indicate that mother’s SBP is significantly associated with child’s SBP while adjusting for the other variables. This demonstrates that a mother with a high baseline SBP and an increased trend in SBP predict for an increase in the child’s SBP over time. It seems that males are at an increased risk for higher SBP vs. females. We also see that with an increase in child’s weight, the child’s SBP increases while adjusting for the other variables. The results for Table 6 indicate that mother’s baseline SBP is significant in predicting child’s SBP over time, which again indicates that mothers with high SBP at the beginning of pregnancy will predict for children with higher SBP and vice versa. Table 6 also highlighted a most important finding which was that birth weight was not significant in it’s association with child’s SBP once maternal SBP, child’s sex, age, and weight were in the model. This indicates that maternal cardiovascular health and child’s cardiovascular health factors may be more important than birth weight in predicting a
Table 4 Mother’s descriptive measurements during pregnancy. Variable
n
Mean
Standard deviation
Minimum
Maximum
Systolic blood pressure (mm Hg) Weight (kg) Socioeconomic score
481 486 486
107.46 69.40 19.29
8.35 13.22 4.37
86.71 37.91 3.00
146.90 124.71 32.00
Brought to you by | University of California - San Diego Authenticated Download Date | 2/20/16 11:55 PM
700 Royal-Thomas et al., Maternal blood pressure in pregnancy and offspring blood pressure through adolescence Table 5 Maternal and children’s factors predicting child’s SBP. Predictors
Estimate (β)
Standard error
P-value
Intercept Mother’s baseline SBP Mother’s SBP change Child’s baseline SBP Age Sex (males) Child’s weight
–0.630 0.087 0.074 0.261 0.063 0.076 0.368
0.044 0.021 0.021 0.021 0.006 0.041 0.025
