Original Article
Beat-to-Beat Heart Rate and Blood Pressure Variability and Hypertensive Disease in Pregnancy Pamela Flood, MD, MA1 Paula McKinley, PhD2 Catherine Monk, PhD3 Paul Muntner, PhD4 Lisandro D. Colantonio, MD, MSc4 Laura Goetzl, MD, MPH5 Maureen Hatch, PhD6 Richard P. Sloan, PhD2 1 Department of Anesthesia, Perioperative and Pain Medicine, Stanford
University, San Francisco, California 2 Department of Behavioral Medicine, Columbia University, New York, New York 3 Department of Psychiatry, Behavioral Medicine and Developmental Neuroscience, Columbia University, New York, New York 4 Department of Epidemiology, University of Alabama at Birmingham, Birmingham, Alabama 5 Department of Obstetrics and Gynecology, Temple University, Philadelphia, Pennsylvania 6 National Institutes of Health/National Cancer Institute, Division of Cancer Epidemiology & Genetics, Radiation Epidemiology Branch Bethesda, Maryland
Address for correspondence Pamela Flood, MD, MA, Department of Anesthesia, Perioperative and Pain Medicine, Stanford University, 300 Pasteur Drive, Palo Alto, CA 94304 (e-mail: pfl[email protected]).
Am J Perinatol 2015;32:1050–1058.
Abstract
Keywords
► hypertensive diseases of pregnancy ► autonomic function ► heart rate variability ► blood pressure variability
received November 30, 2014 accepted after revision January 30, 2015 published online May 13, 2015
Objective The aim of this study is to determine the relationship between heart rate and/or blood pressure variability, measured at 28 weeks’ gestation, and the incidence of pregnancy-induced hypertension or preeclampsia. Study Design Secondary analysis of data from a prospectively enrolled cohort of 385 active military women in whom spectral analysis of continuous heart rate and variability was measured at 28 weeks’ gestation. The primary outcome was the predictive value of spectral analysis of heart rate and blood pressure for hypertensive diseases of pregnancy. Results High-frequency heart rate variability was reduced and low-frequency variability of systolic and diastolic blood pressure increased in women who would develop pregnancy-induced hypertension but not preeclampsia. Low-frequency variability of diastolic blood pressure remained a significant predictor of pregnancy-induced hypertension but not preeclampsia after adjustment for age, weight, and blood pressure in a multivariate model. Conclusion Early identification of pregnancy-induced hypertension can facilitate treatment to avoid maternal morbidity. Understanding the physiological underpinnings of the two very different diseases may lead to improved treatment and prevention. If proven effective in a broader population, the ability to differentiate pregnancy-induced hypertension from preeclampsia may reduce unnecessary iatrogenic interventions or inappropriate preterm delivery.
Copyright © 2015 by Thieme Medical Publishers, Inc., 333 Seventh Avenue, New York, NY 10001, USA. Tel: +1(212) 584-4662.
DOI http://dx.doi.org/ 10.1055/s-0035-1548542. ISSN 0735-1631.
Downloaded by: NYU. Copyrighted material.
1050
Hypertensive disorders of pregnancy, including preeclampsia (PE) and pregnancy-induced hypertension (PIH), affect approximately 10 to 15% of pregnancies and are a significant underlying factor in severe maternal and neonatal morbidity and mortality.1 Although preventive interventions are still lacking, accurate noninvasive, low cost early predictors for these disorders may improve maternal and fetal outcomes through increased surveillance and early treatment of highrisk populations. Furthermore, identification of markers associated with increased risk for PE and PIH will be useful for designing trials of therapeutic interventions. One potential method for early detection of hypertensive disease during pregnancy focuses on associated dysregulation of the autonomic nervous system (ANS). Two of the following noninvasive indices of ANS dysregulation are of special interest: heart rate variability (HRV) and blood pressure variability (BPV). HRV, especially at high frequency (HF) (range, 0.15–0.40 Hz), is an index of cardiac parasympathetic modulation.2,3 In previous studies, women with normal pregnancy (at 34 weeks) had lower HF-HRV compared with control women who were not pregnant. In women who would develop PE, HF-HRV was lower still.4 These data suggest that cardiac autonomic dysfunction may be an early marker of pathophysiology in women who go on to develop hypertensive disorders of pregnancy. Like heart rate (HR), blood pressure (BP) exhibits periodic oscillations at high and low frequencies (range, 0.04–0.15 Hz, LF).5 While HF-BPV is a direct product of respiratory-induced mechanical fluctuations of cardiac output which change arterial pressure,6 some studies suggest that LF-BPV reflects vascular sympathetic drive.7–13 A careful observational study of the natural course and outcome of pregnant patients with autonomic dysfunction has the potential to identify early markers of hypertensive disorders of pregnancy before the disease onset and thereby contribute to the determination of the underlying pathophysiology and development of clinical interventions. Here, we test the hypothesis that ANS function, measured at 28 weeks’ gestation, predicts later development of PE and PIH in a large cohort of military women.
Patients and Methods Subjects From 1996 to 2001, data were collected from pregnant military women stationed at Lackland Air Force Base in San Antonio, Texas for a study of “Cardiovascular responsivity, physical and psychosocial job stress, and the risk of preterm delivery” (Maureen Hatch, PI) that was funded by the Department of Defense. Results addressing the primary aim have been published.14 As part of the protocol, HR and BP variability were measured in a subset of study participants to determine if responsiveness to psychological stress predicted preterm delivery. The study protocol was approved by the relevant institutional review boards. All participants signed informed consent for the original study. Deidentified data from that project were used in this analysis. Here, we report on HRV and BPV data from the resting baseline period that preceded psychological testing in the original study, and
Flood et al.
investigate their relationship to the development of hypertensive disease of pregnancy. A total of 724 pregnant active-duty military women who attended the prenatal clinic between 1997 and 2000 were enrolled early in pregnancy (6–12 weeks of gestation). Overall, 80 women had pregnancy loss or delivery before 28 weeks’ gestation and 63 left the military base and were considered lost to follow-up. Some deliveries before 28 weeks were secondary to hypertensive complications of pregnancy; eliminating early-onset PIH or PET from the study population. Overall, 581 subjects returned for a 28-week assessment visit that included measurement of HRV and BPV and 385 completed the psychophysiology test. Data from these subjects were used in this secondary data analysis. Most women who declined to participate in the psychophysiology study indicated that they did not have time to stay for the test. None of the subjects had a current clinical diagnosis of PIH or PE at the time of autonomic testing. Overall, 21 subjects reported an episode of hypertension before their pregnancy and 4 subjects had required medication to control their BP. Medical data including diagnoses of PIH and PE were abstracted from hospital records by Hatch for the original study; PIH and PE were based on treating physician diagnosis using standard clinical definitions as recommended by ACOG at the time of treatment. Gestational age was based on the due date estimated by ultrasound measurements before 20 weeks’ gestation. Maternal weight was recorded before pregnancy and at delivery.
Measurement of Heart Rate Variability Participants were tested in the seated position. Electrodes were placed on the right shoulder, on the left anterior axillary line at the 10th intercostal space, and in the right lower quadrant for standard three-lead ECG measurement. After a 5-minute resting period, analog electrocardiogram signals were digitized at 500 Hz by a 12-bit analog-to-digital (A/D) conversion board and passed to a microcomputer. The electrocardiogram waveform was submitted to an R wave detection routine implemented by custom-written event detection software, resulting in an RR interval series. Errors in marking of R waves were corrected by visual inspection. Ectopic beats were corrected by interpolation. HF-HRV (range, 0.15–0.50 Hz) from 300-second epochs were computed with an interval method for computing Fourier transforms similar to that described by deBoer et al.6 Before computing Fourier transforms, the mean of the RR interval series was subtracted from each value in the series, then filtered the series with a Hanning window15 and summed the power (i.e., variance, in msec2) over the HF band. Estimates of spectral power were adjusted to account for attenuation produced by this filter.15
Measurement of Blood Pressure and Blood Pressure Variability The beat-to-beat BP waveform was obtained using a Finapres noninvasive BP monitor (Finapres Medical Systems, Amsterdam, The Netherlands) with a finger cuff placed on the middle finger of the nondominant hand during the same session as American Journal of Perinatology
Vol. 32
No. 11/2015
1051
Downloaded by: NYU. Copyrighted material.
Spectral Analysis in Pregnancy
Spectral Analysis in Pregnancy
Flood et al.
HRV measurement earlier. The analog BP waveform also was sampled by the A/D board at 500 samples per second. Systolic and diastolic BP values for each cardiac cycle were identified using custom-written software resulting in a BP time series. Mean BP and spectral power in the low (range, 0.04–0.15 Hz) frequency band of the BP power spectrum for both systolic BP and diastolic BP were computed from these time series. Because the servo adjustment of the Finapres monitor was enabled during the last minute of each 300 second period, spectra were calculated on 240 second epochs using the method described earlier. Power, that is, variance (in mm Hg2), over the LF band was summed and adjusted to account for attenuation produced by the Hanning window.15
Statistical Analyses HF-HRV and LF-BPV data were log-transformed before the analysis as the data were not normally distributed (skewed to the right). Characteristics of women who developed PIH or PE versus those who remained normotensive were compared using t-test or Fisher exact test for continuous and categorical variables, respectively. Unadjusted odds ratios and 95% confidence intervals (95% CI) for developing PIH or PE, separately, versus remaining normotensive associated with age, race, HR, mean systolic and diastolic BP, HF-HRV, and LF-HRV were estimated using logistic regression. In addition, two logistic regression models including progressive adjustment were conducted. The first model included adjustment for age, weight, and HR. The second (fully adjusted) model included adjustment for variables included in the first model plus mean systolic and diastolic BP, HF-HRV, and LF-BPV. Different fully adjusted models were conducted for diastolic or systolic LF-BPV.
All regression models were conducted after performing multiple imputations for missing data. Missing data were imputed for age (n ¼ 6), race (n ¼ 6), HR (n ¼ 16), predelivery weight (n ¼ 6), and HF-HRV (n ¼ 16). No data were missing on hypertensive disorders of pregnancy, mean systolic or diastolic BP, or LF-BPV. For the multiple imputations, 10 data sets using chained equations were assembled based on observed values from all the variables collected and the occurrence of hypertensive disorders of pregnancy.16 All analyses were conducted using Stata/I.C. 13.0 (Stata Corporation, College Station, TX).
Results Of the 385 women who underwent psychophysiological testing, 26 developed PIH, 27 developed PE, and 332 did not develop hypertensive disease of pregnancy. Among the 21 women who reported an episode of hypertension before pregnancy, 12 went on to be diagnosed with PIH and 6 PE and 2 subjects did not become hypertensive during their pregnancy, and 1 was lost to follow-up. Demographic variables for subjects by group are shown in ►Table 1. Women who would go on to develop PIH were older (p < 0.04) and heavier, both before pregnancy and at the time of delivery (p < 0.01, 0.002), compared with normotensive subjects (►Table 1). Maternal weight before pregnancy and at delivery (p < 0.02, 0.01) was also higher in those who developed PE compared with normotensive women. There were no racial differences between women who developed PE or PIH and those who did not in this cohort. There was no difference in neonatal weight or gestational age at birth for infants born to women with PIH. Neonates born to mothers with PE were smaller (2.98 0.84
Table 1 Baseline characteristics of participants included in the analysis by presence of hypertensive disorders of pregnancya Participants included (N ¼ 385) Age (y), mean (SD) Race/Ethnicity, (%)
NT (n ¼ 332)
PE (n ¼ 27)
p Value
PIH (n ¼ 26)
p Value
26.4 (5.1)
27.9 (6.3)
0.18
28.7 (5.7)
0.04
0.19
58.4
0.87
b
White
62.1
48.0
Black
17.9
32.0
20.8
Other
20.0
20.0
20.8
135.1 (18.3)
144.8 (18.4)
Weight prepregnancy (lbs.), mean (SD)
0.02
145.7 (23.4)
0.01
Weight before delivery (lbs.), mean (SD)
175.2 (23.5)
188.2 (25.3)
0.01
190.3 (28.9)
0.002
HR (bpm), mean (SD)c
86.7 (10.4)
86.7 (8.7)
0.98
90.0 (11.6)
0.12
c
108.3 (11.1)
112.5 (12.5)
0.07
119.1 (15.9)
< 0.001
c
58.6 (8.2)
64.3 (10.2)
< 0.001
64.8 (11.7)
< 0.001
358 (314, 408)
363 (197, 668)
0.95
206 (116, 365)
0.03
c
9.7 (9.1, 10.3)
8.9 (7.0, 11.3)
0.51
12.7 (9.5, 17.0)
0.03
c
4.1 (3.9, 4.4)
4.3 (3.5, 5.3)
0.80
6.2 (4.8, 8.0)
0.002
SBP (mmHg), mean (SD)
DBP (mmHg), mean (SD)
HF-HRV, geometric mean (95% CI)c LF-BPV (SBP), geometric mean (95% CI)
LF-BPV (DBP), geometric mean (95% CI)
Abbreviations: 95% CI, 95% confidence interval; bpm, beats per minute; HF-HRV, high frequency heart rate variability; LF-BPV, low frequency blood pressure variability; NT, normotensive; PE, preeclampsia; PIH, pregnancy-induced hypertension. a Nonimputed data. b Fisher exact test. c Blood pressure and heart rate were measured at 28 weeks’ gestation—the time of psychophysiological testing. American Journal of Perinatology
Vol. 32
No. 11/2015
Downloaded by: NYU. Copyrighted material.
1052
Spectral Analysis in Pregnancy
Flood et al.
vs. 3.41 0.50 g, p < 0.00001) and more premature (38 3 vs. 39 2 weeks, p < 0.002) than the neonates born to normotensive women. At 28 weeks’ gestation, there was no difference in resting HR in women who developed PE or PIH compared with those who remained normotensive (►Table 1). In women who would go on to develop PIH, both mean systolic (p < 0.001) and diastolic BP (p < 0.001) were higher (►Fig. 1). Diastolic BP was higher in women who would develop PE (p < 0.001) compared with those who remained normotensive. During psychophysiological testing at 28 weeks’ gestation, only three subjects (one in each group) had diastolic BP values that if repeated and correlated with other clinical criteria would be diagnostic for PIH or chronic hypertension (greater than 90 mm Hg). Six subjects had a systolic BP over 140 at psychophysiological testing, three in the NT, one in the PIH, and two subjects in the PE group. In women who went on to develop PIH, HF-HRV was reduced (►Table 1, ►Fig. 2, p < 0.03) and LF-BPV for both systolic and diastolic BP was increased (►Table 1, ►Fig. 3a, b; p < 0.03 and 0.002, respectively) relative to women with normal pregnancies. Neither HRV nor BPV were predictive of the development of PE. A multivariate model for prediction of PIH was created adjusted for age, weight, and resting HR (►Table 2, adjusted model 1). Model 2a including items in model 1 with additional adjustment for SBP, DBP, HR-HFV, and LF-BPV estimated from systolic BP demonstrated the importance of weight before delivery and LF-BPV derived from the systolic BP time series. Alternative model 2b including additional adjustment for SBP, DBP, HF-HRV, and LF-BPV estimated
from diastolic BP suggested that only maternal weight contributed to the final model. As such, only LF-BPV estimated from systolic BP added significant predictive power to mean systolic and diastolic BP for the prediction of PIH. The multivariate model for prediction of PE was adjusted for age, weight, and resting HR (►Table 3, adjusted model 1). Model 2a including items in model 1 with additional adjustment for SBP, DBP, HR-HFV, and LF-BPV estimated from systolic BP showed a significant contribution from weight before delivery and diastolic BP at 28 weeks’ gestation. Alternative model 2b including additional adjustment for SBP, DBP, HF-HRV, and LF-BPV estimated from diastolic BP was similar. Spectral
Fig. 1 Systolic and diastolic blood pressure distributions by presence of hypertensive disorders during pregnancy. DBP, diastolic blood pressure; NT, normotensive; PE, preeclampsia; PIH, pregnancy-induced hypertension; SBP, systolic blood pressure.
Fig. 3 Low-frequency systolic and diastolic blood pressure variability distributions by presence of hypertensive disorders during pregnancy. DBP, diastolic blood pressure; LF-BPV, low-frequency blood pressure; NT, normotensive; PE, preeclampsia; PIH, pregnancy-induced hypertension; SBP, systolic blood pressure.
1053
American Journal of Perinatology
Vol. 32
No. 11/2015
Downloaded by: NYU. Copyrighted material.
Fig. 2 High-frequency heart rate variability (HF-HRV) distribution by presence of hypertensive disorders during pregnancy. NT, normotensive; PE, preeclampsia; PIH, pregnancy-induced hypertension.
American Journal of Perinatology
Vol. 32
No. 11/2015
2.07 (1.30, 3.32) 1.64 (1.05, 2.59)
LF-BPV (DBP)a
a
0.03
0.002
0.03 1.65 (1.04, 2.62)
2.27 (1.39, 3.68)
0.90 (0.63, 1.29)
1.93 (1.31, 2.84)
2.20 (1.46, 3.31)
< 0.001 0.001
1.53 (1.02, 2.30)
0.13
0.03
0.001
0.56
0.001
< 0.001
0.04
0.006
0.66
0.03
p Value
1.81 (1.07, 3.06)
0.80 (0.55, 1.18)
0.92 (0.45, 1.90)
1.99 (0.95, 4.16)
1.13 (0.59, 2.19)
1.78 (1.15, 2.74)
2.34 (0.72, 7.61)
1.57 (0.48, 5.19)
1 (ref)
1.43 (0.90, 2.26)
OR (95% CI)
Adjusted model 2a
0.03
0.26
0.83
0.07
0.71
0.01
0.35
0.13
p Value
1.39 (0.85, 2.27)
0.84 (0.57, 1.22)
1.08 (0.53, 2.21)
1.96 (0.93, 4.12)
1.16 (0.59, 2.26)
1.77 (1.15, 2.72)
2.17 (0.67, 6.98)
1.57 (0.48, 5.14)
1 (ref)
1.39 (0.88, 2.18)
OR (95%CI)
Adjusted model 2b
0.19
0.35
0.83
0.08
0.67
0.009
0.41
0.16
p Value
Downloaded by: NYU. Copyrighted material.
Abbreviations: 95% CI, 95% confidence interval; DBP, diastolic blood pressure; HF-HRV, high frequency heart rate variability; LF-BPV, low frequency blood pressure variability; OR, odds ratio; SBP, systolic blood pressure; SD, standard deviation. a Log-transformed variables, with 1 unit representing doubling values. Notes: All models include multiple imputation including age, race, weight before delivery, heart rate, SBP, DBP, HF-HRV, LF-BPV, and hypertensive disorders of pregnancy. Model 1 includes adjustment for age, weight, and heart rate. Model 2a includes adjustment for covariates in model 1 plus mean SBP, DBP, HF-HRV, and LF-BPV (estimated using diastolic blood pressure). Model 2b includes adjustment for covariates in model 1 plus mean SBP, DBP, HF-HRV, and LF-BPV (estimated using systolic blood pressure).
LF-BPV (SBP)
1.84 (1.29, 2.62) 0.77 (0.61, 0.97)
HF-HRVa
2.18 (1.49, 3.20)
SBP (per 1 SD)
DBP (per 1 SD)
1.34 (0.92, 1.96)
Heart rate (per 1 SD)
1.77 (1.17, 2.66)
1.65 (0.54, 5.05)
1.81 (1.23, 2.67)
Weight before delivery (per 1 SD)
1.07 (0.37, 3.10)
1 (ref)
1.59 (1.04, 2.44)
1.30 (0.44, 3.90) 0.98
0.93
0.04
OR (95% CI)
Adjusted model 1
1.22 (0.42,3.53)
1 (ref)
1.51 (1.02, 2.23)
p Value
Spectral Analysis in Pregnancy
Other
Black
White
Race/Ethnicity
Age (per 1 SD)
OR (95% CI)
Unadjusted
Table 2 Odds ratio for developing pregnancy-induced hypertension versus being normotensive
1054 Flood et al.
1.44 (0.97, 2.14) 1.81 (1.26, 2.58) 1.00 (0.79, 1.25)
SBP (per 1 SD)
DBP (per 1 SD)
a
1.06 (0.68, 1.65) 0.86 (0.55, 1.35)
LF-BPV (SBP)a 0.51
0.80
0.97
0.001
0.07
0.91
0.01
0.23
0.16
p Value
0.89 (0.56, 1.41)
1.16 (0.74, 1.83)
1.12 (0.79, 1.57)
1.89 (1.30, 2.75)
1.42 (0.95, 2.14)
1.04 (0.68, 1.60)
1.66 (1.10, 2.51)
1.73 (0.58, 5.21)
2.34 (0.88, 6.21)
1 (ref)
1.23 (0.83, 1.85)
OR (95% CI)
Adjusted model 1
0.62
0.52
0.53
0.001
0.09
0.86
0.02
0.21
0.31
p Value
0.99 (0.61, 1.60)
1.19 (0.83, 1.72)
3.15 (1.57, 6.34)
0.53 (0.25, 1.11)
1.16 (0.61, 2.19)
1.89 (1.21, 2.96)
1.55 (0.51, 4.71)
1.84 (0.67, 5.06)
1 (ref)
1.32 (0.85, 2.04)
OR (95% CI)
Adjusted model 2a
0.95
0.35
0.001
0.09
0.65
0.005
0.46
0.22
p Value
0.87 (0.53, 1.43)
1.24 (0.84, 1.82)
3.10 (1.54, 6.25)
0.55 (0.26, 1.16)
1.22 (0.63, 2.34)
1.89 (1.21, 2.96)
1.59 (0.52, 4.87)
1.90 (0.70, 5.20)
1 (ref)
1.33 (0.86, 2.06)
OR (95% CI)
Adjusted model 2b
0.58
0.28
0.002
0.11
0.56
0.006
0.42
0.20
p Value
American Journal of Perinatology
Downloaded by: NYU. Copyrighted material.
Abbreviations: 95% CI, 95% confidence interval; DBP, diastolic blood pressure; HF-HRV, high frequency heart rate variability; LF-BPV, low frequency blood pressure variability; OR, odds ratio; SBP, systolic blood pressure; SD, standard deviation. a Log-transformed variables, with 1 unit representing doubling values. Notes: All models include multiple imputation including age, race, weight before delivery, heart rate, SBP, DBP, HF-HRV, LF-BPV, and hypertensive disorders of pregnancy. Model 1 includes adjustment for age, weight and heart rate. Model 2a includes adjustment for covariates in model 1 plus mean SBP, DBP, HF-HRV, and LF-BPV (estimated using diastolic blood pressure). Model 2b includes adjustment for covariates in model 1 plus mean SBP, DBP, HF-HRV, and LF-BPV (estimated using systolic blood pressure).
LF-BPV (DBP)
a
HF-HRV
0.98 (0.65, 1.47)
Heart rate (per 1 SD)
1.38 (0.47, 4.00)
Other 1.71 (1.14, 2.56)
2.30 (0.89, 5.93)
Black
Weight before delivery (per 1 SD)
1 (ref)
1.31 (0.90, 1.90)
White
Race/Ethnicity
Age (per 1 SD)
OR (95% CI)
Unadjusted
Table 3 Odds ratio for developing preeclampsia versus being normotensive
Spectral Analysis in Pregnancy Flood et al.
Vol. 32
No. 11/2015
1055
Spectral Analysis in Pregnancy
Flood et al.
analysis of HR and BP did not add predictive value for the development of PE.
Comment Given the high prevalence of PIH and PE and the devastating consequences of PE mother and infant, there is substantial interest both in identifying potential early risk factors for and understanding the mechanisms that underlie the development of these syndromes. As expected, based on previous reports, systolic and diastolic BP at 28 weeks’ gestation were associated with the development of PIH and PE. However, the use of mean BP alone to predict subsequent risk of PE and/or PIH is inadequate because of the significant overlap with controls leading to a poor-predictive value17,18 and does not differentiate between PIH and PE. Spectrally defined HF-HRV and LF-BPV enhanced prediction of PIH and allowed differentiation between PIH and PE. In addition, HF-HRV, an index of cardiac parasympathetic modulation, was lower in women who went on to develop PIH compared with those who developed PE or remained normotensive. Our findings support the view that beat-to-beat measurement of BP as well as HR fluctuations yields important information on pregnancy outcomes above and beyond standard BP measurements These findings are novel for several reasons. First, they represent the largest study to demonstrate the clinical value of spectrally defined BPV to obstetrical outcomes. Second, they suggest that measures of cardiovascular autonomic regulation may have predictive power to differentiate PE from PIH at an early stage to facilitate the importantly different clinical management. Lastly, they support the concept that PIH and PE have distinct physiologic mechanisms and autonomic consequences. The finding that measurement of HRV and BPV, measures of autonomic tone, was predictive of PIH but not PE is supportive of the idea that PIH represents a true abnormality in autonomic function, perhaps related to lifetime predilection to hypertension, while PE is inflammatory in nature and results in changes in BP as a downstream effect. Considerable evidence now suggests that BP fluctuations contain important prognostic as well as physiologic information, and may be as clinically important as mean systolic and diastolic BP in health outcomes in adults. Many recent studies in humans and animals have demonstrated that variation in BP exists across a wide time scale, from years to days, hours, and even on a beatto-beat basis, and in most, but not all, studies, this variability is associated with risk of cardiovascular events independent of mean BP.1,19–27 Differences in methods of measurement of BPV are likely to be at least partially responsible for these contrasting findings.28 In addition, some evidence now suggests the possibility of therapeutic benefits from reduction in BPV independent of changes in mean pressure.25,29,30 A role for the involvement of the sympathetic nervous system in PE is suggested by a small treatment study comparing women at low and high risk of PE during the first or early second trimester before the onset of symptoms.31 High-risk women were more likely to have a “hyperdynamic” circulation, characterized by elevated cardiac output and reduced peripheral vascular resistance. These high-risk women were American Journal of Perinatology
Vol. 32
No. 11/2015
randomized to treatment with atenolol or placebo and active treatment significantly reduced the development of PE (3.8% in the atenolol group; 18% in the placebo group), further supporting the involvement of the sympathetic nervous system in PE. This study likely underestimated the extent of autonomic dysfunction as high-risk subjects with the highest levels of cardiac output were directed to a treatment arm of the study. Four small observational studies have examined the relationship between spectral analysis of HR and BP in pregnancy and hypertensive diseases of pregnancy. Two of these studies supported our findings. Silver et al measured baroreflex sensitivity (BRS), an index of reflexive cardiac autonomic control after the onset of disease, in 20 women with PE and 20 women with PIH compared with 20 gestational age-matched controls and 20 control subjects who were not pregnant.32 BRS was reduced in normal pregnancy compared with nonpregnant women and was still further reduced in patients with hypertensive disorders of pregnancy suggesting a pathological abnormality in parasympathetic activity. Yang et al found that preeclamptic women ( 34 weeks) had lower HF-HRV compared with those who were normotensive.4 Two other small studies partially conflict with our findings. Ekholm et al found that women with PIH had higher levels of HF-HRV at the time of their diagnosis compared with women with uncomplicated pregnancies.33 However, they also studied midfrequency BPV and reported that higher levels were more common in hypertensive pregnant women. Midfrequency BPV (range, 0.07–0.15 Hz) overlaps with the LF frequency band evaluated in our study (range, 0.04–0.15 Hz) and as such, this finding is consistent with ours. Finally, Greenwood et al, using a different approach to assessment of ANS function (microneurography) showed that women with PIH had elevated levels of sympathetic nervous system activity compared with women with normal pregnancies.34 All of these studies were limited by small sample size and the fact that the indices of autonomic regulation were evaluated in women who already had been diagnosed with hypertensive disorders of pregnancy. Only one small study has examined the baseline ANS function prospectively. In a sample of 42 women studied on several occasions during pregnancy, Rang et al found no difference in LF-BPV or in BRS between women who went on to develop PE or remained normotensive,35 suggesting no role for the BPV in PE. Our data showing a lack of association between BPV and PE are consistent with this finding. Our study has several strengths. First, the sample size is large for studies of beat-to-beat BPV. Second, the use of military women as subjects provides for homogeneity with respect to many potentially confounding variables including physical health, employment status, health behaviors, and access to prenatal care. Limitations include the fact that not all subjects underwent psychophysiological testing. Women who participated in psychophysiological testing were similar to those who did not in age, education, and military rank but there was a statistically significant increase in risk of preterm delivery in the women who did not participate. Nonparticipants also were more likely to report job-related stress and were somewhat more likely to be black.14 PIH and PE were diagnosed on the basis of clinical criteria at the time. It is
Downloaded by: NYU. Copyrighted material.
1056
Spectral Analysis in Pregnancy
5 Stauss HM. Identification of blood pressure control mechanisms
6
7
8
9
10
11
12
13
14
15 16
Conflicts of Interest The authors report no conflicts of interest.
17
18
Note The cohort was assembled at Lackland Air force Base, San Antonio, TX when Dr. Hatch was a faculty member at Mt. Sinai Hospital, New York, NY. Reprints will not be available
19
20
21
References 1 Macdonald-Wallis C, Lawlor DA, Fraser A, May M, Nelson SM,
Tilling K. Blood pressure change in normotensive, gestational hypertensive, preeclamptic, and essential hypertensive pregnancies. Hypertension 2012;59(6):1241–1248 2 Berntson GG, Bigger JT Jr, Eckberg DL, et al. Heart rate variability: origins, methods, and interpretive caveats. Psychophysiology 1997;34(6):623–648 3 Berntson GG, Hart S, Ruland S, Sarter M. A central cholinergic link in the cardiovascular effects of the benzodiazepine receptor partial inverse agonist FG 7142. Behav Brain Res 1996;74(1-2): 91–103 4 Yang CC, Chao TC, Kuo TB, Yin CS, Chen HI. Preeclamptic pregnancy is associated with increased sympathetic and decreased parasympathetic control of HR. Am J Physiol Heart Circ Physiol 2000; 278(4):H1269–H1273
1057
22
23
24 25
by power spectral analysis. Clin Exp Pharmacol Physiol 2007; 34(4):362–368 deBoer RW, Karemaker JM, Strackee J. Hemodynamic fluctuations and baroreflex sensitivity in humans: a beat-to-beat model. Am J Physiol 1987;253(3 Pt 2):H680–H689 Montano N, Lombardi F, Gnecchi Ruscone T, et al. Spectral analysis of sympathetic discharge, R-R interval and systolic arterial pressure in decerebrate cats. J Auton Nerv Syst 1992;40(1):21–31 Dimier-David L, Billon N, Costagliola D, Jaillon P, Funck-Brentano C. Reproducibility of non-invasive measurement and of short-term variability of blood pressure and heart rate in healthy volunteers. Br J Clin Pharmacol 1994;38(2):109–115 Cevese A, Grasso R, Poltronieri R, Schena F. Vascular resistance and arterial pressure low-frequency oscillations in the anesthetized dog. Am J Physiol 1995;268(1 Pt 2):H7–H16 Mukai S, Hayano J. Heart rate and blood pressure variabilities during graded head-up tilt. J Appl Physiol (1985) 1995;78(1): 212–216 Cooke WH, Hoag JB, Crossman AA, Kuusela TA, Tahvanainen KU, Eckberg DL. Human responses to upright tilt: a window on central autonomic integration. J Physiol 1999;517(Pt 2):617–628 Schächinger H, Weinbacher M, Kiss A, Ritz R, Langewitz W. Cardiovascular indices of peripheral and central sympathetic activation. Psychosom Med 2001;63(5):788–796 Ditor DS, Kamath MV, Macdonald MJ, Bugaresti J, McCartney N, Hicks AL. Reproducibility of heart rate variability and blood pressure variability in individuals with spinal cord injury. Clin Auton Res 2005;15(6):387–393 Hatch M, Berkowitz G, Janevic T, et al. Race, cardiovascular reactivity, and preterm delivery among active-duty military women. Epidemiology 2006;17(2):178–182 Harris FJ. On the use of windows for harmonic analysis with discrete Fourier transorm. Proc IEEE 1978;66(1):51–83 Barros AJ, Hirakata VN. Alternatives for logistic regression in crosssectional studies: an empirical comparison of models that directly estimate the prevalence ratio. BMC Med Res Methodol 2003;3:21 Begum KS, Sachchithanantham K, De Somsubhra S. Maternal obesity and pregnancy outcome. Clin Exp Obstet Gynecol 2011; 38(1):14–20 Salihu HM, De La Cruz C, Rahman S, August EM. Does maternal obesity cause preeclampsia? A systematic review of the evidence. Minerva Ginecol 2012;64(4):259–280 Parati G, Grassi G, Coruzzi P, et al. Influence of cardiopulmonary receptors on the bradycardic responses to carotid baroreceptor stimulation in man. Clin Sci (Lond) 1987;72(5):639–645 Frattola A, Parati G, Cuspidi C, Albini F, Mancia G. Prognostic value of 24-hour blood pressure variability. J Hypertens 1993;11(10): 1133–1137 Sander D, Kukla C, Klingelhöfer J, Winbeck K, Conrad B. Relationship between circadian blood pressure patterns and progression of early carotid atherosclerosis: A 3-year follow-up study. Circulation 2000;102(13):1536–1541 Mancia G, Parati G, Hennig M, et al; ELSA Investigators. Relation between blood pressure variability and carotid artery damage in hypertension: baseline data from the European Lacidipine Study on Atherosclerosis (ELSA). J Hypertens 2001;19(11):1981–1989 Mancia G, Bousquet P, Elghozi JL, et al. The sympathetic nervous system and the metabolic syndrome. J Hypertens 2007;25(5): 909–920 Mancia G. Blood pressure reduction and cardiovascular outcomes: past, present, and future. Am J Cardiol 2007;100(3A):3J–9J Xie HH, Zhang XF, Chen YY, Shen FM, Su DF. Synergism of hydrochlorothiazide and nifedipine on blood pressure variability reduction and organ protection in spontaneously hypertensive rats. Hypertens Res 2008;31(4):685–691
American Journal of Perinatology
Vol. 32
No. 11/2015
Downloaded by: NYU. Copyrighted material.
possible, given current criteria that there may have been misclassification. BPV and HRV are currently research tools. The measurements are simple, noninvasive, and require a short testing period ( 5 minutes of rest followed by 5 minutes of testing). However, in order for these techniques to be clinically useful, they would have to be adapted by industry for clinical use. The strength of a uniform population is also a weakness. This association needs to be replicated in a more normal clinical cohort that would be expected to have greater variability in maternal weight and overall fitness. Another weakness is the overall infrequency of hypertensive disease in this fit population. One potential utility of this noninvasive technique would be to select a population at risk for future study. Testing occurred at 28 weeks’ gestation, assuring that the most severely affected patients would not be part of the study group. Future studies, measuring these indices at 22 to 24 weeks before the usual onset of signs and symptoms would assure the inclusion of the most severe forms of hypertensive disease in pregnancy. In conclusion, 5-minute estimates of LF-BPV and HF-HRV collected at the beginning of the third trimester of pregnancy allow for significant improvement in the ability to identify which patients will develop PIH. These associations were independent and additive to values of BP and maternal weight. Inclusion of these factors into risk stratification for PIH may enhance sensitivity and specificity of prediction for treatment of inclusion in clinical trials. The early identification of subjects at risk has the potential to reduce costs for cohort analyses and would allow for relatively early intervention.
Flood et al.
Spectral Analysis in Pregnancy
Flood et al.
26 Rothwell PM. Limitations of the usual blood-pressure hypothesis
31 Carr DB, McDonald GB, Brateng D, Desai M, Thach CT, Easterling TR.
and importance of variability, instability, and episodic hypertension. Lancet 2010;375(9718):938–948 Muntner P, Shimbo D, Tonelli M, Reynolds K, Arnett DK, Oparil S. The relationship between visit-to-visit variability in systolic blood pressure and all-cause mortality in the general population: findings from NHANES III, 1988 to 1994. Hypertension 2011;57(2): 160–166 Mena L, Pintos S, Queipo NV, Aizpúrua JA, Maestre G, Sulbarán T. A reliable index for the prognostic significance of blood pressure variability. J Hypertens 2005;23(3):505–511 Schillaci G, Pucci G, Parati G. Blood pressure variability: an additional target for antihypertensive treatment? Hypertension 2011;58(2):133–135 Zhang Y, Agnoletti D, Safar ME, Blacher J. Effect of antihypertensive agents on blood pressure variability: the Natrilix SR versus candesartan and amlodipine in the reduction of systolic blood pressure in hypertensive patients (X-CELLENT) study. Hypertension 2011;58(2):155–160
The relationship between hemodynamics and inflammatory activation in women at risk for preeclampsia. Obstet Gynecol 2001;98(6): 1109–1116 Silver HM, Tahvanainen KU, Kuusela TA, Eckberg DL. Comparison of vagal baroreflex function in nonpregnant women and in women with normal pregnancy, preeclampsia, or gestational hypertension. Am J Obstet Gynecol 2001;184(6): 1189–1195 Ekholm EM, Tahvanainen KU, Metsälä T. Heart rate and blood pressure variabilities are increased in pregnancyinduced hypertension. Am J Obstet Gynecol 1997;177(5): 1208–1212 Greenwood JP, Stoker JB, Walker JJ, Mary DA. Sympathetic nerve discharge in normal pregnancy and pregnancy-induced hypertension. J Hypertens 1998;16(5):617–624 Rang S, Wolf H, van Montfrans GA, Karemaker JM. Serial assessment of cardiovascular control shows early signs of developing pre-eclampsia. J Hypertens 2004;22(2):369–376
27
28
29
30
32
33
34
35
Downloaded by: NYU. Copyrighted material.
1058
American Journal of Perinatology
Vol. 32
No. 11/2015
Beat-to-Beat Heart Rate and Blood Pressure Variability and Hypertensive Disease in Pregnancy Pamela Flood, MD, MA1 Paula McKinley, PhD2 Catherine Monk, PhD3 Paul Muntner, PhD4 Lisandro D. Colantonio, MD, MSc4 Laura Goetzl, MD, MPH5 Maureen Hatch, PhD6 Richard P. Sloan, PhD2 1 Department of Anesthesia, Perioperative and Pain Medicine, Stanford
University, San Francisco, California 2 Department of Behavioral Medicine, Columbia University, New York, New York 3 Department of Psychiatry, Behavioral Medicine and Developmental Neuroscience, Columbia University, New York, New York 4 Department of Epidemiology, University of Alabama at Birmingham, Birmingham, Alabama 5 Department of Obstetrics and Gynecology, Temple University, Philadelphia, Pennsylvania 6 National Institutes of Health/National Cancer Institute, Division of Cancer Epidemiology & Genetics, Radiation Epidemiology Branch Bethesda, Maryland
Address for correspondence Pamela Flood, MD, MA, Department of Anesthesia, Perioperative and Pain Medicine, Stanford University, 300 Pasteur Drive, Palo Alto, CA 94304 (e-mail: pfl[email protected]).
Am J Perinatol 2015;32:1050–1058.
Abstract
Keywords
► hypertensive diseases of pregnancy ► autonomic function ► heart rate variability ► blood pressure variability
received November 30, 2014 accepted after revision January 30, 2015 published online May 13, 2015
Objective The aim of this study is to determine the relationship between heart rate and/or blood pressure variability, measured at 28 weeks’ gestation, and the incidence of pregnancy-induced hypertension or preeclampsia. Study Design Secondary analysis of data from a prospectively enrolled cohort of 385 active military women in whom spectral analysis of continuous heart rate and variability was measured at 28 weeks’ gestation. The primary outcome was the predictive value of spectral analysis of heart rate and blood pressure for hypertensive diseases of pregnancy. Results High-frequency heart rate variability was reduced and low-frequency variability of systolic and diastolic blood pressure increased in women who would develop pregnancy-induced hypertension but not preeclampsia. Low-frequency variability of diastolic blood pressure remained a significant predictor of pregnancy-induced hypertension but not preeclampsia after adjustment for age, weight, and blood pressure in a multivariate model. Conclusion Early identification of pregnancy-induced hypertension can facilitate treatment to avoid maternal morbidity. Understanding the physiological underpinnings of the two very different diseases may lead to improved treatment and prevention. If proven effective in a broader population, the ability to differentiate pregnancy-induced hypertension from preeclampsia may reduce unnecessary iatrogenic interventions or inappropriate preterm delivery.
Copyright © 2015 by Thieme Medical Publishers, Inc., 333 Seventh Avenue, New York, NY 10001, USA. Tel: +1(212) 584-4662.
DOI http://dx.doi.org/ 10.1055/s-0035-1548542. ISSN 0735-1631.
Downloaded by: NYU. Copyrighted material.
1050
Hypertensive disorders of pregnancy, including preeclampsia (PE) and pregnancy-induced hypertension (PIH), affect approximately 10 to 15% of pregnancies and are a significant underlying factor in severe maternal and neonatal morbidity and mortality.1 Although preventive interventions are still lacking, accurate noninvasive, low cost early predictors for these disorders may improve maternal and fetal outcomes through increased surveillance and early treatment of highrisk populations. Furthermore, identification of markers associated with increased risk for PE and PIH will be useful for designing trials of therapeutic interventions. One potential method for early detection of hypertensive disease during pregnancy focuses on associated dysregulation of the autonomic nervous system (ANS). Two of the following noninvasive indices of ANS dysregulation are of special interest: heart rate variability (HRV) and blood pressure variability (BPV). HRV, especially at high frequency (HF) (range, 0.15–0.40 Hz), is an index of cardiac parasympathetic modulation.2,3 In previous studies, women with normal pregnancy (at 34 weeks) had lower HF-HRV compared with control women who were not pregnant. In women who would develop PE, HF-HRV was lower still.4 These data suggest that cardiac autonomic dysfunction may be an early marker of pathophysiology in women who go on to develop hypertensive disorders of pregnancy. Like heart rate (HR), blood pressure (BP) exhibits periodic oscillations at high and low frequencies (range, 0.04–0.15 Hz, LF).5 While HF-BPV is a direct product of respiratory-induced mechanical fluctuations of cardiac output which change arterial pressure,6 some studies suggest that LF-BPV reflects vascular sympathetic drive.7–13 A careful observational study of the natural course and outcome of pregnant patients with autonomic dysfunction has the potential to identify early markers of hypertensive disorders of pregnancy before the disease onset and thereby contribute to the determination of the underlying pathophysiology and development of clinical interventions. Here, we test the hypothesis that ANS function, measured at 28 weeks’ gestation, predicts later development of PE and PIH in a large cohort of military women.
Patients and Methods Subjects From 1996 to 2001, data were collected from pregnant military women stationed at Lackland Air Force Base in San Antonio, Texas for a study of “Cardiovascular responsivity, physical and psychosocial job stress, and the risk of preterm delivery” (Maureen Hatch, PI) that was funded by the Department of Defense. Results addressing the primary aim have been published.14 As part of the protocol, HR and BP variability were measured in a subset of study participants to determine if responsiveness to psychological stress predicted preterm delivery. The study protocol was approved by the relevant institutional review boards. All participants signed informed consent for the original study. Deidentified data from that project were used in this analysis. Here, we report on HRV and BPV data from the resting baseline period that preceded psychological testing in the original study, and
Flood et al.
investigate their relationship to the development of hypertensive disease of pregnancy. A total of 724 pregnant active-duty military women who attended the prenatal clinic between 1997 and 2000 were enrolled early in pregnancy (6–12 weeks of gestation). Overall, 80 women had pregnancy loss or delivery before 28 weeks’ gestation and 63 left the military base and were considered lost to follow-up. Some deliveries before 28 weeks were secondary to hypertensive complications of pregnancy; eliminating early-onset PIH or PET from the study population. Overall, 581 subjects returned for a 28-week assessment visit that included measurement of HRV and BPV and 385 completed the psychophysiology test. Data from these subjects were used in this secondary data analysis. Most women who declined to participate in the psychophysiology study indicated that they did not have time to stay for the test. None of the subjects had a current clinical diagnosis of PIH or PE at the time of autonomic testing. Overall, 21 subjects reported an episode of hypertension before their pregnancy and 4 subjects had required medication to control their BP. Medical data including diagnoses of PIH and PE were abstracted from hospital records by Hatch for the original study; PIH and PE were based on treating physician diagnosis using standard clinical definitions as recommended by ACOG at the time of treatment. Gestational age was based on the due date estimated by ultrasound measurements before 20 weeks’ gestation. Maternal weight was recorded before pregnancy and at delivery.
Measurement of Heart Rate Variability Participants were tested in the seated position. Electrodes were placed on the right shoulder, on the left anterior axillary line at the 10th intercostal space, and in the right lower quadrant for standard three-lead ECG measurement. After a 5-minute resting period, analog electrocardiogram signals were digitized at 500 Hz by a 12-bit analog-to-digital (A/D) conversion board and passed to a microcomputer. The electrocardiogram waveform was submitted to an R wave detection routine implemented by custom-written event detection software, resulting in an RR interval series. Errors in marking of R waves were corrected by visual inspection. Ectopic beats were corrected by interpolation. HF-HRV (range, 0.15–0.50 Hz) from 300-second epochs were computed with an interval method for computing Fourier transforms similar to that described by deBoer et al.6 Before computing Fourier transforms, the mean of the RR interval series was subtracted from each value in the series, then filtered the series with a Hanning window15 and summed the power (i.e., variance, in msec2) over the HF band. Estimates of spectral power were adjusted to account for attenuation produced by this filter.15
Measurement of Blood Pressure and Blood Pressure Variability The beat-to-beat BP waveform was obtained using a Finapres noninvasive BP monitor (Finapres Medical Systems, Amsterdam, The Netherlands) with a finger cuff placed on the middle finger of the nondominant hand during the same session as American Journal of Perinatology
Vol. 32
No. 11/2015
1051
Downloaded by: NYU. Copyrighted material.
Spectral Analysis in Pregnancy
Spectral Analysis in Pregnancy
Flood et al.
HRV measurement earlier. The analog BP waveform also was sampled by the A/D board at 500 samples per second. Systolic and diastolic BP values for each cardiac cycle were identified using custom-written software resulting in a BP time series. Mean BP and spectral power in the low (range, 0.04–0.15 Hz) frequency band of the BP power spectrum for both systolic BP and diastolic BP were computed from these time series. Because the servo adjustment of the Finapres monitor was enabled during the last minute of each 300 second period, spectra were calculated on 240 second epochs using the method described earlier. Power, that is, variance (in mm Hg2), over the LF band was summed and adjusted to account for attenuation produced by the Hanning window.15
Statistical Analyses HF-HRV and LF-BPV data were log-transformed before the analysis as the data were not normally distributed (skewed to the right). Characteristics of women who developed PIH or PE versus those who remained normotensive were compared using t-test or Fisher exact test for continuous and categorical variables, respectively. Unadjusted odds ratios and 95% confidence intervals (95% CI) for developing PIH or PE, separately, versus remaining normotensive associated with age, race, HR, mean systolic and diastolic BP, HF-HRV, and LF-HRV were estimated using logistic regression. In addition, two logistic regression models including progressive adjustment were conducted. The first model included adjustment for age, weight, and HR. The second (fully adjusted) model included adjustment for variables included in the first model plus mean systolic and diastolic BP, HF-HRV, and LF-BPV. Different fully adjusted models were conducted for diastolic or systolic LF-BPV.
All regression models were conducted after performing multiple imputations for missing data. Missing data were imputed for age (n ¼ 6), race (n ¼ 6), HR (n ¼ 16), predelivery weight (n ¼ 6), and HF-HRV (n ¼ 16). No data were missing on hypertensive disorders of pregnancy, mean systolic or diastolic BP, or LF-BPV. For the multiple imputations, 10 data sets using chained equations were assembled based on observed values from all the variables collected and the occurrence of hypertensive disorders of pregnancy.16 All analyses were conducted using Stata/I.C. 13.0 (Stata Corporation, College Station, TX).
Results Of the 385 women who underwent psychophysiological testing, 26 developed PIH, 27 developed PE, and 332 did not develop hypertensive disease of pregnancy. Among the 21 women who reported an episode of hypertension before pregnancy, 12 went on to be diagnosed with PIH and 6 PE and 2 subjects did not become hypertensive during their pregnancy, and 1 was lost to follow-up. Demographic variables for subjects by group are shown in ►Table 1. Women who would go on to develop PIH were older (p < 0.04) and heavier, both before pregnancy and at the time of delivery (p < 0.01, 0.002), compared with normotensive subjects (►Table 1). Maternal weight before pregnancy and at delivery (p < 0.02, 0.01) was also higher in those who developed PE compared with normotensive women. There were no racial differences between women who developed PE or PIH and those who did not in this cohort. There was no difference in neonatal weight or gestational age at birth for infants born to women with PIH. Neonates born to mothers with PE were smaller (2.98 0.84
Table 1 Baseline characteristics of participants included in the analysis by presence of hypertensive disorders of pregnancya Participants included (N ¼ 385) Age (y), mean (SD) Race/Ethnicity, (%)
NT (n ¼ 332)
PE (n ¼ 27)
p Value
PIH (n ¼ 26)
p Value
26.4 (5.1)
27.9 (6.3)
0.18
28.7 (5.7)
0.04
0.19
58.4
0.87
b
White
62.1
48.0
Black
17.9
32.0
20.8
Other
20.0
20.0
20.8
135.1 (18.3)
144.8 (18.4)
Weight prepregnancy (lbs.), mean (SD)
0.02
145.7 (23.4)
0.01
Weight before delivery (lbs.), mean (SD)
175.2 (23.5)
188.2 (25.3)
0.01
190.3 (28.9)
0.002
HR (bpm), mean (SD)c
86.7 (10.4)
86.7 (8.7)
0.98
90.0 (11.6)
0.12
c
108.3 (11.1)
112.5 (12.5)
0.07
119.1 (15.9)
< 0.001
c
58.6 (8.2)
64.3 (10.2)
< 0.001
64.8 (11.7)
< 0.001
358 (314, 408)
363 (197, 668)
0.95
206 (116, 365)
0.03
c
9.7 (9.1, 10.3)
8.9 (7.0, 11.3)
0.51
12.7 (9.5, 17.0)
0.03
c
4.1 (3.9, 4.4)
4.3 (3.5, 5.3)
0.80
6.2 (4.8, 8.0)
0.002
SBP (mmHg), mean (SD)
DBP (mmHg), mean (SD)
HF-HRV, geometric mean (95% CI)c LF-BPV (SBP), geometric mean (95% CI)
LF-BPV (DBP), geometric mean (95% CI)
Abbreviations: 95% CI, 95% confidence interval; bpm, beats per minute; HF-HRV, high frequency heart rate variability; LF-BPV, low frequency blood pressure variability; NT, normotensive; PE, preeclampsia; PIH, pregnancy-induced hypertension. a Nonimputed data. b Fisher exact test. c Blood pressure and heart rate were measured at 28 weeks’ gestation—the time of psychophysiological testing. American Journal of Perinatology
Vol. 32
No. 11/2015
Downloaded by: NYU. Copyrighted material.
1052
Spectral Analysis in Pregnancy
Flood et al.
vs. 3.41 0.50 g, p < 0.00001) and more premature (38 3 vs. 39 2 weeks, p < 0.002) than the neonates born to normotensive women. At 28 weeks’ gestation, there was no difference in resting HR in women who developed PE or PIH compared with those who remained normotensive (►Table 1). In women who would go on to develop PIH, both mean systolic (p < 0.001) and diastolic BP (p < 0.001) were higher (►Fig. 1). Diastolic BP was higher in women who would develop PE (p < 0.001) compared with those who remained normotensive. During psychophysiological testing at 28 weeks’ gestation, only three subjects (one in each group) had diastolic BP values that if repeated and correlated with other clinical criteria would be diagnostic for PIH or chronic hypertension (greater than 90 mm Hg). Six subjects had a systolic BP over 140 at psychophysiological testing, three in the NT, one in the PIH, and two subjects in the PE group. In women who went on to develop PIH, HF-HRV was reduced (►Table 1, ►Fig. 2, p < 0.03) and LF-BPV for both systolic and diastolic BP was increased (►Table 1, ►Fig. 3a, b; p < 0.03 and 0.002, respectively) relative to women with normal pregnancies. Neither HRV nor BPV were predictive of the development of PE. A multivariate model for prediction of PIH was created adjusted for age, weight, and resting HR (►Table 2, adjusted model 1). Model 2a including items in model 1 with additional adjustment for SBP, DBP, HR-HFV, and LF-BPV estimated from systolic BP demonstrated the importance of weight before delivery and LF-BPV derived from the systolic BP time series. Alternative model 2b including additional adjustment for SBP, DBP, HF-HRV, and LF-BPV estimated
from diastolic BP suggested that only maternal weight contributed to the final model. As such, only LF-BPV estimated from systolic BP added significant predictive power to mean systolic and diastolic BP for the prediction of PIH. The multivariate model for prediction of PE was adjusted for age, weight, and resting HR (►Table 3, adjusted model 1). Model 2a including items in model 1 with additional adjustment for SBP, DBP, HR-HFV, and LF-BPV estimated from systolic BP showed a significant contribution from weight before delivery and diastolic BP at 28 weeks’ gestation. Alternative model 2b including additional adjustment for SBP, DBP, HF-HRV, and LF-BPV estimated from diastolic BP was similar. Spectral
Fig. 1 Systolic and diastolic blood pressure distributions by presence of hypertensive disorders during pregnancy. DBP, diastolic blood pressure; NT, normotensive; PE, preeclampsia; PIH, pregnancy-induced hypertension; SBP, systolic blood pressure.
Fig. 3 Low-frequency systolic and diastolic blood pressure variability distributions by presence of hypertensive disorders during pregnancy. DBP, diastolic blood pressure; LF-BPV, low-frequency blood pressure; NT, normotensive; PE, preeclampsia; PIH, pregnancy-induced hypertension; SBP, systolic blood pressure.
1053
American Journal of Perinatology
Vol. 32
No. 11/2015
Downloaded by: NYU. Copyrighted material.
Fig. 2 High-frequency heart rate variability (HF-HRV) distribution by presence of hypertensive disorders during pregnancy. NT, normotensive; PE, preeclampsia; PIH, pregnancy-induced hypertension.
American Journal of Perinatology
Vol. 32
No. 11/2015
2.07 (1.30, 3.32) 1.64 (1.05, 2.59)
LF-BPV (DBP)a
a
0.03
0.002
0.03 1.65 (1.04, 2.62)
2.27 (1.39, 3.68)
0.90 (0.63, 1.29)
1.93 (1.31, 2.84)
2.20 (1.46, 3.31)
< 0.001 0.001
1.53 (1.02, 2.30)
0.13
0.03
0.001
0.56
0.001
< 0.001
0.04
0.006
0.66
0.03
p Value
1.81 (1.07, 3.06)
0.80 (0.55, 1.18)
0.92 (0.45, 1.90)
1.99 (0.95, 4.16)
1.13 (0.59, 2.19)
1.78 (1.15, 2.74)
2.34 (0.72, 7.61)
1.57 (0.48, 5.19)
1 (ref)
1.43 (0.90, 2.26)
OR (95% CI)
Adjusted model 2a
0.03
0.26
0.83
0.07
0.71
0.01
0.35
0.13
p Value
1.39 (0.85, 2.27)
0.84 (0.57, 1.22)
1.08 (0.53, 2.21)
1.96 (0.93, 4.12)
1.16 (0.59, 2.26)
1.77 (1.15, 2.72)
2.17 (0.67, 6.98)
1.57 (0.48, 5.14)
1 (ref)
1.39 (0.88, 2.18)
OR (95%CI)
Adjusted model 2b
0.19
0.35
0.83
0.08
0.67
0.009
0.41
0.16
p Value
Downloaded by: NYU. Copyrighted material.
Abbreviations: 95% CI, 95% confidence interval; DBP, diastolic blood pressure; HF-HRV, high frequency heart rate variability; LF-BPV, low frequency blood pressure variability; OR, odds ratio; SBP, systolic blood pressure; SD, standard deviation. a Log-transformed variables, with 1 unit representing doubling values. Notes: All models include multiple imputation including age, race, weight before delivery, heart rate, SBP, DBP, HF-HRV, LF-BPV, and hypertensive disorders of pregnancy. Model 1 includes adjustment for age, weight, and heart rate. Model 2a includes adjustment for covariates in model 1 plus mean SBP, DBP, HF-HRV, and LF-BPV (estimated using diastolic blood pressure). Model 2b includes adjustment for covariates in model 1 plus mean SBP, DBP, HF-HRV, and LF-BPV (estimated using systolic blood pressure).
LF-BPV (SBP)
1.84 (1.29, 2.62) 0.77 (0.61, 0.97)
HF-HRVa
2.18 (1.49, 3.20)
SBP (per 1 SD)
DBP (per 1 SD)
1.34 (0.92, 1.96)
Heart rate (per 1 SD)
1.77 (1.17, 2.66)
1.65 (0.54, 5.05)
1.81 (1.23, 2.67)
Weight before delivery (per 1 SD)
1.07 (0.37, 3.10)
1 (ref)
1.59 (1.04, 2.44)
1.30 (0.44, 3.90) 0.98
0.93
0.04
OR (95% CI)
Adjusted model 1
1.22 (0.42,3.53)
1 (ref)
1.51 (1.02, 2.23)
p Value
Spectral Analysis in Pregnancy
Other
Black
White
Race/Ethnicity
Age (per 1 SD)
OR (95% CI)
Unadjusted
Table 2 Odds ratio for developing pregnancy-induced hypertension versus being normotensive
1054 Flood et al.
1.44 (0.97, 2.14) 1.81 (1.26, 2.58) 1.00 (0.79, 1.25)
SBP (per 1 SD)
DBP (per 1 SD)
a
1.06 (0.68, 1.65) 0.86 (0.55, 1.35)
LF-BPV (SBP)a 0.51
0.80
0.97
0.001
0.07
0.91
0.01
0.23
0.16
p Value
0.89 (0.56, 1.41)
1.16 (0.74, 1.83)
1.12 (0.79, 1.57)
1.89 (1.30, 2.75)
1.42 (0.95, 2.14)
1.04 (0.68, 1.60)
1.66 (1.10, 2.51)
1.73 (0.58, 5.21)
2.34 (0.88, 6.21)
1 (ref)
1.23 (0.83, 1.85)
OR (95% CI)
Adjusted model 1
0.62
0.52
0.53
0.001
0.09
0.86
0.02
0.21
0.31
p Value
0.99 (0.61, 1.60)
1.19 (0.83, 1.72)
3.15 (1.57, 6.34)
0.53 (0.25, 1.11)
1.16 (0.61, 2.19)
1.89 (1.21, 2.96)
1.55 (0.51, 4.71)
1.84 (0.67, 5.06)
1 (ref)
1.32 (0.85, 2.04)
OR (95% CI)
Adjusted model 2a
0.95
0.35
0.001
0.09
0.65
0.005
0.46
0.22
p Value
0.87 (0.53, 1.43)
1.24 (0.84, 1.82)
3.10 (1.54, 6.25)
0.55 (0.26, 1.16)
1.22 (0.63, 2.34)
1.89 (1.21, 2.96)
1.59 (0.52, 4.87)
1.90 (0.70, 5.20)
1 (ref)
1.33 (0.86, 2.06)
OR (95% CI)
Adjusted model 2b
0.58
0.28
0.002
0.11
0.56
0.006
0.42
0.20
p Value
American Journal of Perinatology
Downloaded by: NYU. Copyrighted material.
Abbreviations: 95% CI, 95% confidence interval; DBP, diastolic blood pressure; HF-HRV, high frequency heart rate variability; LF-BPV, low frequency blood pressure variability; OR, odds ratio; SBP, systolic blood pressure; SD, standard deviation. a Log-transformed variables, with 1 unit representing doubling values. Notes: All models include multiple imputation including age, race, weight before delivery, heart rate, SBP, DBP, HF-HRV, LF-BPV, and hypertensive disorders of pregnancy. Model 1 includes adjustment for age, weight and heart rate. Model 2a includes adjustment for covariates in model 1 plus mean SBP, DBP, HF-HRV, and LF-BPV (estimated using diastolic blood pressure). Model 2b includes adjustment for covariates in model 1 plus mean SBP, DBP, HF-HRV, and LF-BPV (estimated using systolic blood pressure).
LF-BPV (DBP)
a
HF-HRV
0.98 (0.65, 1.47)
Heart rate (per 1 SD)
1.38 (0.47, 4.00)
Other 1.71 (1.14, 2.56)
2.30 (0.89, 5.93)
Black
Weight before delivery (per 1 SD)
1 (ref)
1.31 (0.90, 1.90)
White
Race/Ethnicity
Age (per 1 SD)
OR (95% CI)
Unadjusted
Table 3 Odds ratio for developing preeclampsia versus being normotensive
Spectral Analysis in Pregnancy Flood et al.
Vol. 32
No. 11/2015
1055
Spectral Analysis in Pregnancy
Flood et al.
analysis of HR and BP did not add predictive value for the development of PE.
Comment Given the high prevalence of PIH and PE and the devastating consequences of PE mother and infant, there is substantial interest both in identifying potential early risk factors for and understanding the mechanisms that underlie the development of these syndromes. As expected, based on previous reports, systolic and diastolic BP at 28 weeks’ gestation were associated with the development of PIH and PE. However, the use of mean BP alone to predict subsequent risk of PE and/or PIH is inadequate because of the significant overlap with controls leading to a poor-predictive value17,18 and does not differentiate between PIH and PE. Spectrally defined HF-HRV and LF-BPV enhanced prediction of PIH and allowed differentiation between PIH and PE. In addition, HF-HRV, an index of cardiac parasympathetic modulation, was lower in women who went on to develop PIH compared with those who developed PE or remained normotensive. Our findings support the view that beat-to-beat measurement of BP as well as HR fluctuations yields important information on pregnancy outcomes above and beyond standard BP measurements These findings are novel for several reasons. First, they represent the largest study to demonstrate the clinical value of spectrally defined BPV to obstetrical outcomes. Second, they suggest that measures of cardiovascular autonomic regulation may have predictive power to differentiate PE from PIH at an early stage to facilitate the importantly different clinical management. Lastly, they support the concept that PIH and PE have distinct physiologic mechanisms and autonomic consequences. The finding that measurement of HRV and BPV, measures of autonomic tone, was predictive of PIH but not PE is supportive of the idea that PIH represents a true abnormality in autonomic function, perhaps related to lifetime predilection to hypertension, while PE is inflammatory in nature and results in changes in BP as a downstream effect. Considerable evidence now suggests that BP fluctuations contain important prognostic as well as physiologic information, and may be as clinically important as mean systolic and diastolic BP in health outcomes in adults. Many recent studies in humans and animals have demonstrated that variation in BP exists across a wide time scale, from years to days, hours, and even on a beatto-beat basis, and in most, but not all, studies, this variability is associated with risk of cardiovascular events independent of mean BP.1,19–27 Differences in methods of measurement of BPV are likely to be at least partially responsible for these contrasting findings.28 In addition, some evidence now suggests the possibility of therapeutic benefits from reduction in BPV independent of changes in mean pressure.25,29,30 A role for the involvement of the sympathetic nervous system in PE is suggested by a small treatment study comparing women at low and high risk of PE during the first or early second trimester before the onset of symptoms.31 High-risk women were more likely to have a “hyperdynamic” circulation, characterized by elevated cardiac output and reduced peripheral vascular resistance. These high-risk women were American Journal of Perinatology
Vol. 32
No. 11/2015
randomized to treatment with atenolol or placebo and active treatment significantly reduced the development of PE (3.8% in the atenolol group; 18% in the placebo group), further supporting the involvement of the sympathetic nervous system in PE. This study likely underestimated the extent of autonomic dysfunction as high-risk subjects with the highest levels of cardiac output were directed to a treatment arm of the study. Four small observational studies have examined the relationship between spectral analysis of HR and BP in pregnancy and hypertensive diseases of pregnancy. Two of these studies supported our findings. Silver et al measured baroreflex sensitivity (BRS), an index of reflexive cardiac autonomic control after the onset of disease, in 20 women with PE and 20 women with PIH compared with 20 gestational age-matched controls and 20 control subjects who were not pregnant.32 BRS was reduced in normal pregnancy compared with nonpregnant women and was still further reduced in patients with hypertensive disorders of pregnancy suggesting a pathological abnormality in parasympathetic activity. Yang et al found that preeclamptic women ( 34 weeks) had lower HF-HRV compared with those who were normotensive.4 Two other small studies partially conflict with our findings. Ekholm et al found that women with PIH had higher levels of HF-HRV at the time of their diagnosis compared with women with uncomplicated pregnancies.33 However, they also studied midfrequency BPV and reported that higher levels were more common in hypertensive pregnant women. Midfrequency BPV (range, 0.07–0.15 Hz) overlaps with the LF frequency band evaluated in our study (range, 0.04–0.15 Hz) and as such, this finding is consistent with ours. Finally, Greenwood et al, using a different approach to assessment of ANS function (microneurography) showed that women with PIH had elevated levels of sympathetic nervous system activity compared with women with normal pregnancies.34 All of these studies were limited by small sample size and the fact that the indices of autonomic regulation were evaluated in women who already had been diagnosed with hypertensive disorders of pregnancy. Only one small study has examined the baseline ANS function prospectively. In a sample of 42 women studied on several occasions during pregnancy, Rang et al found no difference in LF-BPV or in BRS between women who went on to develop PE or remained normotensive,35 suggesting no role for the BPV in PE. Our data showing a lack of association between BPV and PE are consistent with this finding. Our study has several strengths. First, the sample size is large for studies of beat-to-beat BPV. Second, the use of military women as subjects provides for homogeneity with respect to many potentially confounding variables including physical health, employment status, health behaviors, and access to prenatal care. Limitations include the fact that not all subjects underwent psychophysiological testing. Women who participated in psychophysiological testing were similar to those who did not in age, education, and military rank but there was a statistically significant increase in risk of preterm delivery in the women who did not participate. Nonparticipants also were more likely to report job-related stress and were somewhat more likely to be black.14 PIH and PE were diagnosed on the basis of clinical criteria at the time. It is
Downloaded by: NYU. Copyrighted material.
1056
Spectral Analysis in Pregnancy
5 Stauss HM. Identification of blood pressure control mechanisms
6
7
8
9
10
11
12
13
14
15 16
Conflicts of Interest The authors report no conflicts of interest.
17
18
Note The cohort was assembled at Lackland Air force Base, San Antonio, TX when Dr. Hatch was a faculty member at Mt. Sinai Hospital, New York, NY. Reprints will not be available
19
20
21
References 1 Macdonald-Wallis C, Lawlor DA, Fraser A, May M, Nelson SM,
Tilling K. Blood pressure change in normotensive, gestational hypertensive, preeclamptic, and essential hypertensive pregnancies. Hypertension 2012;59(6):1241–1248 2 Berntson GG, Bigger JT Jr, Eckberg DL, et al. Heart rate variability: origins, methods, and interpretive caveats. Psychophysiology 1997;34(6):623–648 3 Berntson GG, Hart S, Ruland S, Sarter M. A central cholinergic link in the cardiovascular effects of the benzodiazepine receptor partial inverse agonist FG 7142. Behav Brain Res 1996;74(1-2): 91–103 4 Yang CC, Chao TC, Kuo TB, Yin CS, Chen HI. Preeclamptic pregnancy is associated with increased sympathetic and decreased parasympathetic control of HR. Am J Physiol Heart Circ Physiol 2000; 278(4):H1269–H1273
1057
22
23
24 25
by power spectral analysis. Clin Exp Pharmacol Physiol 2007; 34(4):362–368 deBoer RW, Karemaker JM, Strackee J. Hemodynamic fluctuations and baroreflex sensitivity in humans: a beat-to-beat model. Am J Physiol 1987;253(3 Pt 2):H680–H689 Montano N, Lombardi F, Gnecchi Ruscone T, et al. Spectral analysis of sympathetic discharge, R-R interval and systolic arterial pressure in decerebrate cats. J Auton Nerv Syst 1992;40(1):21–31 Dimier-David L, Billon N, Costagliola D, Jaillon P, Funck-Brentano C. Reproducibility of non-invasive measurement and of short-term variability of blood pressure and heart rate in healthy volunteers. Br J Clin Pharmacol 1994;38(2):109–115 Cevese A, Grasso R, Poltronieri R, Schena F. Vascular resistance and arterial pressure low-frequency oscillations in the anesthetized dog. Am J Physiol 1995;268(1 Pt 2):H7–H16 Mukai S, Hayano J. Heart rate and blood pressure variabilities during graded head-up tilt. J Appl Physiol (1985) 1995;78(1): 212–216 Cooke WH, Hoag JB, Crossman AA, Kuusela TA, Tahvanainen KU, Eckberg DL. Human responses to upright tilt: a window on central autonomic integration. J Physiol 1999;517(Pt 2):617–628 Schächinger H, Weinbacher M, Kiss A, Ritz R, Langewitz W. Cardiovascular indices of peripheral and central sympathetic activation. Psychosom Med 2001;63(5):788–796 Ditor DS, Kamath MV, Macdonald MJ, Bugaresti J, McCartney N, Hicks AL. Reproducibility of heart rate variability and blood pressure variability in individuals with spinal cord injury. Clin Auton Res 2005;15(6):387–393 Hatch M, Berkowitz G, Janevic T, et al. Race, cardiovascular reactivity, and preterm delivery among active-duty military women. Epidemiology 2006;17(2):178–182 Harris FJ. On the use of windows for harmonic analysis with discrete Fourier transorm. Proc IEEE 1978;66(1):51–83 Barros AJ, Hirakata VN. Alternatives for logistic regression in crosssectional studies: an empirical comparison of models that directly estimate the prevalence ratio. BMC Med Res Methodol 2003;3:21 Begum KS, Sachchithanantham K, De Somsubhra S. Maternal obesity and pregnancy outcome. Clin Exp Obstet Gynecol 2011; 38(1):14–20 Salihu HM, De La Cruz C, Rahman S, August EM. Does maternal obesity cause preeclampsia? A systematic review of the evidence. Minerva Ginecol 2012;64(4):259–280 Parati G, Grassi G, Coruzzi P, et al. Influence of cardiopulmonary receptors on the bradycardic responses to carotid baroreceptor stimulation in man. Clin Sci (Lond) 1987;72(5):639–645 Frattola A, Parati G, Cuspidi C, Albini F, Mancia G. Prognostic value of 24-hour blood pressure variability. J Hypertens 1993;11(10): 1133–1137 Sander D, Kukla C, Klingelhöfer J, Winbeck K, Conrad B. Relationship between circadian blood pressure patterns and progression of early carotid atherosclerosis: A 3-year follow-up study. Circulation 2000;102(13):1536–1541 Mancia G, Parati G, Hennig M, et al; ELSA Investigators. Relation between blood pressure variability and carotid artery damage in hypertension: baseline data from the European Lacidipine Study on Atherosclerosis (ELSA). J Hypertens 2001;19(11):1981–1989 Mancia G, Bousquet P, Elghozi JL, et al. The sympathetic nervous system and the metabolic syndrome. J Hypertens 2007;25(5): 909–920 Mancia G. Blood pressure reduction and cardiovascular outcomes: past, present, and future. Am J Cardiol 2007;100(3A):3J–9J Xie HH, Zhang XF, Chen YY, Shen FM, Su DF. Synergism of hydrochlorothiazide and nifedipine on blood pressure variability reduction and organ protection in spontaneously hypertensive rats. Hypertens Res 2008;31(4):685–691
American Journal of Perinatology
Vol. 32
No. 11/2015
Downloaded by: NYU. Copyrighted material.
possible, given current criteria that there may have been misclassification. BPV and HRV are currently research tools. The measurements are simple, noninvasive, and require a short testing period ( 5 minutes of rest followed by 5 minutes of testing). However, in order for these techniques to be clinically useful, they would have to be adapted by industry for clinical use. The strength of a uniform population is also a weakness. This association needs to be replicated in a more normal clinical cohort that would be expected to have greater variability in maternal weight and overall fitness. Another weakness is the overall infrequency of hypertensive disease in this fit population. One potential utility of this noninvasive technique would be to select a population at risk for future study. Testing occurred at 28 weeks’ gestation, assuring that the most severely affected patients would not be part of the study group. Future studies, measuring these indices at 22 to 24 weeks before the usual onset of signs and symptoms would assure the inclusion of the most severe forms of hypertensive disease in pregnancy. In conclusion, 5-minute estimates of LF-BPV and HF-HRV collected at the beginning of the third trimester of pregnancy allow for significant improvement in the ability to identify which patients will develop PIH. These associations were independent and additive to values of BP and maternal weight. Inclusion of these factors into risk stratification for PIH may enhance sensitivity and specificity of prediction for treatment of inclusion in clinical trials. The early identification of subjects at risk has the potential to reduce costs for cohort analyses and would allow for relatively early intervention.
Flood et al.
Spectral Analysis in Pregnancy
Flood et al.
26 Rothwell PM. Limitations of the usual blood-pressure hypothesis
31 Carr DB, McDonald GB, Brateng D, Desai M, Thach CT, Easterling TR.
and importance of variability, instability, and episodic hypertension. Lancet 2010;375(9718):938–948 Muntner P, Shimbo D, Tonelli M, Reynolds K, Arnett DK, Oparil S. The relationship between visit-to-visit variability in systolic blood pressure and all-cause mortality in the general population: findings from NHANES III, 1988 to 1994. Hypertension 2011;57(2): 160–166 Mena L, Pintos S, Queipo NV, Aizpúrua JA, Maestre G, Sulbarán T. A reliable index for the prognostic significance of blood pressure variability. J Hypertens 2005;23(3):505–511 Schillaci G, Pucci G, Parati G. Blood pressure variability: an additional target for antihypertensive treatment? Hypertension 2011;58(2):133–135 Zhang Y, Agnoletti D, Safar ME, Blacher J. Effect of antihypertensive agents on blood pressure variability: the Natrilix SR versus candesartan and amlodipine in the reduction of systolic blood pressure in hypertensive patients (X-CELLENT) study. Hypertension 2011;58(2):155–160
The relationship between hemodynamics and inflammatory activation in women at risk for preeclampsia. Obstet Gynecol 2001;98(6): 1109–1116 Silver HM, Tahvanainen KU, Kuusela TA, Eckberg DL. Comparison of vagal baroreflex function in nonpregnant women and in women with normal pregnancy, preeclampsia, or gestational hypertension. Am J Obstet Gynecol 2001;184(6): 1189–1195 Ekholm EM, Tahvanainen KU, Metsälä T. Heart rate and blood pressure variabilities are increased in pregnancyinduced hypertension. Am J Obstet Gynecol 1997;177(5): 1208–1212 Greenwood JP, Stoker JB, Walker JJ, Mary DA. Sympathetic nerve discharge in normal pregnancy and pregnancy-induced hypertension. J Hypertens 1998;16(5):617–624 Rang S, Wolf H, van Montfrans GA, Karemaker JM. Serial assessment of cardiovascular control shows early signs of developing pre-eclampsia. J Hypertens 2004;22(2):369–376
27
28
29
30
32
33
34
35
Downloaded by: NYU. Copyrighted material.
1058
American Journal of Perinatology
Vol. 32
No. 11/2015
