Multiscale asymmetry reveals changes in the maternal short-term heart rate dynamics of preeclamptic women during labor

Abstract

BACKGROUND:

It is known that acceleration and deceleration patterns in heart rate variability (HRV) are asymmetrically distributed in healthy subjects. Accordingly, novel approaches for assessing the asymmetrical properties of HRV, such as the multiscale asymmetry (MSA), have been applied in the perinatal field.

OBJECTIVE:

To study the asymmetry of accelerations and decelerations of maternal short-term cardiac dynamics of thirty-six normotensive and preeclamptic women during labor/nonlabor by MSA analysis.

METHODS:

The RR interval time series obtained from these participants were classified into four groups: normotensive (control) without labor C-NL, $n=$ 10; control with labor C-L, $n=$ 10; and two preeclamptic groups with absence or presence of labor P-NL, $n=$ 6; and P-L, $n=$ 10, respectively. Multiscale indices of heart rate asymmetry (HRA) such as Porta (P%), Guzik (G%) and Ehlers (E) were used to explore the changes of HRA in the normotensive and preeclamptic groups in the presence or absence of labor.

RESULTS:

The main result of this study shows that preeclamptic women manifest decreased magnitude of decelerations of heart rate dynamics compared to normotensive women indicated by G% and E. We speculate that a lower cardiac parasympathetic response may be manifested in preeclamptic women during labor/nonlabor compared to normotensive women.

CONCLUSIONS:

These observations represented a new insight into the autonomic cardiovascular regulation in preeclampsia, which could contribute to the perinatal field in the future.

Keywords

Pre-Eclampsia Autonomic regulation Asymmetry Labor HRV

1. Introduction

The more recent evidence confirms that preeclampsia is associated with an altered maternal cardiac autonomic function; specifically, preeclamptic women manifest increased sympathetic activity and decreased parasympathetic activity [1, 2].

Nonlinear signal processing analysis of heart rate variability (HRV) has provided useful computational tools for assessing the autonomic nervous system (ANS) [3]. Given that nonlinear approaches improve the predictive capacity for determining illness states linked with ANS function, these approaches to HRV analysis have been gaining popularity. Heart rate asymmetry (HRA) is a nonlinear feature based on the Poincaré plot, which allows determining the irreversibility of a heart rate or RR interval time series [4]. The HRA analysis suggests that the physiological regulation of heart rate deceleration (prolongation of successive RR intervals) and acceleration (reduction of consecutive RR intervals) appears as asymmetric [5]. Some metrics for evaluating HRA have been proposed in the biomedical signal processing field, e.g., the Porta’s index (P%) that assesses the asymmetry of the amounts of points situated in two regions of the Poincaré plot divided by the line of identity [6]. Also, the Guzik’s index (G%) that measures the asymmetry based on the cumulative distance of each point to the line of identity [5], and the Ehlers index (E), which is based on the evaluation of the skewness of the distribution of differences of time series [7].

After reviewing the relevant literature there seem to be no current reports of changes in HRA of preeclamptic women in the presence or absence of labor. In addition, our recent research proposes that the nonlinear analysis of HRV, such as that obtained by HRA, can provide relevant features for a better understanding of neuroinflammatory processes in preeclamptic women [8]. Therefore, this study aimed to analyze whether HRA is manifested in preeclamptic and normotensive pregnant women associated with the presence or absence of labor.

2. Methods

In this preliminary cross-sectional study, Mexican women aged between 18 and 35 years, who attended the obstetrical surgery and emergency areas of the Maternal-Perinatal Hospital “Mónica Pretelini Sáenz”, located in Toluca, Mexico State, Mexico, from February 2019 until October 2021, were invited to participate in this study. The participants were enrolled during their clinical follow-up routines in both clinical areas of the hospital. The Ethics Research Committee from this institution approved this study protocol (registration number: 2021-03-719), and all participants provided informed consent. Abdominal electrocardiogram (ECG) data were recorded in term pregnant women in semi-fowler position using a portable data acquisition device (Mobi, TMSi Systems, The Netherlands) with a bipolar electrodes placement and sampling frequency of 1000 Hz during 5 minutes. Clinical characteristics from participants were also collected before recordings. Subsequently, data from participants were classified into four groups: normotensive (control) without labor C-NL, $n=$ 10; control with labor C-L, $n=$ 10; and two preeclamptic groups with absence or presence of labor P-NL, $n=$ 6; and P-L, $n=$ 10, respectively. The medical diagnosis of preeclampsia (with and without severe features) was based on blood pressure ( $>$ 140/90 mm/Hg) and the presence of proteinuria (urine protein-creatinine ratio or UPCR $>$ 0.28) [9]. This biochemical indicator differentiated preeclampsia from gestational hypertension [10]. We also included normotensive and preeclamptic participants during the latent and active phases of labor. Furthermore, exclusion criteria to establish groups were the presence of diabetes mellitus, chronic or gestational hypertension, renal, autoimmune, or cardiovascular diseases, and administration of epidural analgesia during labor [11].

Abdominal ECGs were preprocessed using a bandpass filter with cutoff frequencies of 5–15 Hz; subsequently, the Pan-Tompkins algorithm was applied to filtered ECG time series for detecting QRS complexes [12]. The maternal short-term RR interval time series from the ECG were preprocessed by applying adaptive filtering to eliminate any ectopic beats from the series [13], and traditional time-domain linear indices (mean RR, SDNN, RMSSD, and pNN50) were computed using the Kubios software [14]. Besides, the following nonlinear indices were computed from the RR interval time series using the PyBioS software [15]: the multiscale asymmetry (MSA) index of Porta (P%), Guzik (G%), and Ehlers (E), and multiscale entropy (MSE).

Porta’s index P% [6] is calculated by dividing the percentage of negative $\Delta$ RR by the total of $\Delta$ RR $\neq$ 0. $\Delta$ RR corresponds to the arithmetic difference of two consecutive RR intervals, $N(\Delta\text{RR}^{-})$ denotes the number of negative $\Delta$ RR, and $N(\Delta\text{RR}\neq 0)$ indicates the number of all non-zero $\Delta$ RR:

$\displaystyle P\%=\frac{N(\Delta\text{RR}^{-})}{N(\Delta\text{RR}\neq 0)}% \times 100$ (1)

Guzik’s index G% [5, 6] is based on the percentage evaluation of the cumulative sum of the squared values of positive $\Delta$ RR to the cumulative sum of all squared $\Delta$ RRs:

$\displaystyle G\%=\frac{\sum_{i=1}^{N\left({\Delta\text{RR}^{+}}\right)}\Delta% \text{RR}^{+}(i)^{2}}{\sum_{i=1}^{N\left({\Delta\text{RR}}\right)}\Delta\text{% RR}(i)^{2}}\cdot 100$ (2)

Ehlers’ index [6, 7] is calculated as the skewness of the probability distribution of the $\Delta$ RR expressed as the appropriate moments of the time series:

$\displaystyle\text{E}=\frac{\sum_{i=1}^{N\left({\Delta\text{RR}}\right)}\Delta% \text{RR}(i)^{3}}{\left({\sum_{i=1}^{N\left({\Delta\text{RR}}\right)}\Delta% \text{RR}(i)^{2}}\right)^{3/2}}$ (3)

The multiscale asymmetry method corresponds to the calculation of P%, G%, and E for various $\tau$ -scaled versions of the original RR time series. For a given time series $\Delta\text{RR}$ of length $L$ , each point in a $\tau$ -scaled signal is defined as $\Delta\text{RR}^{\tau}\left[n\right]=\frac{1}{{\tau}}\sum_{i=\left({n-1}\right% ){\tau}+1}^{n{\tau}}\text{RR}\left[i\right],1\leqslant n\leqslant L/{\tau}$ . Then, P%, G%, and E are calculated for all $\Delta\text{RR}^{\tau}\left[n\right]$ signals. This study used a range of scales ( $\tau=$ 1–10) to analyze the asymmetry of maternal RR intervals.

To compare the mean values of clinic characteristics and time-domain linear HRV parameters among groups, one-way ANOVA was used in case of normality (Shapiro-Wilk test); otherwise, Kruskal-Wallis was applied. Subsequently, we applied a two-way ANOVA followed by an Uncorrected Fisher’s LSD posthoc test for comparing the MSA and MSE among groups. As a complementary analysis, associations between specific MSA indices and linear time-domain measures of HRV were studied using a two-tailed Spearman correlation test by combining the P-NL and P-L groups. The statistical significance was considered when $p<$ 0.05 for all tests. Statistical analyses were performed on GraphPad Prism version 8 (GraphPad Software, La Jolla, CA, USA).

Table 1

Clinical characteristics for the studied groups

Characteristics	C-NL $n=$ 10	C-L $n=$ 10	P-NL $n=$ 6	P-L $n=$ 10	$p$ -value
Age (years)	24.5 $\pm$ 2.7	23.90 $\pm$ 4.9	21.7 $\pm$ 6.5	28.4 $\pm$ 6.0	0.097
Gestational age (weeks)	39.02 $\pm$ 1.3	39.02 $\pm$ 0.9	38.5 $\pm$ 1.2	38 $\pm$ 1.0	0.211
BMI (kg/m ${}^{2}$ )	31.05 $\pm$ 6.9	27.90 $\pm$ 2.5	30.2 $\pm$ 5.4	28.8 $\pm$ 4.7	0.561
Systolic blood pressure (mm/Hg)	119.1 $\pm$ 14.2	112.9 $\pm$ 9.9	149.7 $\pm$ 17.0	146.6 $\pm$ 14.4	0.0001
Diastolic blood pressure (mm/Hg)	73.5 $\pm$ 12.6	71.7 $\pm$ 11.0	100.7 $\pm$ 17.3	95.5 $\pm$ 10.6	0.0001
UPCR (mg/mg)	–	–	0.52 $\pm$ 0.3	1.1 $\pm$ 0.4	0.476
Cervical dilation (cm)	–	3.6 $\pm$ 1.6	–	2.9 $\pm$ 1.4	0.282
Active phase of labor?
Yes	–	5(50%)	–	2(20%)
No	–	5(50%)	–	8(80%)

BMI: Body Mass Index, UPCR: Urine protein-creatinine ratio. Bold text indicates a significant p-value from non-bold text, assessed by one-way ANOVA or Kruskal-Wallis test.

3. Results and discussion

Table 1 depicts the mean $\pm$ SD of maternal clinical characteristics for all groups. No differences in clinical measures were found among the groups, except for systolic and diastolic pressures blood pressure between C-NL/C-L and P-NL/P-L, which are in accordance with the diagnosis of preeclampsia. In addition, time-domain linear measures of HRV were also not significantly different between groups (Fig. 1).

Figure 1.

Time-domain linear indices for normotensive (control) women with the absence of labor (C-NL), control with presence of labor (C-L), preeclampsia with the absence of labor (P-NL), and preeclampsia with the presence of labor (P-L). SDNN: Standard Deviation of Normal RR intervals (NN). RMSSD: Root Mean Square of the RR intervals Successive Differences. pNN50: Percentage of consecutive RR intervals differences that differ by more than 50 ms. Data are shown as mean $\pm$ SD. No significant differences were found among groups.

G% values more than 50 and E values greater than 0 suggest that the averaged magnitude of beat-to-beat decelerations was larger than that of accelerations. Values of G%, which are significantly lower than 50, and values of E, which are significantly smaller than 0, indicate the opposite [6]. MSA analysis did not reveal significant differences in P% values between groups (Fig. 2a). Thus, the percentage of cardiac accelerations seems to be equal in preeclampsia and normotensive women, and it does not depend on the presence or absence of labor.

Figure 2.

Multiscale asymmetry indices for normotensive (control) women with the absence of labor (C-NL), control with presence of labor (C-L), preeclampsia with the absence of labor (P-NL) and preeclampsia with the presence of labor (P-L). a) Porta index (P%), b) Guzik index (G%), c) Ehlers index (E), d) Multiscale Entropy (MSE). Data are shown as mean $\pm$ SEM. Significant differences ( $p<$ 0.05) between groups are indicated by the following symbols: # C-NL vs P-NL; *C-NL vs P-L; ^C-L vs P-NL and $+$ C-L vs P-L.

G% values in C-L and C-NL were significantly higher (0.009 $<p<$ 0.04) than P-NL and P-L on several scales, Fig. 2b. On scales 4–9, E values in C-NL were significantly higher ( $p<$ 0.05) than P-L, Fig. 1c. Likewise, E revealed significantly lower values ( $p<$ 0.05) in P-NL compared to C-NL on scales 7 and 8. E values in C-NL were significantly higher (0.008 $<p<$ 0.04) than and P-L on several scales, Fig. 2c. Moreover, MSE showed no significant differences in any group, Fig. 2d. These results indicate that preeclampsia does affect changes in the HRA of decelerations independently of the absence or presence of labor. Furthermore, nonlinear features of heart rate fluctuations appear to be preserved in all groups (as indicated by MSE).

Table 2

Correlation of multiscale asymmetry (MSA) indices of Guzik and Ehlers on scales 4 and 8 ( $G_{4}$ %, E ${}_{4}$ and G ${}_{8}$ %, E ${}_{8}$ , respectively) with time-domain linear indices of HRV for preeclampsia

	$G_{4}$ %		$E_{4}$		$G_{8}$ %		$E_{8}$
	$r$	$p$	$r$	$p$	$r$	$p$	$r$	$p$
Mean RR	0.23	0.37	0.26	0.31	0.32	0.22	0.32	0.21
SDNN	0.42	0.10	0.45	0.08	0.45	0.08	0.44	0.08
RMSSD	0.47	0.06	0.50	0.04	0.47	0.06	0.50	0.04
pNN50	0.33	0.20	0.40	0.11	0.28	0.28	0.30	0.24

Note: To increase the number of samples for the correlation analysis, we combined the preeclamptic groups with labor and nonlabor (P-L $+$ P-NL) in the preeclampsia group ( $n=$ 16) Bold text indicates a significant $p$ -value assessed by Spearman’s two-tailed correlation test.

Spearman correlation analysis revealed that women with preeclampsia exhibited a moderately positive significant correlation between the RMSSD and both G% and E on scales 4 and 8 (Table 2).

The main results of this study then show that preeclamptic laboring and nonlaboring women manifest decreased magnitude of decelerations of heart rate compared to normotensive laboring and nonlaboring women. Thus, the presence or absence of labor did not affect HRA in preeclamptic women. According to our results, the MSA analysis seems more appropriate to identify short-term maternal cardiac dynamics changes in preeclamptic women in the presence or absence of labor than traditional time-domain linear indices of HRV.

A non-zero value of E from cardiac time series indicates time-irreversibility. Interestingly, the C-NL group exhibited the largest value (E $>>$ 0) compared to the remaining groups. Thus, preeclamptic women manifest a lower average magnitude of decelerations than normotensive control women, indicated by E and G%.

It is recognized that pregnancy increases the heart rate and may reduce the adaptive capacity of the cardiovascular system [16]. Notwithstanding that previous studies have suggested an increased sympathetic activity in preeclampsia [2, 17], the decreased magnitude of decelerations reported here for preeclamptic laboring and nonlaboring women could also indicate a reduction in cardiac parasympathetic response in comparison with normotensive women. The positive association between RMSSD and both E and G% in preeclampsia on scales 4 and 8 also support such relationship between the magnitude of decelerations of heart rate dynamics and the parasympathetic cardiac response.

In addition, a diminished cardiac vagal response might be associated with alteration in the anti-inflammatory cholinergic pathway activity during preeclampsia [8].

The potential limitations of this study involve the sample itself. The samples size was relatively small. For this reason any $p$ -values obtained in this study should be treated only as exploratory tools. Therefore, a considerable amount of data must be analyzed in future studies to endorse these results.

4. Conclusion

The MSA analysis of short-term maternal heart rate dynamics revealed changes in the magnitude of decelerations of women diagnosed with preeclampsia. We speculate that a lower cardiac parasympathetic response may be manifested in preeclamptic women during labor/nonlabor compared to normotensive women. Thus, we consider the MSA analysis as a potential promising biomarker for evaluating autonomic cardiac changes in preeclampsia.

Footnotes

Acknowledgments

The authors thank the study participants and the Mónica Pretelini Hospital staff. We especially thank Dr. Carlos Arturo Paniagua Coahuila, head of the obstetrical surgery area, and Dr. Arturo Harold Von Johann Juárez Díaz, head of the emergency area at Mónica Pretelini Hospital.

Conflict of interest

The authors declare that they have no known competing financial interests or personal relationships that could have influenced the work reported in this paper.

Funding

Rosselin Gabriela Ceballos-Juárez was supported by the Consejo Mexiquense de Ciencia y Tecnología (COMECYT) (Grant No. EESP2021-0004, programa de Estancias de Investigación Especializadas COMECYT-EDOMÉX 2021-2022). J. Javier Reyes-Lagos thanks the Mexican Public Education Ministry (Secretaría de Educación Pública: SEP) for providing financial support (project number: 511-6/2020-7841).

References

Musa

Adam

Lutfi

. Heart rate variability and autonomic modulations in preeclampsia. PLoS ONE [Internet]. 2016; 11(4): 1-9. doi: 10.1371/journal.pone.0152704.

Tejera

Rodrigues

Areias

Rebelo

Nieto-Villar

. Network centrality and multiscale transition asymmetry in the heart rate variability analysis of normal and preeclamptic pregnancies. Communications in Nonlinear Science and Numerical Simulation [Internet]. 2011; 16(3): 1589-96. doi: 10.1016/j.cnsns.2010.07.009.

Constantinescu

Matei

Costache

Cuciureanu

Arsenescu-Georgescu

. Linear and nonlinear parameters of heart rate variability in ischemic stroke patients. Neurologia i Neurochirurgia Polska [Internet]. 2018; 52(2): 194-206. doi: 10.1016/j.pjnns.2017.10.002.

Rohila

Sharma

. Asymmetric spread of heart rate variability. Biomedical Signal Processing and Control [Internet]. 2020; 60(101985): 1-11. doi: 10.1016/j.bspc.2020.101985.

Guzik

Piskorski

Krauze

Wykretowicz

Wysocki

. Heart rate asymmetry by Poincaré plots of RR intervals. In: Biomedizinische Technik. 2006; pp. 272-5.

Porta

Casali

Gnecchi-Ruscone

Tobaldini

Montano

, et al. Temporal asymmetries of short-term heart period variability are linked to autonomic regulation. American Journal of Physiology – Regulatory Integrative and Comparative Physiology [Internet]. 2008; 295(2): 550-7. doi: 10.1152/ajpregu.00129.2008.

Ehlers

Havstad

Prichard

Theiler

. Low doses of ethanol reduce evidence for nonlinear structure in brain activity. Journal of Neuroscience. 1998; 18(18): 7474-86.

Reyes-lagos

Abarca-castro

. Nonlinear analysis of heart rhythm in preeclampsia: a route for translational clinical applications in neuroinflammation. Clinical Hypertension. 2021; 27(24): 1-8.

Instituto Mexicano del Seguro Social. Detección, diagnóstico y tratamiento de las enfermedades hipertensivas del embarazo. Instituto Mexicano del Seguro Social [Internet]. 2017; 63. http//www.imss.gob.mx/profesionales-salud/gpc%0Ahttp://www.cenetec.salud.gob.mx/contenidos/gpc/catalogoMaestroGPC.html#.

10.

Rana

Lemoine

Granger

Karumanchi

. Preeclampsia: Pathophysiology, Challenges, and Perspectives. Circulation Research. 2019; 124(7): 1094-112.

11.

Deschamps

Kaufman

Backman

Plourde

. Autonomic nervous system response to epidural analgesia in laboring patients by wavelet transform of heart rate and blood pressure variability. Anesthesiology. 2004; 101(1): 21-7.

12.

Sedghamiz

. Complete Pan Tompkins Implementation ECG QRS detector – File Exchange – MATLAB Central. MATLAB Central File Exchange. 2020.

13.

Wessel

Voss

Malberg

Ziehmann

Voss

Schirdewan

, et al. Nonlinear analysis of complex phenomena in cardiological data. Herzschrittmachertherapie und Elektrophysiologie. 2000; 11(3): 159-73.

14.

Tarvainen

Niskanen

J-P

Lipponen

Ranta-Aho

Karjalainen

. Kubios HRV – heart rate variability analysis software. Computer Methods and Programs in Biomedicine. 2014; 113(1): 210-20.

15.

Silva

LEV

Fazan

Marin-Neto

. PyBioS: A freeware computer software for analysis of cardiovascular signals. Computer Methods and Programs in Biomedicine [Internet]. 2020; 197: 105718. doi: 10.1016/j.cmpb.2020.105718.

16.

Sanghavi

Rutherford

. Cardiovascular physiology of pregnancy. Circulation. 2014; 130(12): 1003-8.

17.

Naranjo Orellana

De La Cruz Torres

. La entropía y la irreversibilidad temporal multiescala en el análisis de sistemas complejos en fisiología humana. Acta Médica Colombiana. 2010; 3(1): 29-32.