Fetal autonomic response to severe acidaemia during labour

Abstract

Please cite this paper as: van Laar J, Peters C, Vullings R, Houterman S, Bergmans J, Oei S. Fetal autonomic response to severe acidaemia during labour. BJOG 2010;117:429–437.

Objective  Spectral analysis of heart-rate variability is used to monitor autonomic nervous system fluctuations. The low-frequency component is associated with sympathetic and parasympathetic modulation and the high-frequency component is associated with parasympathetic modulation. The objective was to study whether changes in low-frequency or high-frequency power of heart-rate variability occur in case of fetal distress.

Design  Case–control study.

Setting  Obstetric unit of a tertiary-care teaching hospital.

Population  Twenty healthy human fetuses during labour at term of which ten had an umbilical artery pH < 7.05 (cases), and ten had an arterial pH > 7.20 (controls) after birth.

Methods  Spectral information about fetal beat-to-beat heart rate, calculated from direct fetal electrocardiogram registrations, was obtained by using a short-time Fourier transform.

Main outcome measures  Absolute power and normalised power in the low-frequency and high-frequency bands.

Results  No differences were found between fetuses with and without acidaemia in absolute low or high frequency power (P = 0.2 and P = 0.3, respectively). During the last 30 minutes of labour, acidaemic fetuses had significantly increased normalised low-frequency power (P = 0.01) and decreased normalised high-frequency power (P = 0.03) compared with non-acidaemic fetuses. These differences were not observed from 3 to 2 hours before birth (P = 0.7 and P = 0.9, respectively).

Conclusion  The autonomic nervous system of human fetuses at term responds adequately to severe stress during labour. Normalised low and high frequency power of heart-rate variability might be able to discriminate between normal and abnormal fetal condition.

Introduction

Detection of fetal distress during labour is complex. The cardiotocogram (CTG), a simultaneous recording of fetal heart rate and uterine contractions, is the method used worldwide for fetal surveillance. Poor specificity of this method has resulted in increased rates of operative deliveries without a decrease in perinatal mortality or cerebral palsy.¹ It has been shown that during labour combined use of CTG and automatic ST-waveform analysis of the fetal electrocardiogram (ECG; STAN^®, Neoventa Medical, Moelndal, Sweden) reduces the rates of severe metabolic acidosis at birth and instrumental vaginal delivery for fetal distress.² However, slow progressive deterioration of the CTG without pathological changes in ST-waveform (ST-events) has been reported as a cause of neonatal metabolic acidosis despite monitoring using STAN^®3. Furthermore, STAN^® remains dependent on the assessment of the CTG, which has a high inter- and intra-observer variability.⁴ Although the CTG has a high sensitivity, ST-events occur at a similar frequency for normal and abnormal CTG patterns.⁵ This illustrates the need for more detailed information on the fetal cardiovascular response to hypoxaemia.

Hypoxaemia activates the autonomic nervous system, which subsequently modulates beat-to-beat heart rate.⁶ Spectral analysis is a method that can be used to detect and quantify these changes in heart rate objectively.^7,8 Spectral analysis decomposes sequential R-R interval series into a sum of sinusoids of different amplitudes and frequencies by the fast Fourier transform algorithm.⁶ The power spectrum reflects the magnitude of heart-rate variability (power) present at different frequency ranges.⁶ Spectral power (variability) in the low-frequency (LF) range is associated with sympathetic and parasympathetic nervous system modulation and spectral power (variability) in the high-frequency (HF) range is associated with parasympathetic modulation.⁷ This is because the fetal heart rate fluctuates under the influence of the sympathetic and parasympathetic nervous system. As impulses from the parasympathetic nervous system are conveyed much faster than impulses of the sympathetic nervous system, sympathetic modulation is solely present in the LF range while parasympathetic modulation is also present in the HF range.^7,9 In other words, sympathetic modulation of the fetal heart rate leads to slow oscillations while parasympathetic modulation also leads to fast oscillations. As spectral analysis evaluates oscillations in beat-to-beat fetal heart rate, it has the potential to monitor the autonomic nervous system modulation and may provide an early diagnostic tool for assessing fetal distress.

Previous studies have demonstrated that spectral power in the LF and HF ranges decreased in case of fetal hypoxaemia or acidaemia.^10–14 This was thought to be the result of immaturity or of decompensation of the fetal autonomic nervous system. However, these studies used absolute values of LF and HF power, whereas changes in total power (total heart rate variability) influence LF and HF power in the same direction. Normalised values of LF and HF power seem more suitable for fetal monitoring because they detect relative changes, that cannot be masked by changes in total power.⁹ Normalised LF (LFn) and normalised HF (HFn) power are calculated by dividing LF and HF power respectively by total power and represent the controlled and balanced behaviour of the two branches of the autonomic nervous system.⁹ We hypothesised that the autonomic cardiovascular control is functional in fetuses at term, and that LFn power would gradually increase in case of distress because of increased sympathetic nervous system modulation. The aim of our study was to compare spectral values in healthy and distressed fetuses during labour at term to determine whether differences in spectral values exist, which could be used in future research to improve fetal monitoring.

Methods

A case–control study was performed. Healthy fetuses of at least 36 weeks of gestation with intrapartum fetal ECG recordings, whose umbilical cord arterial and venous acid–base status was analysed immediately after delivery, were studied. Only good-quality fetal ECG recordings, free of ectopic beats and missing data, until at least 10 minutes before birth were included. Furthermore, only fetal ECG registrations from healthy mothers who experienced an uncomplicated pregnancy and did not use any medication, except oxytocin or epidural analgesia, were included. Pregnancies complicated by intrauterine growth restriction or fetal congenital anomalies were excluded. The controls (i.e. fetuses without acidaemia) were defined as fetuses with an umbilical arterial pH > 7.20 after birth. The included registrations were made in the Máxima Medical Centre, Veldhoven and were selected consecutively from the period January 2007 to August 2007. The cases (i.e. fetuses with acidaemia) were defined as having an umbilical artery pH < 7.05. These registrations were selected consecutively from the period January 2006 to December 2007. As a result of the strict inclusion and exclusion criteria only five fetuses with acidaemia could be included. In addition, five fetuses with acidaemia from the University Medical Centre Utrecht were selected consecutively from the period January 2001 to July 2002. Both hospitals are tertiary-care teaching hospitals.

Data acquisition and signal processing

A frequency-specific assessment of heart-rate variability by spectral analysis is a well-established means of characterising autonomic cardiac control.⁷ Our method has been described in more detail in previous studies.^15,16 The fetal ECG was recorded during delivery using a single-helix scalp electrode (Goldtrace™, Neoventa Medical, Moelndal, Sweden) and a maternal skin electrode. The scalp electrode was connected to a STAN S31^® monitor (Neoventa Medical). The included registrations were analysed including accelerations and decelerations.

The STAN^® device detected the R-peak location of each heart beat exactly, from the scalp ECG signal at a rate of 500 Hz. Although separate ECG waveforms can be aligned to calculate average fetal ECG waveforms to improve the signal quality of the waveform used for ST analysis, RR-intervals are provided on a true beat-to-beat basis. These beat-to-beat RR-interval data sets were stored on a PC hard disc and analysed off line.

For spectral analysis a fast Fourier transform algorithm was used. Based on previous studies as well as the physiological range of fetal heart and respiratory movement rates the following frequency bands were chosen: total frequency band: 0.04–1.5 Hz, LF: 0.04–0.15 Hz and HF: 0.4–1.5 Hz.^16,17 The Fourier transform is a mathematical method that transforms data from quantity (fetal heart rate) varying over time to amplitude (spectral power) varying over a range of frequencies. In other words, it performs the conversion of a function in the time domain to the frequency domain.

Before applying spectral analysis, the beat-to-beat RR-intervals are pre-processed in a standardised way to increase reliability of spectral information.¹⁵ The Fourier transform requires sampled data to be equidistant. As RR intervals are not equidistant they have to be resampled. The Nyquist criterion states that to obtain reliable spectral information, the signal has to be resampled at least twice the frequency of the highest frequency of interest (1.5 Hz). Because the fast Fourier transform requires 2ⁿ samples, data were resampled at 4 Hz. Because the rapidly fluctuating fetal autonomic system results in non-stationarity of the fetal heart-rate signal, a minimum window length of 64 seconds was chosen. This window length was considered reliable because it included at least twice the wavelength of the lower frequency of interest (0.04 Hz). The signal’s mean was subtracted and to reduce the effect of discontinuities at each side of the window the signal was multiplied with a triangular window (Parzen).¹⁵

The signal is then ready for Fourier transformation. To calculate the power spectrum of the pre-processed RR-series, 256-point Fourier transforms were calculated for partly overlapping, intervals of 64 seconds, which were shifted every 0.25 seconds.¹⁵ Hence, after 64 seconds, spectral values are provided every 0.25 seconds. Besides calculating the absolute spectral power of fetal heart rate variability in the specified frequency bands, normalised values were calculated by dividing LF and HF power respectively by total power. Spectral power data were given in arbitrary units. We refer to earlier papers for further details on our method of spectral analysis.^15,16

Statistical methods

For statistical analysis SPSS 12.0 (SPSS Inc., Chicago, IL, USA) was used. Fetal characteristics were tested with Student’s t test (continuous variables) and Fisher’s exact test (categorical variables). Significance was tested two-sided at an α-level of 0.05. For the last 30 minutes of labour, mean power in the LF, HF, LFn, HFn bands was calculated for intervals of 5 minutes. Values were calculated for 4 minutes overlapping providing one new value every minute for the preceding 5-minute continuous signal segment. These longitudinal mean 5-minute LF, HF, LFn, HFn power values were compared between fetuses with and without acidaemia using an analysis of variance (ANOVA) for repeated measures. In addition, to compare mean 5-minute LFn and HFn power longitudinally between fetuses with and without acidaemia during the first stage of labour, the repeated measures ANOVA was also performed for the 1-hour period from 3 to 2 hours before birth.

For all 20 fetuses good quality ECG data were available until 9 minutes before birth. From four fetuses with acidaemia and five fetuses without acidaemia beat-to-beat heart-rate data could be obtained until the last minute before birth. Because the ANOVA for repeated measures automatically deletes observations with missing values, two separate analyses were performed, one for 20 fetuses until 9 minutes before birth and one for nine fetuses until the last minute before birth. Results of both analyses were comparable; therefore we present only P-values for the whole group of 20 fetuses.

Results

In total, 20 fetuses were included, of which 10 had an umbilical artery pH < 7.05 (cases), and 10 had an arterial pH > 7.20 (controls). As shown in Table 1, fetuses with and without acidaemia were comparable with regard to gestational age (P = 0.3) and birthweight (P = 0.3). The neonatal 5-minute Apgar score was significantly lower in the acidaemia group compared with the control group (P ≤ 0.01). Four fetuses with acidaemia and none of those without were admitted to the neonatal ward (P = 0.09). Five mothers in the case group and six mothers in the control group used epidural analgesia (P = 0.65). Seven women in the case group and three in the control group needed oxytocin augmentation (P = 0.4). According to the STAN^® criteria, in the case group nine women had an abnormal and one had an intermediate CTG during the last 30 minutes of labour. In the control group eight women had an abnormal and two had an intermediate CTG (P = 1.0). During the last hour of labour, seven fetuses with acidaemia and four without had one or more ST-events (P = 0.4); of these, six and three, respectively, were significant ST-events indicating the need for intervention (P = 0.4). All ST-events were based on a rise in T/QRS ratio.

Table 1.   Clinical characteristics of the included fetuses

	Fetuses with acidaemia (Cases; n = 10)	Fetuses without acidaemia (Controls; n = 10)
Values are expressed as mean (standard deviation).
Gestational age (days)	283 (8)	278 (11)
Birthweight (g)	3414 (423)	3643 (562)
5-minute Apgar score	8 (1)	10 (0)
Cord arterial pH	6.98 (0.07)	7.26 (0.03)
Cord venous pH	7.09 (0.1)	7.33 (0.04)
Cord arterial base excess (mmol/l)	−17 (14)	−5 (2)

As shown in Figure 1, during the last 30 minutes of labour, there were no significant differences in absolute LF or HF spectral power of heart-rate variability between fetuses with and fetuses without acidaemia, no significant differences within subjects over time and there was no significant interaction between acidaemic and non-acidaemic fetuses.

Figure 1.  Absolute low-frequency (LF) and high-frequency (HF) spectral power of heart-rate variability during the last minutes of labour in fetuses with acidaemia (pH < 7.05) and control fetuses without acidaemia (pH > 7.20). There were no differences in absolute LF or HF power between fetuses with acidaemia and those without (P = 0.2 and P = 0.3, respectively), no significant differences within subjects over time (P = 0.9 and P = 0.7, respectively) and no significant interaction was observed (P = 0.7 and P = 0.7, respectively). au, arbitrary units.

As shown in Figure 2, during the last 30 minutes before birth, LFn spectral power of heart-rate variability was significantly higher in the acidaemia group and HFn spectral power of heart-rate variability was significantly higher in the control group during the whole study period. We found no significant differences in LFn or HFn power within subjects over time and no significant interaction for these spectral values between fetuses with acidaemia those without acidaemia.

Figure 2.  Normalised low-frequency (LF) and high-frequency (HF) spectral power of heart-rate variability during the last minutes of labour in fetuses with acidaemia (pH < 7.05) and non-acidaemic control fetuses (pH > 7.20). Normalised LF power was significantly higher (P = 0.01) and normalised HF power was significantly lower (P = 0.03) in the acidaemia group. There were no differences within subjects over time (P = 0.7 and P = 0.2, respectively) and no significant interaction was observed (P = 0.8 and P = 0.6, respectively).

The observed significant differences in LFn and HFn power of heart rate variability between fetuses with and without acidaemia were not observed during the 1-hour period from 3 to 2 hours before birth (Figure 3).

Figure 3.  Normalised low-frequency (LF) and high-frequency (HF) spectral power of heart-rate variability from 3 to 2 hours before birth in fetuses with acidaemia (solid black line; pH < 7.05) and control fetuses without acidaemia (solid grey line; pH > 7.20). No differences in normalised LF or HF power between fetuses with and those without acidaemia were observed (P = 0.7 and P = 0.9, respectively).

Discussion

Spectral power of heart-rate variability in the LFn band was significantly higher while spectral power of heart-rate variability in the HFn band was significantly lower during the last 30 minutes of labour in the case of fetal acidaemia compared with fetuses in the non-acidaemic control group. In other words, slow oscillations in heart rate increase while fast oscillations in heart rate decrease in cases of fetal distress. This reflects a shift in autonomic modulation (sympathovagal balance) towards sympathetic predominance in fetusues with acidaemia. Spectral analysis objectively quantifies the variability in fetal heart rate as a function of frequency. Therefore, the observed differences in normalised spectral power between fetuses with and without acidaemia are the result of differences in fetal heart-rate pattern. Despite this, CTG classification alone was not able to show differences between fetuses with acidaemia and those without. Therefore, it seems that spectral analysis is able to observe more subtle changes in fetal heart-rate patterns that are difficult to observe in CTG.

Several studies have reported on normal values for the umbilical artery pH,¹⁸ pH values between 7.0 and 7.20 have been suggested as the cutoff for fetal acidaemia.¹⁹ We chose an umbilical artery pH of 7.05 as the cutoff value for severe fetal acidaemia, because this threshold is associated with major neurological morbidity¹⁹ and therefore, this cutoff is considered as clinically important. Furthermore, this value has been applied as a marker of severe fetal acidaemia in previous studies.^2,10 Because an umbilical cord arterial pH > 7.20 is considered normal after an uncomplicated vaginal delivery at term (>10th percentile)²⁰ and as fetal acidaemia has traditionally been defined as an arterial pH < 7.20,^21,22 the controls were defined as fetuses with an arterial pH > 7.20.

Previous studies using spectral analysis of fetal heart-rate variability for fetal monitoring during labour used 2-minute continuous signal segments.^10,14 We used shorter signal segments of 64 seconds to produce power spectra, assuming that in such a short window the data have a greater likelihood of presenting a stable state. We chose a 64-second window because the longest cycle we were interested in is 0.04 Hz. It is required to have at least two repetitions of the longest cycle included and for computational efficiency the lowest power of two, exceeding two times this longest cycle, was chosen. Furthermore, because a moving, partly overlapping window was used for Fourier transformation, which was shifted every 0.25 seconds, the influence of non-stationarity was minimised.

In the absence of sudden catastrophic events, acidaemia in term fetuses is reported to develop in a period of at least 90 minutes.²³ Therefore, analysis was also performed for the 1-hour period from 3 to 2 hours before birth (i.e. before the second stage of labour). During this period a normal fetal pH is expected. Because the differences between fetuses with acidaemia and those without were not seen during this 1-hour period, it seems that the LFn and HFn response during the last 30 minutes of labour are acidaemia-associated alterations. Therefore, it is assumed that the increase in LFn power and the decrease in HFn power were the result of the stress of labour resulting in acidaemia instead of intrinsic differences in fetal autonomic modulation that already exist in an early stage of labour.

Fetal heart rate variability is positively related to gestational age and negatively associated with birthweight.¹⁰ It is important to emphasise that none of the included fetuses was growth retarded and that gestational age and birthweight were comparable between the groups. Therefore, in our study the observed differences between fetuses with and without acidaemia were not the result of differences in these fetal baseline characteristics. Furthermore, our results were not biased by maternal use of medication influencing fetal autonomic modulation. Epidural injection may cause minimal, transient changes in fetal heart rate.²⁴ Although a small and non-significant difference exists between fetuses with and without acidaemia in epidural use, these differences are unlikely to influence our results, because these epidural-induced changes in fetal heart rate resolve within 30 minutes.²⁴

Although accelerations and decelerations influence fetal heart-rate variability, spectral analysis was performed on continuous signal segments, because ignoring accelerations and decelerations would lead to results with little physiological relevance. Accelerations were not seen during the last 30 minutes of labour in any of the included registrations and the CTG classification according to the STAN^® criteria during the last 30 minutes of monitoring was not different between the groups. Therefore, accelerations and decelerations per se are unlikely to account for the observed differences between fetuses with acidaemia and those without.

There is a strong correlation between a rise in T/QRS ratio and the level of circulating catecholamines²⁵ and hence anaerobic metabolism. Siira et al. found that an increase in T/QRS ratio is associated with an increase in LF heart-rate variability.¹⁰ In our study we found a non-significant trend toward increased rates of ST-events based on a rise in T/QRS ratio in fetal acidaemia.

Previous studies in human fetuses found a decrease in absolute LF and HF power during severe fetal compromise.^10–14 This was thought to represent a sign of decompensation of the fetal circulatory system.¹⁰ In our study we could not confirm this fetal autonomic decompensation. On the contrary, we found that even in severe fetal metabolic acidaemia, although HFn power was low, LFn power was high until the last minute before birth. Possibly the observed differences are the result of using absolute values in previous studies. If spectral components are expressed in absolute units, the changes in total power influence LF and HF power, concealing the relative distribution of the energy.⁹ Because we found significant differences between fetuses with and those without acidaemia using normalised spectral values, normalisation seems to be more suitable. Although the sample size is quite small, it is unlikely that the observed differences are the result of chance alone because they are fully in line with the autonomic response to stress in adults.²⁶ On the other hand, the lack of significant differences in absolute spectral values between fetuses with and those without acidaemia could be the result of the small number of fetuses examined in our study. However, because of the strict inclusion and exclusion criteria we were not able to include more cases. To determine the diagnostic value for identification of fetuses at risk for severe acidaemia our results have to be verified in larger prospective studies.

Another possible limitation of our study is that the accuracy with which the beat-to-beat RR-intervals are available is 2 milliseconds, because the STAN^® device has a sampling rate of 500 Hz. This is inadequate for short-term variability measurement in the time domain as described by Dawes, because this method requires an accuracy of 1 millisecond.²⁷ However, for short-term recordings frequency-domain measures are preferred over time-domain measures, because more knowledge exists on the physiological interpretation.⁹ The European Society of Cardiology and the North American Society of Pacing and Electrophysiology state that the optimal sampling range for frequency domain measures (spectral analysis) is 250–500 Hz.⁹ This is in line with previous studies, which showed that increasing the sampling rate over 500 Hz barely changed spectral power values.²⁸ For spectral analysis sampling rates over 500 Hz are only required in people with extremely low heart-rate variability (e.g. after cardiac transplantation),²⁹ which is not the case for the included fetuses.

Low-frequency and high-frequency power increases during the second and third trimesters of pregnancy.³⁰ The sympathetic nervous system is effective as early as mid-gestation, while the parasympathetic nervous system matures much later in pregnancy and begins to exert typical reflex responses at term and reaches adult levels only after birth.³¹ Furthermore, the fetal cardiovascular response to neurotransmitters increases with gestational age because of maturation of the neuroeffector system and this increase in sensitivity continues after birth.³¹ This maturation of the autonomic nervous system might affect the fetal haemodynamic responses to stress. However, the observed increase in LFn and the decrease in HFn power indicate a shift from sympathovagal balance towards sympathetic predominance and reduced vagal modulation in the human fetus as a response to the stress of labour. This reaction is comparable to the autonomic nervous system response to physical stress in human adults.²⁶ Therefore, despite its incomplete development, the autonomic nervous system in human fetuses at term is capable of exerting a strong response to severe stress.

It has been known for a long time that decreased fetal heart rate variability is associated with fetal distress.³² Both autonomic withdrawal and a maximally stimulated sympathetic input can lead to diminished heart-rate variability.²⁶ Yu et al. found during fetal hypoxaemia (without acidaemia) an increase in LF and HF power of heart-rate variability.³³β-Adrenoceptor blockade had no effect on the changes in the power spectrum induced by hypoxaemia, whereas fetal heart rate decreased.³³ Therefore, they stated that this sympathetic influence on the fetal heart during hypoxaemia must be predominantly the result of increased adrenomedullary secretion of catecholamines instead of cardiac sympathetic neural activity. Gardner et al. showed that acute hypoxia in fetuses with acidaemia is also associated with great concentrations of plasma catecholamines.³⁴ It is unlikely that these high levels of circulating catecholamines cause LF fluctuations in heart rate. We found an increase in LFn power in fetal acidaemia. Therefore, we suggest that a saturating high level of circulating catecholamines and sympathetic neural activity, might be the conceptual basis of decreased heart-rate variability in severe fetal distress in human fetuses at term. It might be speculated that the high level of catecholamines sensitises the sympathetic nervous system, which results in a relative increase of low-frequency heart-rate variability.

Conclusion

Spectral analysis of fetal heart rate variability improves our understanding of the physiological response of the human fetus to hypoxaemia and can provide valuable insight into pathophysiological conditions. Our study indicates that normalised LF and HF power are promising markers for fetal distress and could probably be useful to improve fetal monitoring. To determine the diagnostic value for identification of fetuses at risk for severe acidaemia, prospective studies are needed.

Disclosure of interest

There are no known financial, personal, political, intellectual, or religious interests of any of the authors.

Contribution to authorship

S.G.O. and J.W.M.B. were, besides being the initiators, also the supervisors of this study. J.O.E.H.v.L. designed the study, collected the data, analysed the data and wrote the manuscript. C.H.L.P and R.V. designed the beat-to-beat fetal heart rate and the spectral analysis software and were involved in writing the manuscript. S.H. was the leading force in the set up of the methodological layout and performed the statistical analyses of the data. All authors discussed the results and commented on the manuscript.

Details of ethics approval

Because the study was performed retrospectively informed consent and institutional review board approval were not required.

Funding

The study has been performed without the support of funding.

Acknowledgements

The authors would like to thank A. Kwee, Department of Obstetrics and Gynaecology, University Medical Centre Utrecht, for provision of their STAN^® registrations. We thank J.J.M. Rijpkema, Department of Mathematics and Computer Science, Eindhoven University of Technology, for statistical advice. We are grateful to P.F.F. Wijn, Department of Applied Physics, Eindhoven University of Technology, for his useful comments on our manuscript.