Dr JK Cleal, The Institute of Developmental Sciences, University of Southampton, Southampton General Hospital (MP 887), Southampton SO16 6YD, UK. Email: firstname.lastname@example.org
Please cite this paper as: Cleal J, Thomas M, Hanson M, Paterson-Brown S, Gardiner H, Green L. Noninvasive fetal electrocardiography following intermittent umbilical cord occlusion in the preterm ovine fetus. BJOG 2010;117:438–444.
Objective To investigate whether a noninvasive fetal electrocardiography (fECG) system can identify cardiovascular responses to fetal hypoxaemia and validate the results using standard invasive fECG monitoring techniques.
Design Prospective cohort study.
Setting Biological research facilities at The University of Southampton.
Population or Sample Late gestation ovine fetuses; n = 5.
Methods Five fetal lambs underwent implantation of vascular catheters, umbilical cord occluder and invasive ECG chest electrodes under general anaesthesia (3% halothane/O2) at 119 days of gestation (term ∼147 days of gestation). After 5 days of recovery blood pressure, blood gases, glucose and pH were monitored. At 124 and 125 days of gestation following a 10-minute baseline period a 90-second cord occlusion was applied. Noninvasive fetal ECG was recorded from maternal transabdominal electrodes using advanced signal-processing techniques, concurrently with invasive fECG recordings.
Main outcome measures Comparison of T:QRS ratios of the ECG waveform from noninvasive and invasive fECG monitoring systems.
Results Our fECG monitoring system is able to demonstrate changes in waveforms during periods of hypoxaemia similar to those obtained invasively, which could indicate fetal distress.
Conclusions These findings may indicate a future use for noninvasive electrocardiography during human fetal monitoring both before and during labour in term and preterm pregnancies.
Monitoring fetal wellbeing in labour remains an inexact science with intermittent auscultation being comparable to continuous electronic monitoring by cardiotocography in the low-risk pregnancy.1 Cardiotocographic monitoring for high-risk pregnancies is recommended2 but is associated with a high false-positive rate and therefore increased interventions even when backed up by fetal blood sampling.1 Randomised trials of the most advanced technique combining cardiotocography with fetal electrocardiography (fECG) parameters (specifically the ST waveform analyser termed STAN) suggested improvement in the detection of acidosis with fewer false positives,3 although subsequent reports of its use in routine clinical practice have identified a number of problems.4–6 A common disadvantage with either form of electronic monitoring is that of interpretation of the cardiotocography trace, on which the STAN guidelines also hinge.7,8
Specific disadvantages of the STAN technique are that the signal pick-up is by a spiral electrode screwed into the fetal scalp, which is invasive and cannot be applied until there is both cervical dilatation and membrane rupture. As STAN technology analyses the change in ratio of the T to QRS waveforms assuming the initial ratio is normal, the ‘critical’ ST event will be missed if a fetus is already compromised before the monitoring is commenced. This is indeed the explanation given for some of the babies monitored with STAN that have developed hypoxic ischaemic encephalopathy.7 Another disadvantage is that STAN can only be used in pregnancies after 36 weeks of gestation whereas it is the preterm babies that are most at risk of cerebral insult. If a fECG signal could be picked up transabdominally it could be used in the antenatal period, in early labour, with intact membranes and on preterm babies, addressing many of these problems.
Our group has demonstrated the off-line feasibility and ease of use of fECG in more than 600 pregnancies from 15 weeks of gestation using skin electrodes.9 We have described normal time intervals for fECG and reported a pilot study of its use in labour which confirmed that fECG waveforms, heart-rate variability, the maternal ECG and uterine contractions could be recorded simultaneously from a single device avoiding the potential risks of invasive monitoring with the fetal scalp electrode.10 More recently another study used superficial maternal electrodes to pick up the fECG signal to record fetal heart rate and its variability both at home in the antenatal period and during labour.11 Large prospective clinical trials will be required to demonstrate the ability of superficial electrode recordings to detect human fetal hypoxaemia in a low-risk population in labour, but before such a trial is undertaken a comparative study of the ECG signal obtained using a superficial noninvasive electrode and that obtained with an invasive electrode in response to hypoxaemia is needed.
We used an ovine model of umbilical cord occlusion to investigate the ability of our noninvasive fECG system to identify cardiovascular responses to fetal hypoxaemia introduced by cord occlusion from sensors placed remotely on the maternal abdomen. The results were validated using standard invasive fECG monitoring techniques in the ovine fetus.
Material and methods
Welsh Mountain ewes (n = 5) of uniform body condition score and age were group-housed on wheat straw. From −16 days of gestational age they were fed a complete pelleted diet (89.2% dry matter as fed, providing 10.7 MJ/kg dry matter [metabolisable energy] and 14.8% protein; adjusted to gestational age12) with free access to water. The estrous cycle was synchronised by withdrawing vaginal medroxyprogesterone acetate impregnated sponges (Veramix, Upjohn Ltd, Crawley, UK) 12 days after insertion. A Welsh Mountain ram was introduced for 3 days, and 0 days of gestational age was taken as the first day on which an obvious raddle mark was observed. On study day 60 of gestation each ewe was scanned for pregnancy using ultrasound and the number of fetuses was confirmed: there were three singleton and two twin pregnancies.
At 119 days of gestation, general anaesthesia was induced using thiopentone sodium (10 mg/kg, intravenously; Link Pharmaceuticals, Horsham, UK) and maintained with 2% halothane (Concord Pharmaceuticals Laboratory Ltd, Reading, UK) in O2 (1 l/min). A midline incision was made in the lower abdominal wall, and the fetal head, chest and proximal portion of the umbilical cord were exteriorised through an incision in the uterine wall. Polyvinyl catheters (1.0 mm internal diameter, 2.0 mm external diameter; Portex Ltd, Hythe, UK) filled with heparinised saline (100 U/ml, heparin sodium; Leo Pharmaceuticals, Princes Risborough, UK and Saline 0.9%; 3S-Healthcare, Enfield, UK) were placed in the left jugular vein, right carotid artery and amniotic cavity. A catheter was placed in a maternal jugular vein for administration of drugs and blood sampling.
Three stainless-steel electrodes were sutured to the fetal chest for ‘invasive’ ECG recording and an inflatable occluder cuff (OCHD16; In Vivo Metric, Healdsburg, CA, USA) was positioned around the proximal portion of the umbilical cord and secured to the abdominal skin. The volume required for complete inflation of the occluder cuff was determined at surgery. The uterine and abdominal incisions were closed, and the catheters were exteriorised through the maternal flank and secured to each ewe’s back in a plastic pouch.
To record the ‘noninvasive’ fECG, electrodes (Ethicon, Edinburgh, UK) were sewn onto the exterior abdominal wall of the ewe (shaved and cleaned before surgery). The electrodes were sutured in place on the ewes’ abdomens because the wool and skin lanolin reduced the adhesiveness of standard electrodes. Twelve electrodes were evenly placed over the whole abdomen using the standardised orientation used in humans,9 a reference electrode was sewn in the centre and a ground electrode was sewn to the left hind leg: no previous knowledge of fetal position was required.
At surgery, antibiotics were administered to the ewe (Terramycin topically to incision sites [Pfizer, Eastleigh, Northants, UK]; intramuscular Betamox [Norbrook Laboratories Ltd, Carlisle, UK]; Crystapen intravenously [Britannia Pharmaceuticals Ltd, Redhill, UK]; and gentamicin intravenously [Mayne Pharmaceuticals Plc, Royal Leamington Spa, UK]), fetus (Crystapen intravenously), and amniotic cavity (Crystapen and gentamicin).
A 5-day postoperative recovery period was allowed before experimentation during which fetal and maternal catheters were flushed daily with heparinised saline to maintain their patency and antibiotic treatment was given to the fetus, ewe and amniotic cavity. Ewes and fetuses received an overdose of barbiturate (pentobarbitone sodium Ph. Eur 0.8 ml/kg intravenously, 200 mg/ml; Animalcare Ltd, York, UK) at the end of the study.
Fetal carotid artery and amniotic pressures (Capto AS, N-3193, Horten, Norway/NL 108, Digitimer Ltd, Welwyn Garden City, Herts, UK), and ECG (NL 100/104/125, filter windows set between 0 and 500 Hz) data were captured (sampling rate 1000 samples per second [Hz]; Maclab/8, AD Instruments Pty Ltd, Castle Hill, Australia) and recorded (Chart, AD Instruments, Chalgrove, UK). Blood pressure measurements were corrected for amniotic pressure (to remove pressure changes due to factors such as amniotic fluid volume, fetal and maternal movement).
Noninvasive fECG (system and methodology described previously9,10,13) was monitored via maternal abdominal electrodes connected to a 16-bit multichannel digital recorder with a sampling rate of 1024 Hz, and electrode impedance was <1 kΩ (comparable with previous studies9,10). A high-pass filter of 2 Hz and a low-pass filter of 120 Hz were applied to the data and a Blind Source Separation technique was used to estimate the independent biological sources. Although it is possible to attribute a selected independent source with appropriate weighting to each electrode location, when one independent source is present this contains all of the representative information. The independent source equivalent to the fECG was identified by manual inspection and where twins were present (Ewes 2 and 5) two fECG signals were identified. For the twin pregnancies the heart rate variability (Figure 1A) and fECG morphology (Figure 1B), were determined before and in response to the cord occlusion, with the fetus experiencing no change in variability being rejected. The selection was confirmed by comparison of the heart rate with the invasive recording, with a 100% correlation. Where an independent source was not continuously identifiable during the period for comparison, due to excessive movement activity of the ewe, a no result was recorded.
On 124 and 125 days of gestation fetal blood pressure, noninvasive and invasive fECGs were recorded. Following a 10-minute baseline period, a 90-second umbilical cord occlusion was carried out by complete inflation of the occluder cuff (∼4 ml saline). Fetal femoral arterial blood (0.25 ml) was collected before and 60 seconds into the occlusion into heparinised syringes for glucose, lactate, pH and blood gas measurements.
Plasma glucose, oxygen partial pressure, carbon dioxide partial pressure, pH and lactate in arterial blood were measured using a blood gas analyser (ABL700, Radiometer, Copenhagen, Denmark) and were corrected for temperature (39°C).
Mean arterial blood pressure (mmHg) and heart rate (beats per minute) were acquired from the blood pressure recording.
The noninvasive fECG components, as previously described,13 were identified by inspection and synchronised with the invasive reference source. To enable the comparison of ECG complex changes due to umbilical cord occlusion, a 20-beat baseline sample before umbilical cord occlusion and a 20-beat sample immediately before release of the cord occlusion (≈80–90 seconds) were taken for both noninvasive and invasive measurements. Each sample of 20 fECG complexes was peak detected from the R waves, coherently averaged and the T:QRS ratio was calculated for each averaged complex by manual inspection. The approximately 10-second period (or 20 beats) was chosen through experimentation as the lowest number of beats to provide a PQRST complex of consistent quality that allowed measurement of the QRS and T wave amplitudes. It is in the interests of any technique to have the number of averaged complexes as low as possible, particularly when detecting acute changes.
Biochemical data are expressed as mean ± SEM and analysed by paired Student’s t test. Fetal electrocardiography data are expressed as mean ± SEM and type of fECG system (invasive or noninvasive) and time point during UCO (0 or 90 s) analysed using two-way ANOVA. The response to UCO was compared between 124 and 125 dGA by paired Student’s t test. Statistical significance was accepted when P < 0.05.
Fetal arterial blood measurements
The 90-second umbilical cord occlusion on day 125 of gestation resulted in a significant fall in fetal arterial partial pressure of oxygen (25.9 ± 0.7 to 7.2 ± 1.0 mmHg, P < 0.005), pH (7.36 ± 0.01 to 7.29 ± 0.01, P < 0.01) and, a significant rise in partial pressure of CO2 (36.0 ± 1.4 to 42.0 ± 1.5 mmHg, P < 0.05). No significant change in glucose (0.64 ± 0.07 to 0.48 ± 0.08 mmol/l) or lactate (0.70 ± 0.04 to 0.88 ± 0.13 mmol/l) was observed. This is in agreement with our previous studies.14
Invasive measures of blood pressure and heart rate
Over the course of a 90-second umbilical cord occlusion on day 125 of gestation, mean arterial blood pressure increased (Figure 2A) and fetal heart rate decreased (Figure 2B), comparable with our previous studies.14
Separation of the noninvasive fECG signal was obtained and P, Q, R, S and T waves were identified in both the noninvasive and invasive fECG in all five fetuses. Figure 3 illustrates a noninvasive measurement and an invasive independent fECG component obtained using blind source separation with the waveforms in this case well correlated with a correlation coefficient of 0.96 using a point-by-point comparison. In both the noninvasive and invasive cases a marked change was seen between the waveforms and T:QRS ratios before and after umbilical cord occlusion (Figure 4). The fECG was not obtained for two fetuses on 124 days of gestation because of excessive movement of the ewe. On days 124 and 125 of gestation both noninvasive and invasive fECG showed significant changes in T:QRS ratio following umbilical cord occlusion (P < 0.05, Figure 5). Table 1 summarises the change in T:QRS ratio for each subject on each day. There was no significant change in T:QRS ratio responses between days 124 and 125. There was no significant effect of fECG recording system on the change we observed in T:QRS ratio and no statistical interaction between response to umbilical cord occlusion and fECG recording system (indicating no difference in their individual responses) on gestational days 124 and 125.
Table 1. Electrocardiograms for T:QRS ratio change in response to a 90-second umbilical cord occlusion for each fetus on each day
This study confirms that the noninvasive fECG monitoring system is able to record changes in the fECG during and after umbilical cord occlusion in this ovine model. The noninvasive fECG signatures were similar to those obtained invasively and therefore this system may be useful in monitoring fetal wellbeing in pregnancy and labour before and after membrane rupture. We obtained contemporaneous recordings successfully in five lambs.
Invasive measures of recording the ECG are not suitable for use in people, except during labour, and much of our knowledge comes from invasive monitoring of ovine fetuses.15,16 The STAN technique using the T:QRS ratio is used in preference to absolute measures of ST elevation or depression because of the technical difficulties associated with determining the isoelectric line in the fetus. It is an effective automated means of determining hypoxaemia, but because it only detects a change in the ratio, the recording may reflect worsening fetal distress rather than its onset. Clearly the most important advantage of noninvasive fECG is that recordings can be obtained before labour begins and in its early stages making it less likely that the critical ST event will be missed.
Blind source separation techniques in this study were able to successfully obtain the P QRS and T elements of the signal in five fetuses and these were confirmed by direct comparison with invasive fECG recordings. The T wave amplitude reflects the repolarisation wave of the ECG complex and is inherently low frequency, making it vulnerable to high-pass filtering effects from preprocessing methods and also the propagation medium within the maternal volume.13 This has manifested itself as a suppressed T-wave in the noninvasive case, primarily because of the 2 Hz high-pass filter which was required to suppress the significantly increased breathing artefact found in the ovine model that is generally not observed in the human fetus. However, our study suggests that an alteration in the T:QRS ratio was easily identified in response to umbilical cord occlusion in this ovine model. It is important to stress that this noninvasive fECG equipment samples at 1024 Hz, which is equivalent to ECG systems in clinical use and three times that of currently available commercial models.11 There are differences in the morphology of waveforms and these are the result of the orientation of the heart vector in the cardiac volume and its associated propagation pathway, which sometimes suppresses the T wave. Moreover, several independent fECG sources may be obtained from a single heart, particularly in mid-gestation, but this is less common towards term when labour generally occurs. However, these are significant factors affecting quality and interpretation of transabdominal recordings and are in agreement with those who have looked at ST segment elevation.13 Electrode pads were not suitable for monitoring in sheep because they do not stick to the ewe’s abdomen. We have not found this a problem in the pregnant women we have studied, even in labour, and dry electrodes produce even better signal acquisition.
This study demonstrated an observable change of the ST segment in noninvasive recordings from a single independent component extracted by blind source separation in response to a hypoxaemic event. Comparison of T wave amplitude in our model demonstrated less increase in the amplitude of the T wave compared with the invasive fECG recordings, but the change in the T:QRS ratio in eight of 12 cases for noninvasive, compared with 13 of 15 for invasive, would be considered an abnormal event according to the clinical guidelines of ST segment analysis.7 There is scope for improvement of this statistic in women because in humans this typically separates with a 0.5-Hz high-pass preprocessing filter. The changes in invasive fECG, blood gases, blood pressure and heart rate confirmed responses to cord occlusion in the ovine model similar to those reported by our group previously.14
Our noninvasive fECG monitoring system is able to demonstrate changes in fECG waveforms during periods of hypoxaemia that could indicate fetal distress. These findings may indicate a future use for this system during human fetal monitoring both before and during labour in term and preterm pregnancies.
Disclosure of interest
Contribution to authorship
J.C., M.T., M.H., H.G. and L.G. designed the study. J.C. and L.G. carried out the experiments and analysed the data. M.T. processed the ECG data. S.P.B. provided expert advice on interpreting the data. All contributed to the writing of the paper.
Details of ethics approval
All procedures were conducted with strict adherence to the regulations of the UK Home Office Animals (Scientific Procedures) Act, 1986, under project licence (PPL 30/1858).
This work was supported by QinetiQ and M.A.H. is supported by the British Heart Foundation. Funding for analysis and support for J.C. was provided by Tiny Tickers charity. H.M.G. is supported by the Biomedical Research Centre and the Institute of Obstetrics and Gynaecology Trust at Queen Charlotte’s and Chelsea Hospital.
We are grateful to the Biological Research facilities at The University of Southampton and The Royal Veterinary College for their expert animal care.