Portal de Periódicos da CAPES

Commentary on ‘Antenatal cardiotocogram quality and interpretation using computers’

2014; Wiley; Volume: 121; Issue: s7 Linguagem: Inglês

10.1111/1471-0528.13151

1471-0528

PJ Steer,

Phonocardiography and Auscultation Techniques

Stillbirth is one of the major challenges facing modern obstetrics. Antenatal assessment of fetal health remains problematic. Cardiotocography (continuous assessment of the fetal heart rate and uterine activity) can give vital short-term information about fetal wellbeing, but its assessment by ‘eyeballing’ (subjective pattern recognition) is complicated by high intra-observer and inter-observer variability. Computer numerical analysis is objective and reproducible, and provides a firmer basis for categorisation into normal and pathological. Dawes, Redman and colleagues produced a commercially viable system for such analysis that is still in use today. The principle has recently been extended to intrapartum care. For most of the existence of the species recognisable as Homo sapiens, not only was it impossible to assess the health of the fetus in utero, it was also impossible to know whether it was alive or dead, or even if there was a fetus at all. For example, from September 1554 through to June 1555, Queen Mary I of England was thought to be pregnant by her marriage to Prince Philip of Spain, and her half sister Elizabeth (later Queen Elizabeth I) was released from house arrest to witness the birth—which never happened, as in fact Mary had never been pregnant. In the seventeenth century, the French physician Marsac was reported by his friend Philippe LeGaust to have heard fetal heart sounds by placing his ear on the mother's abdomen, but it was not until the nineteenth century that any attempt was made to capitalise upon this observation.1 In 1818, Francois Mayor, a Swiss surgeon, and in 1821 a French nobleman Jean-Alexandre Le Jumeau, Vicomte de Kergaradec independently confirmed that the beating of the fetal heart could be heard. In 1834 Evory Kennedy of Dublin published an essay on ‘obstetric auscultation, or means of detecting life or death of a foetus before birth’.2 By 1906, Cremer3 had described the detection of the fetal electrocardiogram (ECG), using electrodes placed on the mother's abdomen and in her vagina. But even with modern technology, detecting and analysing the fetal ECG noninvasively remains challenging because the maternal ECG signal is much larger and often swamps the small fetal signal, and the electrical activity of the abdominal muscles creates substantial interference. Effective fetal monitoring using the fetal ECG requires either rupture of the amniotic membranes, or the passage of an electrode through the abdominal wall to contact the fetus. A major step forward was when the first commercial fetal monitor (the Hewlett-Packard 8020a) was designed by Hammacher in Germany (protected under U.S. Pat. No. 3,318,303). It used a microphone to produce a beat-to-beat analysis of the fetal heart rate, and it picked up the signal from the mother's abdominal wall. Unfortunately, recordings were difficult to make if the mother was restless or obese, and the next step was the use of ultrasound radiated from an abdominal transducer to detect fetal heart movement using the Doppler principle. This more powerful approach was introduced by the British company Sonicaid in their first fetal monitor, the FM2, which came into widespread use in the early 1970s. Technology was later developed to improve the quality of signal detection and processing, and the technique was rapidly adopted by other fetal monitoring companies. Initially, use of continuous fetal heart rate monitoring was predominantly during labour, but it soon became apparent that it might also be useful antenatally. Ultrasound was also the key to the development of fetal imaging techniques that were less harmful than X-rays, still widely used when I began training in the 1960s. Pioneered by Ian Donald in Glasgow, by the 1970s the concept of using ultrasound to measure fetal size and thereby detect fetal growth restriction had been widely adopted. However, many studies have shown that simply detecting small babies in the third trimester does not improve outcome. For example, a 2008 Cochrane review of eight trials of routine late pregnancy ultrasound screening in unselected populations, which included over 27 000 women, concluded that ‘routine late pregnancy ultrasound in low-risk or unselected populations does not confer benefit on mother or baby’.4 By the early 1990s, it had become apparent that the proper measure of pathological fetal growth restriction is function and not size.5, 6 But how to measure function? In the late 1960s, measurement of urinary estriol excretion was introduced as a means of assessing the function of the feto-placental unit,7 and it became very popular during the 1970s. However, by the 1980s, reports had begun to emerge suggesting that estriol measurements in either the urine or plasma, were of little clinical value8 and soon the same verdict was being reached for human placental lactogen.9 Measurement of flow velocity waveforms in the umbilical artery was suggested by FitzGerald and Drumm in 197710 but it was not until the mid-1980s that the resistance index was introduced as a screen for increased placental blood flow resistance and the detection of babies that were small for gestational age.11 However, the GRIT (Growth Restriction Intervention Trial) study in the 1990s showed that Doppler flow velocity waveforms are unable to give guidance as to the precise stage of growth restriction at which delivery is indicated.12 By the early 1990s, a practical clinical protocol had developed in which fetal growth was monitored by ultrasound biometry and if a fall off in growth was noted, umbilical Doppler flow velocity waveforms were measured at weekly intervals. If these became abnormal, daily (or even twice daily) fetal heart rate monitoring was commenced.13 This brings us to the question of how the fetal heart rate pattern should be interpreted, a controversy that continues to this day. The idea that changes in the fetal heart rate in response to uterine contractions could give an indication of fetal condition was first introduced by Edward Hon in the USA, documented in a landmark publication in 1958.14 In the UK, fetal heart rate interpretation in labour generally followed the scheme proposed by Beard et al. in 1971.15 By the second half of the 1970s, fetal heart rate and uterine activity traces (cardiotocograms) were being used to assess the fetus antenatally.16, 17 However, by the mid-1980s, several randomised controlled trials had failed to find any beneficial effect of the technique.16, 18 Attempts were made to refine it by the addition of fetal acoustic stimulation (essentially, a buzzer applied to the maternal abdominal wall),19 but it transpired that the stress induced in the fetus could precipitate decompensation and it increased the premature delivery rate without an improvement in outcome. A similar fate befell the ‘contraction stress test’, in which an oxytocin infusion was used to stimulate uterine contractions to assess fetal response20—there was a 30% false-positive rate. Attention reverted to trying to refine the analysis of the non-stress test, and an important step forward was the realisation that the fetal heart rate pattern did not stay constant through gestation. In particular, the discovery of fetal rest activity cycles (which appear as the fetus matures, usually between 28 and 32 weeks) indicated that it was necessary to monitor the fetal heart rate for a sufficient period of time to allow the fetus in quiet sleep (associated with episodes of reduced fetal heart rate baseline variability, lasting for an average of 15 minutes21) to change into active sleep. This is when fetal breathing occurs, associated with an increase in variability equivalent to sinus arrhythmia.22 Geoffrey Dawes was born in 1918 and qualified in medicine in 1943. He was appointed director of the Nuffield Institute for medical research in Oxford in 1948, only 5 years after qualification. He embarked on a programme of research into fetal physiology that established the unit as one of the premier centres for this type of study in the world, in recognition of which he was elected a Fellow of the Royal Society in 1971. He retired from the Nuffield Institute in 1985, but continued his research at the Sunley Research Centre at Charing Cross Hospital and remained research active until shortly before he died in 1996. He was particularly interested in the fetal circulation and fetal breathing movements. Much of his initial work was done in sheep.23-30 However, he increasingly turned his attention to the human, and being of a mathematical turn of mind, became interested in the numerical analysis of the human fetal heart rate, and in particular, its modulation by breathing and movement.31 His paper in 1981 in the British Journal of Obstetrics and Gynaecology on this topic31 marked the start of a very productive collaboration with Christopher Redman, newly appointed as a consultant in obstetric medicine in the Oxford maternity unit. Together, they are named on 33 papers on PubMed, the last being in 1999, 3 years after Geoffrey had died.32 Once Geoffrey had turned his interest to the human fetus, he soon produced publications on the effect on the human fetal heart rate of fetal breathing and body movements,33 fetal micturition,34 diurnal variation,35 external physical stimulation36 and maternal smoking.37 The key to his approach was the avoidance of the purely descriptive system still routinely used to categorise fetal heart rate trace patterns in clinical practice, and instead the use of computerised numerical analysis,38 using the microprocessors that were just becoming available.39 By 1990, Dawes and Redman had worked with Sonicaid to produce the ‘system 8000’.40 Software had been written in C (a programming language), which ran on a personal computer ‘with hard disk’ (!), interfaced with an electronic fetal monitor. This categorised signal quality (many monitoring errors in clinical practice occur because of reliance on poor-quality signals), detected uterine contraction peaks, calculated baseline fetal heart rate variation, and quantified accelerations and decelerations. The system was designed to take into account the episodic changes in fetal heart rate and fetal movements characteristic of sleep states. (It is notable that Geoffrey was always careful to talk about sleep states; he was aware that the fetus is almost never in a state that we would characterise as ‘awake’ in utero, because this would result in a high oxygen consumption rate without having any obvious value in the monotonous intrauterine environment. Popular mythology that babies are ‘awake’ when they are in active fetal movement is just that, a myth.) They noted that in clinical practice, inter-observer and intra-observer variation is greatest in assessing fetal heart rate variation, although a decrease in variation is the most reliable index of fetal deterioration. They pointed out that computerised analysis not only measured fetal heart rate variation accurately, it was also reproducible (unlike human ‘eyeball’ analysis). They also made the important observation that the overall baseline rate was surprisingly insensitive to progressive hypoxia, and that extreme tachycardia is usually due to infection or maternal pyrexia, not hypoxia. The next important observation was to distinguish between short-term and long-term variation of the baseline fetal heart rate.41 An hypoxic fetus with diminished autonomic function can still show substantial long-term variations (sometimes called sinusoidal), and so they defined a new index of short-term fetal heart rate variation (the 1/16 minute epoch–epoch variation) that was insensitive to long-term sinusoidal variation, a pattern which still causes important clinical misinterpretations. Having developed the technology over the better part of a decade, it was time to put it to the test. With support from ‘Birthright’ (the charity now renamed ‘Wellbeing of Women’), system 8000 was tested prospectively for a year at the Luton and Dunstable district general hospital, and the Princess Anne Hospital Southampton. A combined total of 6530 recordings were taken from 2864 ‘patients’. Importantly, they used a ‘randomised’ approach to see if feedback from the system about signal quality would improve the overall quality of the recording, and therefore the interpretation obtained. Randomisation was based upon the hospital number, an approach which would nowadays not be acceptable, but in the late 1980s, randomisation itself was relatively novel, especially in studies of this size. In the ‘revealed’ or feedback group, audible and visible warnings were given to adjust the transducer if signal loss was high (and therefore quality of recording poor). In Southampton there was specific training to encourage midwives to pay attention to this feedback, whereas in Luton there was not. Accordingly, signal quality improved substantially more in Southampton than in Luton. Overall, in 546 records (8.4% of the total), recording was stopped before the computer had advised stopping (because the heart rate pattern had satisfied the ‘Dawes–Redman criteria’), or before 60 minute had passed (whichever was the shorter). This meant that these traces were inadequate for clinical assessment because the fetal heart rate variation was low, and insufficient time had been allowed for an episode of high variation to appear. The traces were stopped despite the computer displaying a message ‘to continue’. The authors were clearly exasperated by this behaviour and commented that ‘it is evident that simply pointing out the high signal loss on the screen and printed record is insufficient to ensure that the second record is made’ and ‘it is a matter of concern that the instructions on record taking were often ignored’. This illustrates an important issue that has still not been resolved to this day, which is the tendency of clinicians not to follow simple basic procedures, and to ignore abnormalities even when they have been detected. Why this happens remains unexplained. For example, in 1990, Ennis and Vincent reported on 64 cases of ‘obstetric accidents’ notified to the Medical Protection Society (a body providing medicolegal indemnity to clinicians). They found that there were six cases in which the quality of the recording was unsatisfactory, and moreover in 14 cases, an obvious cardiotocogram abnormality was either not noticed, or ignored.42 These problems are currently leading to a medicolegal crisis in the UK. The annual expenditure of the NHS litigation authority has risen from £70 million in 1998 to £1.3 billion in 2012–13. More than a third of the total expenditure relates to obstetric negligence, and by 2012–13 it had reached £482 million, approximately £1.32 million per day. A final paragraph in the paper revealed an important limitation to the use of the system 8000. In the two hospitals during the year of the study, there were 45 stillbirths. In only two of these pregnancies had a recent fetal heart rate record been taken, both because of a history of reduced fetal movements. Even if computerised analysis of the fetal heart rate was able to give warning of impending fetal demise, it was clearly impractical to use it several times a day on all pregnancies, and so the challenge of predicting those cases in which it might be useful remained (and still remains to this day; in recent decades the rate of stillbirth in the UK has stubbornly failed to decline43). The system 8000 is still commercially available from Huntleigh (who acquired Sonicaid in the 1990s)44 but to date no other manufacturer has adopted a similar approach, and antenatal computerised fetal heart rate analysis has not been widely adopted, because of the difficulty of predicting the cases in which it might be useful. The 2012 Cochrane systematic review of antenatal cardiotocography included only six randomised studies involving 2105 women, and these were mostly of poor quality.45 The meta-analysis was therefore underpowered to assess potentially preventable deaths; there were only two studies in which this aspect could be assessed, and the relative risk of 0.23 had 95% confidence intervals from 0.04 to 1.29. Although the relative risk is encouraging, the numbers are clearly too small to conclude that antenatal monitoring per se is beneficial. No randomised trial of the computerised system versus ‘eyeballing’ the traces has been carried out, and therefore it has not been possible to assess whether the computerised approach is of value. However, although Dawes and Redman never extended their computerised analysis to intrapartum traces, regarding them as too complicated for their 1990s software, their approach did stimulate such an attempt by the Plymouth group led by Keith Greene and Robert Keith. In their 1995 paper in the British Journal of Obstetrics and Gynaecology they described an ‘intelligent computer system’ for managing labour using algorithms to analyse the cardiotocogram.46 Their software, which has been given the acronym of INFANT (INtelligent Fetal AssessmeNT), has been implemented on the electronic intrapartum care monitoring system called ‘Guardian’ and manufactured by K2 Medical Systems. With a grant of £5.9 million from the UK Health Technology Assessment programme, a randomised prospective trial of the software including more than 46 000 women has been completed to test the hypothesis that it will reduce adverse fetal outcomes during labour. The results should be available by early 2015. In 1988 in an earlier paper in the British Journal of Obstetrics and Gynaecology, Dawes and Redman reported on the computerised analysis of the fetal heart rate of 588 women admitted in labour and who were monitored for up to an hour as a screening test.47 6.8% of recordings had reduced fetal heart rate variability, but none of these required resuscitation at birth or special care. They commented that as the tests were revealed, they might have influenced management and prevented poor outcome, and went on to conclude that there was a basis for a randomised controlled trial of numerically analysed fetal heart rate screening in early labour ‘but that this would need to be large (more than 10 000 patients) to test the benefit of detecting the most sinister pattern (decelerations with reduced variation) of which only one example was found in this study’. I am sure that Prof Dawes would have been delighted that such a trial has now been carried out. I have no interests to disclose.