Article first published online: 22 DEC 2003
BJOG: An International Journal of Obstetrics & Gynaecology
Volume 110, Issue 1, pages xv–xvi, January 2003
How to Cite
Grant, J. M. (2003), Editor's Choice. BJOG: An International Journal of Obstetrics & Gynaecology, 110: xv–xvi. doi: 10.1046/j.1471-0528.2003.51101.x
- Issue published online: 22 DEC 2003
- Article first published online: 22 DEC 2003
Birth asphyxia and the fetal heart rate trace
Asphyxia is a term that is misunderstood by obstetricians, but in connection with intrapartum fetal hypoxia, it has a precise definition. The word is derived from the Greek asphyxos, meaning pulseless, which accurately describes the infant born with the end stages of fetal hypoxia and metabolic acidosis. There are also many understandings about electronic fetal monitoring, for many obstetricians assume that it gives information about the function of the infant's heart, when actually it gives information about the function of the infant's brain. We got off to a bad start by calling the investigation a fetal heart rate trace, or cardiotocogram, which focusses attention on the infant's heart, when we should have called it a fetal brain trace, or encephalotocogram, which focusses attention on the infant's brain. Many obstetricians assume that decelerations are due to a decrease in myocardial function, when in the early stages of fetal hypoxia the infant's heart is working overtime and cardiac output is increased. Decelerations are due to cardiovascular reflexes mediated by the infant's central nervous system. It is only with progressive fetal hypoxia with severe metabolic acidosis that, if a threshold is crossed, cardiac failure occurs, cardiac output decreases dramatically, cerebral blood flow all but ceases, and widespread cerebral cortical necrosis occurs, to result in stillbirth or death a few hours or days after birth or survival with the worst form of cerebral palsy, spastic quadriplegia. In these circumstances, the infant is born pulseless or asphyxiated. It is all or nothing; there is no such thing as ‘mild asphyxia’.
Birth asphyxia is the most tragic consequence of childbirth, for the infant, for his parents, and for the obstetricians and midwives responsible for the care of the woman in labour. In an attempt to clarify the muddled thinking about birth asphyxia, the International Cerebral Palsy Task Force published three essential criteria for the diagnosis of cerebral palsy due to birth asphyxia (although the authors forbear to use this term): severe metabolic acidosis, hypoxic–ischaemic encephalopathy, and spastic quadriplegia1. Catherine Greenwood et al. (pages 6–11) performed an analysis of 138 infants with cerebral palsy of whom 27 were the subjects of a legal claim. These infants were more likely to fulfil the three essential criteria for birth asphyxia than those who were not the subjects of a legal claim. However, information necessary for establishing the essential criteria was often missing from the case notes, especially concerning metabolic acidosis; and the majority of claims were settled without the three essential criteria. Only time will tell whether the concept of the ‘three essential criteria’ of birth asphyxia will be accepted in claims for medical negligence.
The purpose of electronic fetal monitoring is to anticipate birth asphyxia by the diagnosis of progressive fetal hypoxia in labour before it has reached the stage of cardiovascular collapse, yet there is still uncertainty about the value of electronic fetal monitoring. There is confusion over the reliability of this clinical test, some investigators finding it reliable and others just the opposite. Ellen Blix et al. (pages 1–5) measured the reliability of electronic fetal monitoring by conducting an observer agreement study in 845 women who underwent an admission test. The authors compared the interpretations of the obstetricians and midwives caring for the women with the interpretations of two experts who separately examined the fetal heart rate traces. The agreement among the experts and between each expert and the obstetricians and midwives was poor, the kappa statistics being about 0.30. The conclusion of this study must be that electronic fetal monitoring is an unreliable test.
We cannot fault the conduct of this study, or its analysis by using the weighted kappa statistic. Close inspection of the data, however, may allow an alternative conclusion. Table 2 shows that Expert 1 judged 29 of the 845 fetal heart rate traces to be ominous, compared with 2 judged by Expert 2; and that Expert 1 judged 58 fetal heart rate traces to be equivocal, compared with 137 judged by Expert 2. There is a systematic bias between the experts in their interpretations of the fetal heart rate traces. This illustrates an important requirement in the conduct of an observer agreement study: not only should the experts be experienced and should interpret the fetal heart rate traces independently of each other, they should also undergo a period of training in order to agree the criteria of abnormality. This is because textbooks of electronic fetal monitoring define the criteria of the components of a fetal heart rate trace, such as accelerations and baseline variability, in ways that differ slightly but significantly. It may be that the poor agreement among the experts and between the experts and the obstetricians and midwives is due to the failure of the experts to agree on the criteria of abnormality before the study started. Paradoxically, the interpretations by the obstetricians and midwives are likely to give a more realistic frequency of pathological fetal heart rate traces, since any tendency of some obstetricians and midwives to overdiagnose ominous and equivocal traces will be balanced by a tendency of other obstetricians and midwives to underdiagnose ominous and equivocal traces. Certainly, the overall proportion of fetal heart rate traces that are ominous or equivocal is what one would expect in a group of unselected women admitted in spontaneous labour (46/845, 5%).
Analysis of the results of the study by the weighted kappa statistic is entirely appropriate, but this supposes that the assessments made by the observers are static, that is, they do not change. But in electronic fetal monitoring, the assessments are dynamic. A fetal heart rate trace thought to be equivocal may change as labour gathers momentum; observation, therefore, will resolve an equivocal trace into one that is reactive or one that is ominous. Much of the disagreement in the study is due to the number of equivocal traces. The really important disagreements are those in which one observer thinks that a trace is reactive which the other observer thinks is ominous, for on the one hand it may result in an unnecessary caesarean section, or on the other to severe fetal hypoxia being overlooked. But this degree of disagreement hardly ever happens. When Expert 1 is compared with Expert 2, it occurred on seven occasions in 845 fetal heart rate traces (0.8%); when Expert 1 is compared with the obstetricians and midwives, it occurred on 18 occasions (2%); and when Expert 2 is compared with the obstetricians and midwives, it occurred on 1 occasion (0.1%). It is entirely appropriate to analyse the results of the study by the weighted kappa statistic, but we must not ignore the clinical logic underlying these results. These clinical considerations suggest that the agreement may be better than is suggested by formal mathematical analysis.
One hindrance to research in electronic fetal monitoring is the language used to define abnormalities of the fetal heart rate. A major strength of Blix et al.'s study is that the terms ominous, equivocal and reactive are defined precisely, but this does not occur always in studies of electronic fetal monitoring. My Webster's New College Dictionary defines ominous as of or being an omen, especially an evil one; threatening; menacing. Unless the word ominous is defined in terms of abnormalities of the fetal heart rate, it is an inappropriate criterion to use in investigations of electronic fetal monitoring. By itself, use of this word will result in disagreement in the interpretation of fetal heart rate traces, for the observers will differ in their understanding of it. Another confusing term is non-reassuring. This participle implies that that there is a verb to non-reassure, which is nonsense. Imagine saying to a woman in labour, ‘I can non-reassure you that your fetal heart rate trace is abnormal’. Yet this word is widely used in clinical practice and in research in electronic fetal monitoring. Too often the interpretation of fetal heart rate traces in clinical practice descends into a recipe of rules for action, often accompanied by a system of mnemonics. In clinical practice, these rules and these systems are harmful, for they divert attention from the proper interpretation of the fetal heart rate trace, which involves understanding of the cardiovascular reflexes in the infant brought about by mechanical and biochemical disturbances during labour.
The proper introduction of a test into clinical practice involves estimation of its reliability by measurement of observer agreement; its efficiency, by measurement of its sensitivity, specificity, positive and negative predictive values and likelihood ratios; and its efficacy, by a randomised trial. Reliability, efficiency and efficacy are the three basic attributes of a clinical test, which are instrinsic to the test and which are inextricably intertwined. If a test is unreliable, it cannot be efficient and it cannot be efficacious. A test may be reliable, but if it is inefficient, it cannot be efficacious. And even if a test is reliable and efficient, it may not be efficacious, if a randomised trial shows that introduction of the test makes no difference to an important clinical outcome. Electronic fetal monitoring has been in clinical practice for several decades, but still we are uncertain about the first step in the evaluation of a clinical test, its reliability. The definitive study of the reliability of electronic fetal monitoring should have the following requirements: the observers should be experienced in the interpretation of fetal heart rate traces, should undergo a period of training to agree the criteria of abnormality, and should examine the traces independently; the fetal heart rate traces of infants who have been born with birth asphyxia should be compared with a random sample of those not born with birth asphyxia; the language used to interpret the traces should be unequivocal; and the agreement should be assessed both clinically and mathematically. Only then can we make authoritative statements about the reliability of electronic fetal monitoring.
Why case reports?
Case reports and small uncontrolled case series are considered the least scientific forms of clinical research, yet we continue to publish them. In our view, case reports and case series serve a useful function by suggesting hypotheses that may be tested by more formal methods. In practice, however, case reports and uncontrolled case series may influence clinical practice to a degree out of all proportion to their scientific worth. We have published two small case series on the treatment of postpartum haemorrhage by the B-Lynch braces1 and the Rusch balloon2, which we believe have influenced clinical practice greatly and possibly have prevented many a caesarean hysterectomy. Our philosophy regarding case reports and uncontrolled case series is that they should describe a new treatment, a new clinical event or an important new complication of a treatment. We are unable therefore to publish case reports of clinical events that are not original, unless we think that these events will be unknown to obstetricians and gynaecologists. The staff in the Journal office tease us by saying that we will publish any case report with the word necrotising in the title or if it accompanied by a gruesome picture; not so, as this month's five case reports attest. Selective termination of a twin pregnancy in severe pre-eclampsia (Francois Audibert et al., pages 68–69), arterial embolisation to treat choriangioma of the placenta (Tze Kin Lau et al., pages 70–73), neonatal gangrene in the infant of a diabetic mother (A. Moazzam et al., pages 74–76), ovarian failure caused by a herbal medicine (Simon Edmonds and Julia Montgomery, pages 77–78), and perineal pain due to the tension-free vaginal tape (Paul Hilton et al., pages 79–82) are all examples of new treatments, new clinical events or new complications of a treatment. These case reports suggest hypotheses that can be tested in formal clinical studies, and in their own right may result in the adoption of treatments or the avoidance of treatments. Roll on case reports!