Letters

Cardiotocography v Doppler auscultation

Guidelines highlight gaps in research evidence

Conclusions do not recognise difference between statistical and clinical significance

All unbiased comparative studies should be published

Trials are underpowered

Evidence is not as good as NICE suggests

Author of editorial's reply

Authors of paper's reply

Guidelines highlight gaps in research evidence

EDITORThe clinical effectiveness support unit of the Royal College of Obstetricians and Gynaecologists has recently developed the guideline of the National Institute for Clinical Excellence (NICE) on the use of electronic fetal monitoring.¹ One of the potential benefits of guidelines is to highlight gaps in research evidence. As part of the guideline development process we conducted a survey of maternity units in England and Wales to evaluate the current use of electronic fetal monitoring in intrapartum care.²

The guideline recommends that in women who are healthy and have an uncomplicated pregnancy, intermittent auscultation is a suitable method of monitoring during labour. This recommendation was based on the evidence from nine randomised controlled trials that have enrolled 18 000 and are included in three systematic reviews.¹ This conclusion was reached because screening with continuous electronic fetal monitoring should not be conducted in the absence of evidence of benefit. This recommendation will perhaps have a minimal impact in practice as the survey found that only 2.5% of units report using continuous electronic fetal monitoring for all women in labour.

Most maternity units (79%), however, reported using admission cardiotocography in the initial assessment of women who are healthy and have had an otherwise uncomplicated pregnancy. When evaluating the research that exists to support the use of admission tests in populations at low risk we were surprised to find only a single cohort study that addressed this issue; this cohort study included only 1041 women and related admission tests to fetal distress.³ The trial by Mires et al is the first published randomised controlled trial to address this clinically important question. The study is also nearly four times the sample size of previous studies. As such it would have been important to publish this trial irrespective of whether the results were positive, negative, or equivocal.⁴

The results of the trial are in line with those of the earlier cohort study and do not change the recommendation in the guideline. We are aware of unease in some quarters that the guideline did not recommended admission cardiotocography and that this recommendation does not concur with current practice in many maternity centres. We hope that this knowledge will act as an incentive to researchers and clinicians to conduct and enrol patients in randomised controlled trials to answer this important question.

Jane Thomas, director. 
Shantini Paranjothy, clinical research fellow. 
Tony Kelly, clinical research fellow. 
Josephine Kavanagh, research fellow. 
Royal College of Obstetricians and Gynaecologists, Clinical Effectiveness Support Unit, London NW1 4RG

Conclusions do not recognise difference between statistical and clinical significance

EDITORMires et al conclude that admission cardiotocography does not benefit neonatal outcome in women at low risk, but that it does increase the rate of operative delivery.¹ No information is given regarding the classification of the admission traces in their study. Other electronic responses, by Murphy and Chalmers (published as the next letter in this cluster), ^{2 3} have addressed the likely cause of this increased operative delivery rate and the suggestion that lack of effect can be proved.

An earlier study, although not randomised, was a prospective blinded study of women designated as low risk on admission in labour and included a similar number of women to the cardiotocography arm of the current study once defined as low risk at labour onset (1041 compared with 1186).⁴ This study showed a 1% incidence of an "ominous" fetal heart rate trace on admission testing and, of these 10 infants, one died three hours after admission, without stethoscopic auscultation detecting any fetal compromise.

This again highlights the point made by Chalmers (next letter).³ The conclusions drawn from the current study do not recognise the difference between statistical and clinical significance in an individual case, particularly when the adverse outcome is rare. Intermittent auscultation will not detect the development of all abnormalities in fetal heart rate and depends on the ability to provide adequate levels of midwifery care. The inability to take the cardiotocography off after an admission test is as much a training issue as the inability to recognise as abnormal the cardiotocography identified by the Confidential Enquiry into Stillbirths and Deaths in Infancy.⁵

Simon Grant, consultant in obstetrics and gynaecology. 
Royal Cornwall Hospital (Treliske), Truro TR1 3LJ simon.grant{at}rcht.swest.nhs.uk

All unbiased comparative studies should be published

EDITORGoldbeck-Wood's invitation to respond to questions raised by the first randomised comparison of cardiotocography versus auscultation of the fetal heart at admission in labour prompts me to emphasise the following points.¹ The term "negative trial" should be outlawed, as should editorial acquiescence in statements implying that it is possible to prove negatives.² For example, the BMJ report of the admission cardiotocography trial states that this intervention does not benefit neonatal outcome in women at low risk, yet the estimate of its effect on the incidence of hypoxic ischaemic encephalopathy is compatible with the reduction in neonatal seizures found in other randomised controlled trials of fetal monitoring during labour.³

Although it is worth making pretrial estimates of the sample sizes needed to rule in or rule out, with specified statistical confidence, effects of a magnitude likely to be relevant to patients, how many participants you needed in a trial depends on what you found, not on what you thought you would find.⁴

Statements that it is unethical to embark on controlled trials unless an arbitrarily defined level of statistical power can be assured make no sense if the alternative is acquiescence in ignorance of the effects of healthcare interventions. Unbiased estimates should be generated through randomised trials because they leave us less ignorant than we were before.⁵ The importance of the admission cardiotocography study is that it provides the first unbiased estimates of the ranges within which the effects of this healthcare intervention are likely to lie. All unbiased comparative studies should be published, so that the totality of the relevant evidence can be evaluated in the systematic reviews that are needed by those making choices and decisions in health care and about further research. It is now possible to register and publish successive versions of trial protocols and all subsequent reports of registered studies. Reluctance to exploit this potential does not reflect scientific or technical problems but rather academic and commercial interests.

Iain Chalmers, director. 
UK Cochrane Centre, Oxford OX2 7LG ichalmers{at}cochrane.co.uk

Trials are underpowered

EDITORGoddard in her editorial acknowledged the fact that trials of routine electronic fetal monitoring are underpowered, but she did not draw the logical conclusion.¹ With her figures for intrapartum stillbirth and cerebral palsy of intrapartum origin, a trial powered (80% power, alpha 5%, two-sided) to show a 50% decrease in intrapartum stillbirth would need to recruit 83 500 women to each arm, and a trial powered to detect a 50% decrease in cerebral palsy due to intrapartum causes would require 670 000 women in each arm. The current meta-analysis has less than 19 000 cases in total.² The evidence is not, as she states, strongly against the routine use of electronic fetal monitoring. The evidence simply does not exist and, given the numbers required, it probably never will.

Routine electronic fetal monitoring halves the risk of neonatal seizures even when applied to a low risk population.² Statements to the effect that electronic fetal monitoring has not been shown to reduce the risk of intrapartum stillbirth and cerebral palsy have the same fundamental weaknesses as statements associated with some previous public health disasters, such as there is no evidence to suggest that blood products can transmit HIV or that there is no evidence that bovine spongiform encephalopathy can be transmitted to humans. Falsely equating absence of evidence with evidence of absence does not have a glorious history and should be avoided.

Gordon C S Smith, subspecialist trainee, maternal-fetal medicine. 
Department of Obstetrics and Gynaecology, The Queen Mother's Hospital, Glasgow G3 8SH gcs4{at}cornell.edu

Evidence is not as good as NICE suggests

EDITORThe recently published clinical guideline from the National Institute for Clinical Excellence (NICE) on the use of electronic fetal monitoring and the clinical guideline from the Royal College of Obstetricians and Gynaecologists from which the NICE guideline is derived recommend that intermittent auscultation is the most appropriate method of fetal monitoring for low risk women in labour. ^{1 2}

The guideline development groups erroneously conclude that there is grade A evidence (at least one randomised controlled trial addressing the specific recommendation) to support intermittent auscultation every 15 minutes in the first stage of labour and every 5 minutes in the second stage. The College guideline, however, states that there are no studies evaluating different protocols for frequency of intermittent auscultation. Any recommendation on frequency of auscultation has to be grade C (evidence from expert committee reports or opinions or clinical experience of respected authorities, or both). This is an important point because the NICE guideline is likely to be held by the courts as the required standard of care in labour. This degree of fetal monitoring may be impossible to achieve in hard pressed labour wards and may encourage the use of continuous electronic fetal monitoring instead.³ For women in our unit who are at low risk we are currently auscultating the fetal heart every 30 minutes in the first stage and after every contraction in the second stage of labour.

I P Stuart, consultant in obstetrics and gynaecology. 
Diana, Princess of Wales Hospital, Grimsby DN33 2BA

Author of editorial's reply

EDITORThe publication of the paper by Mires et al and my accompanying editorial have resulted in a wide range of responses from both sides of the argument. This issue has long been controversial and, as indicated in the letters by Smith and Grant, is likely to remain so as it may prove impossible to carry out a large enough trial with sufficient power to answer the key question: Can electronic fetal monitoring either on admission or during labour reduce the risk of poor obstetric outcome by identifying those infants who are stressed by labour where early intervention will improve the outcome?

Unfortunately, cardiotocography is a poor screening test, especially in low risk women, which has been widely introduced into practice without the evidence for its value. There are wide variations in interpreting traces. It is also a test that is highly sensitive for a rare disease but has a low specificity.¹ As a consequence, the false positive rate is high, leading to a high intervention rate. Although many obstetricians such as Smith and Grant feel that any increase in caesarean rate is justified if there is perceived to be a neonatal benefit, many others disagree particularly where epidemiological data would suggest that there has been no effect on mortality. In particular, midwives, pregnant women themselves, and other professionals who may have to deal with the long term effects of our management are dismayed at the escalating intervention rate that has occurred as a consequence of using electronic monitoring in low risk women. ^{2 3}

There is no grade A evidence for the form of intermittent auscultation. There are randomised controlled trials of electronic fetal monitoring versus intermittent auscultation but no up to date studies comparing different methods of intermittent auscultation.¹ The aim of the guideline was to look at electronic fetal monitoring and not intermittent auscultation. The recommendations on the form of intermittent auscultation used are therefore based on randomised controlled trials comparing electronic fetal monitoring and intermittent auscultation not different methods of intermittent auscultation. Many units do not have the staffing levels to achieve the level of auscultation suggested by the guideline. Thomas et al rightly say that a properly formed trial of which is the best method of intermittent auscultation needs to be done. Although the paper by Mires et al is flawed, it has stimulated this important debate, which will hopefully provide the impetus for a larger trial to answer this important question.

Ros Goddard, specialist registrar. 
Royal United Hospital, Bath BA1 3NG ros_goddard{at}hotmail.com    A longer version of this letter is published on bmj.com

Authors of paper's reply

EDITORThomas et al recognise the importance of our study as the first randomised trial of the value of the admission cardiotocograph in low risk labour. Despite the paucity of evidence to support its use, some 80% of obstetric units in the United Kingdom have introduced the admission cardiotocograph into routine clinical practice. The evidence based guideline from the Royal College of Obstetricians and Gynaecologists on the use of electronic fetal monitoring did not recommend the use of the admission cardiotocograph, and the results of our trial are in line with this recommendation.¹

Grant and Chalmers raise concerns about our conclusion that the admission cardiotocograph does not benefit neonatal outcome in women at low risk. Our conclusion relates to the presence of metabolic acidosis at delivery, which was our primary outcome. Umbilical arterial acid base status is one of the recognised intermediate fetal or neonatal measures of fetal hypoxia and as such is an appropriate surrogate for fetal condition at birth.¹

We acknowledge that our sample size was not adequate to provide definitive information on secondary outcomes such as hypoxic ischaemic encephalopathy and that our findings are compatible with a reduction in this hypoxic complication. Trials to evaluate this and other uncommon outcomes would need to be much larger. Chalmers indicates that our trial provides the first unbiased estimates of the ranges within which the effects of this healthcare intervention are likely to lie. Our data will therefore inform future sample size calculations.

Grant refers to a previous study in which the only intrapartum death was in a woman with an ominous admission test in which intermittent auscultation did not identify a problem with the fetal heart pattern.² He raises concerns about the difference between statistical and clinical significance in an individual case, particularly when the adverse outcome is rare. The trial by Ingemarsson et al was not randomised, and the admission test was not available to the clinician managing the labour, with the interpretation and classification of traces being retrospectively assigned. It cannot be assumed that if the admission test had been available to the clinician that the outcome would have been different.

It is feasible that when the admission test was performed that the fetus could have already suffered irreversible hypoxic damage. In addition few details are given about this baby, which was in a low risk group defined in the study as patients recruited in the first stage of labour with a pregnancy of 34 weeks or more, practically all of whom had attended antenatal clinics. Our study showed a significantly higher incidence of operative delivery in the admission test group. As well as the fetal outcome, maternal outcome needs to be considered. Caesarean section is associated with an increased risk of maternal death compared with vaginal delivery, and both caesarean section and deliveries by forceps or ventouse are associated with increased risk of severe maternal morbidity. ^{3 4} This is in addition to the increased potentially serious neonatal morbidity associated with forceps and ventousefor example, cephalhaematoma.⁴

We recognise that further trials need to be undertaken to confirm our findings, and provide more evidence for the use of the admission cardiotocograph. In association with Thomas et al, we hope that the results of our trial will act as an incentive to researchers and clinicians to conduct and enrol patients into randomised trials to further evaluate the role of the admission cardiotocograph.

G J Mires, senior lecturer, department of obstetrics and gynaecology. 
F L R Williams, lecturer, department of epidemiology and public health. 
P W Howie, professor, department of obstetrics and gynaecology. 
Ninewells Hospital and Medical School, Dundee DD1 9SY