1. Top of page
  2. Introduction
  3. Electronic fetal monitoring and neonatal outcome
  4. The intervention-benefit ratio and research models
  5. References

New technologies generate much optimism at their inception, but this is often followed by disappointment and much criticism. This pattern is exemplified by the introduction of continuous electronic fetal monitoring (EFM) in obstetric care1–8. The scientific reasoning underlying this technology was well-founded and the initial results appeared favourable2. Individual randomised trials, however, failed to demonstrate a significant benefit, and have led many to question the value of the technology9. It was only with meta-analysis of many studies10 and in a large-scale trial5, that a reduction in neonatal seizures following prolonged, augmented or induced labours could be demonstrated. Even so, a reduction in other important outcome measures, such as residual neurological disability, remain unproved.

This lack of demonstrable benefit has provoked wide criticism of this technology from the community, mid-wives and obstetricians. It was pointed out that whenever EFM is introduced, there is an increase in obstetric intervention without any reduction in morbidity. On analysis however, this criticism appears at least as fallacious as the initial optimism. This commentary attempts to put in perspective the relationship between intervention and benefit for conditions of very low prevalence.

Electronic fetal monitoring and neonatal outcome

  1. Top of page
  2. Introduction
  3. Electronic fetal monitoring and neonatal outcome
  4. The intervention-benefit ratio and research models
  5. References

A variety of neonatal outcome parameters have been used extensively in the assessment of EFM. These include Apgar scores, umbilical artery pH at birth, need for intubation, neonatal seizures, ischaemic encephalopathy and perinatal mortality11. Table 1 shows the prevalence of these outcomes in developed countries. Although their correlation with significant long term morbidity is poor except at the extreme of their ranges, they nevertheless reflect the physical and metabolic condition of the baby at birth.

Table 1.  Prevalence of common indicators of obstetric intervention for fetal distress (without EFM) and neonatal outcome. The reference source is given under each heading
IndicatorsPrevalence (%)
 Caesarean section230.2–9.1
Neonatal outcome 
 Apgar score < 5 at 5 rain240-90
 Umbilical artery pH < 7-0250-65
 Intubation at birth250-50
 Neonatal seizures250-20
 Perinatal mortality attributable to hypoxia260-18
 Ischaemic encephalopathy240-06

While the sensitivity and specificity of antepartum EFM to detect adverse fetal outcomes has been extensively evaluated, the data on intrapartum EFM is limited (Table 2). In short, EFM is characterised by a relatively high sensitivity but a low specificity, and the human element in the interpretation has an effect on its accuracy12,13. This, in combination with the low prevalence of poor perinatal outcome, results in a high false positive rate leading to a high intervention rate.

Table 2.  Sensitivity and specificity of electronic fetal monitoring
MethodOutcomeSensitivity (%)Specificity (%)
Intrapartum EFM27LowApgar54-8444-93
Intrapartum EFM28Low Apgar + low cord pH8185
Intrapartum EFM29CordpH <7-157682
Antepartum EFM30Low Apgar, low cord pH, neonatal ITU admission2-9037-100

The intervention-benefit ratio and research models

  1. Top of page
  2. Introduction
  3. Electronic fetal monitoring and neonatal outcome
  4. The intervention-benefit ratio and research models
  5. References

The positive predictive value (i.e. the proportion of those tested positive with the predicted outcome) can be derived from test sensitivity, specificity and disease prevalence14.

If intervention occurs whenever a test is positive, the intervention rate is equivalent to the test positive rate. However, such an intervention would only be beneficial if the test result is a true positive. The intervention-benefit ratio (IBR) can therefore be defined as the inverse of the positive predictive value. In the context of EFM the IBR represents the number of interventions that would be necessary to prevent a case of adverse neonatal outcome:

  • image

Although the sensitivity and specificity of a test tend to be stable, its usefulness—that is the IBR—depends on the prevalence of the adverse outcome in a particular population. Figure 1 demonstrates this relationship for an assumed sensitivity of EFM of 0.9 at four levels of specificity; Fig. 2 shows the same, for an assumed specificity of 0.9 at four levels of sensitivity. From these, it appears that variations in test specificity have a greater effect on the IBR than variations in sensitivity. The relationship between the IBR and decreasing prevalence can be seen to be exponential. This means that, for conditions of very low prevalence, a greater increase in intervention is necessary to achieve the same reduction in morbidity. Assuming a sensitivity and specificity of 0.9 (Tables 1 and 2), the use of EFM needs to increase the number of operative deliveries by 13 for each case of low Apgar score prevented, and by 186 to prevent a case of encephalopathy (Table 3).


Figure 1. The intervention-benefit ratio for EFM is plotted against prevalence of adverse neonatal outcome, assuming a sensitivity of 90% at four levels of specificity.

Download figure to PowerPoint


Figure 2. The intervention-benefit ratio for EFM is plotted against prevalence of adverse neonatal outcome, assuming a specificity of 90% at four levels of sensitivity.

Download figure to PowerPoint

Table 3.  Intervention to benefit ratios for electronic fetal monitoring, assuming a sensitivity of 90% at three levels of specificity.
Outcome variable90%70%50%
Apgar score < 5 at 5 min133862
Umbilical artery pH < 7.0185286
Intubation at birth2367112
Neonatal seizures56167278
Perinatal mortality attributable to hypoxia63186309
Ischaemic encephalopathy186556926

In addition to the disparity between intervention and benefit, the wide discrepancy between the rate of obstetric intervention and perinatal morbidity (Table 1) constitutes a pitfall for researchers designing studies which examine the relationship between these two parameters. Figure 3 shows the number of cases required in each arm of a study to detect changes in interventions and outcomes, using a 0.8 power and a type 1 error of 0.0515. To detect a doubling or halving of the caesarean section rate from 10% requires 220 and 475 cases in each arm, respectively. However, a similar doubling or halving in the incidence of low Apgar scores from around 0.9% would require 2799 and 5634 cases, and for encephalopathy 85,012 and 42,488 cases. It can be seen that nearly all reported EFM studies were only capable of detecting an increase in intervention, but not a reduction in morbidity, even if this did occur3,4,8,16–20.


Figure 3. The sample size per arm of study to detect changes in interventions and outcomes is plotted against prevalence, using a 0.8 power and a type 1 error of 0.05.

Download figure to PowerPoint

Currently, there is an increase in emphasis on evidence-based medicine, especially so in obstetrics10. Our analysis demonstrates that the flaws in much of the evidence has given special prominence to the negative aspects of EFM, whereas the possible benefits have been obscured by studies with insufficient numbers. This has provoked a drive to limit the use of EFM in labour10 before its potential is fully explored.

Even in a meta-analysis10, and in the trial conducted in Dublin5, the size of the studies would still have been insufficient to easily detect a reduction in immediate neonatal morbidity (Fig. 3). That both studies did so suggests that the sensitivity and specificity of EFM may be better than described in the literature.

A general principle is that, for conditions of very low prevalence, a new test or management policy can be shown to reduce intervention without showing rise in morbidity, if the study is small. In other words, it may be possible to reduce the intervention-benefit ratio by combining two tests, but this will decrease the sensitivity. This is well demonstrated in two small studies21,22 when—in addition to EFM—an additional positive test was required to make the diagnosis of fetal distress. These were able to demonstrate an apparent decrease in intervention without a significant increase in morbidity, but again the trials were too small to be able to comment on morbidity.

The separation of the intervention-benefit ratio and sensitivity also explains the apparent success of many modem hospital management practices. Significant reductions in services and therefore costs can be rapidly demonstrated before rises in adverse outcomes become apparent, simply because there is a great difference in the prevalence and hence the time and sample sizes required to detect these two different aspects of a service.


  1. Top of page
  2. Introduction
  3. Electronic fetal monitoring and neonatal outcome
  4. The intervention-benefit ratio and research models
  5. References
  • 1
    Quilligan E, Paul RH. Fetal monitoring: is it worth it Obstet Gynecol 1975; 45:96100.
  • 2
    Shenker L, Post RC, Seiler JS. Routine electronic monitoring of the fetal heart rate and uterine activity during labour. Obstet Gynecol 1976; 46: 185189.
  • 3
    Tutera G, Newman RL. Fetal monitoring; its effect on perinatal mortality and caesaeran section rates and its complications. Am J Obstet Gynecol 1975; 122: 750754.
  • 4
    Amato JC. Fetal monitoring in a community hospital. A statistical analysis. Obstet Gynecol 1977; 50: 269274.
  • 5
    Macdonald D, Grant A, Sheridan-Pereira M, Boylan P, Chalmers I. The Dublin randomised controlled trial of intrapartum fetal heart rate monitoring. Am J Obstet Gynecol 1985; 152: 524539.
  • 6
    Neilson JP. Cardiotocography during labour. An unsatisfactory technique but nothing better yet. BMJ 1993; 306: 347348.
  • 7
    Umstad MP. The predictive value of abnormal fetal heart rate patterns in early labour. Aust N Z J Obstet Gynecol 1993; 33: 145149.
  • 8
    Vintzileos AM, Antsaklis A, Varvarigos I, Papas C, Sofatzis I, Montgomery JT. A randomised trial of intrapartum electronic fetal heart rate monitoring versus intermittent auscultation. Obstet Gynecol 1993; 81: 899907.
  • 9
    Prentice A, Lind T. Fetal heart rate monitoring during labour—too frequent intervention, too little benefit Lancet 1987; 2: 13751377.
  • 10
    Neilson JP. EFM + scalp sampling vs intermittent auscultation in labour. In: NeilsonJP, CrowtherCA, HodnettED, HofmeyrGJ, KeirseMJNC, editors. The Cochrane Pregnancy and Childbirth Database; Issue 1; 1995. Oxford : Update Software, 1995.
  • 11
    SpencerJAD, WardRHT, editors. Intrapartum Fetal Surveillance. London : RCOG Press, 1993.
  • 12
    Nielsen PV, Stigsby B, Nickelson C, Nim J. Intra and inter-observer variability in the assessment of intrapartum cardiotocograms. Acta Obstet Gynecol Scand 1987; 66: 421424.
  • 13
    Donker DK. Inter-observer variation in the assessment of the fetal heart recordings. Eur J Obstet Gynecol Reprod Biol 1993; 52: 2128.
  • 14
    Beck GJ. The clinical application of epidemiology and the interpretation of diagnostic tests. In: MatzenRN, SlangR, editors. Principles of Preventive Medicine and Clinical Prevention. St Louis , Missouri : Mosby, 1993:106.
  • 15
    Sokel RR, Rolf FJ. Biometry. New York : WH Freeman & Co, 1981: 765.
  • 16
    Renou P, Chang A, Anderson 1, Wood C. Controlled trial of fetal intensive care. Am J Obstet Gynecol 1976; 126:470476.
  • 17
    Haverkamp AD, Orleans M, Langendoerfer S, McFee J, Murphy J, Thompson HE. A controlled trial of the differential effects of intra- partum fetal monitoring. Am J Obstet Gynecol 1979; 134: 399412.
  • 18
    Wood C, Renou P, Oats J, Farrell E, Beischer N, Anderson I. A con- trolled trial of fetal heart rate monitoring in a low-risk obstetric popu- lation. Am J Obstet Gynecol 1981; 141: 527534.
  • 19
    Neldam S, Osier M, Hansen PK, Nim J, Smith SF, Hertel J. Intrapartum fetal monitoring in a combined low- and high-risk popu- lation: a controlled clinical trial. Ear J Obstet Gynecol Reprod Biol 1986; 23: 111.
  • 20
    Luthy DA, Shy KK, Van Belle G, Larson EB, Hughes JP, Benedetti TJ, Brown ZA, Effer S, King JF, Stenchever MA. A randomised trial of electronic fetal monitoring in preterm labor. Obstet Gvnecol 1987; 69:687695.
  • 21
    Beard RW, Morris ED, Clayton SG. pH of capillary blood as an indi- cator of the condition of the fetus. J Obstet Gynaecol Br Comnwlth 1974; 74: 812822.
  • 22
    Westgate J, Harris M, Curnow JSH, Greene KR. Plymouth ran- domised trial of cardiotocography alone versus ST waveform analysis plus cardiotocography for intrapartum monitoring in 2400 cases. Am J Obstet Gynecol 1993; 169: 11511160.
  • 23
    Thacker SB. Effectiveness and safety of intrapartum fetal monitoring. In: SpencerJAD. Fetal Monitoring. Castle House Publications, 1989: 211217.
  • 24
    Ellenberg JH, Nelson KB. Cluster of perinatal events identifying infants at high risk for death or disability. J Pediatr 1988; 113: 546552.
  • 25
    Gilstrap LC, Leveno KJ, Bums J, Williams ML, Little BB. Diagnosis of birth asphyxia on the basis of fetal pH, apgar score, and newborn cerebral dysfunction. Am J Obstet Gynecol 1989; 161: 825.
  • 26
    Vintzileos AM, Nochimson DJ, Guzman ER, Knuppel 10. RA, Lake M, Schifrin BS. Intrapartum electronic fetal heart monitoring versus intermittent auscultation: A meta-analysis. Obstet Gynecol 1995; 85: 149155.
  • 27
    Banta DH, Thacker SB. Assessing the costs and benefits of electronic fetal monitoring. Obstet Gynecol Surv 1979; 34: 627642.
  • 28
    Murphy KW, Johnson P, Moorcraft J, Pattison R, Russell V, Turnbull SA. Birth asphyxia and the intrapartum cardiotocograph. Br J Obstet Gynaecol 1990; 97: 470479.
  • 29
    Chung TKH, Mohajer MP, Yang ZJ, Chang AMZ. The pre- diction of fetal acidosis at birth by computerised analysis of intrapartum cardiotocography. Br J Obstet Gynaecol 1995; 102: 454460.
  • 30
    Devoe LD. The nonstress test. In: EdenRD, BoehmFH. Assess- ment and Care of the Fetus. Prentice-Hall International Inc, 1990: 376.