Discussion Questions

• ■
  

  Why is this topic important?

  
  
• ■
  

  Might selection bias have affected this study?

  
  
• ■
  

  Why is it important to identify the direction of potential bias?

  
  
• ■
  

  What are the pros and cons of interpreting tracings with computer analysis?

  
  
• ■
  

  Would you have included women who delivered by cesarean section?

  
  
• ■
  

  Should we start using this 5-tier system in practice?

  
  
• ■
  

  Are the authors’ conclusions congruent with their findings?

  
  
• ■
  

  What are future directions for research?

Introduction

It is no secret to obstetricians that intrapartum electronic fetal monitoring (EFM) is the most commonly used instrument in obstetrics. What continues to surprise us is the paucity of available data associating EFM patterns with clinically relevant outcomes, even though the technology is employed daily to manage patients in labor.

While it is widely accepted that EFM has the potential to alert physicians to signs of fetal compromise, allowing timely intervention and the prevention of permanent neonatal morbidity, rigorous investigation is needed before this is proven to be true. A new study by Elliot et al highlights the importance of further scrutiny.

Alison G. Cahill, MD, MSCI and George A. Macones, MD, MSCE, Associate Editor

Study Design

Cahill: Why is this topic important?

O’Neill: The direct costs of EFM, which has become an integral part of labor management in approximately 85% of live births, exceed $150 million annually. As a result, it is valuable to investigate this tool’s efficacy, effectiveness, and systems for interpretation of results. This study and many of the references it lists aim to assess the clinical importance of the information generated by EFM and to explore suitable management algorithms. The study of this widely used modality is particularly important, since research done to date has shown that in comparison to intermittent auscultation, continuous EFM increases the overall cesarean and operative delivery rates but does not reduce perinatal mortality or cerebral palsy rates. As obstetricians, we rely on EFM even though the limited data we have do not show that its use provides a clear benefit over intermittent monitoring.

Cahill: Can you remind us of what selection bias is?

Stout: Selection bias refers to a statistical bias introduced when individuals chosen to participate in a study share an inherent confounding characteristic that might distort the study results. For example, it may appear that people who jog as rehabilitation from a heart attack have better outcomes. Yet, there may be something inherently different—diet, bodyweight, physical conditioning, and so on—about those patients willing to participate in a jogging program.

Cahill: Is there a chance that selection bias may have affected this study?

Tuuli: There is evidence in the design of this study to suggest that selection bias may have affected the results. First, cases with metabolic acidosis and encephalopathy were selected from hospital series and medicolegal files, while the reference groups were convenience samples of consecutive vaginal births from an urban teaching hospital. It is likely that patients whose labor and delivery record made it into the hospital series or resulted in medicolegal action differed significantly from those who did not meet those criteria. Indeed, those cases are likely to be the worst of the worst.

Second, since an important factor influencing neonatal outcome is the physician’s response to information from EFM, applied interventions must be comparable among cases and controls in order to produce valid results. Since these cases and controls were selected from different institutions, it is possible that they differed in terms of interventions selected in response to abnormal EFM patterns. And, because they ended up in a hospital series or as medicolegal cases, some of the cases might have been managed inappropriately.

Third, while the authors indicate that controls were delivered vaginally, there is no information on mode of delivery for the cases. If the cases included cesarean deliveries, it is likely that patients who required cesarean delivery, especially emergency procedures for nonreassuring fetal tracings, differed significantly from those who had a vaginal delivery.

Cahill: Those are all great points. Going back to the second point, do you think that you can describe how interventions might have differed between patients with EFM abnormalities?

Tuuli: In most institutions, standard management for abnormal EFM patterns includes repositioning the patient; administration of oxygen and often, fluids; ruling out common causes; and at times, administration of a uterine relaxant to give the fetus some relief. It is possible that different institutions used these measures to different extents.

Cahill: How might selection bias have occurred with respect to the primary outcome of metabolic acidosis?

Tuuli: The mode of selection of cases and controls also has potential for selection bias in terms of the outcome of metabolic acidosis. The authors created 3 restrictive categories: metabolic acidosis with encephalopathy; metabolic acidosis without encephalopathy; and neither metabolic acidosis nor encephalopathy. Subjects were selected from different populations. For the reasons discussed above, this had the potential of distorting the relationship between metabolic acidosis and EFM patterns.

Cahill: Why is it important to identify the direction of potential bias?

Tuuli: Bias can be positive or negative. Positive bias tends to increase the association between a factor and an outcome, while negative bias tends to decrease the association. Identifying the direction of potential bias is important for correct interpretation of study results and assessment of validity. For example, if potential bias in a study is positive, an observed association could be due solely to the bias and therefore cannot be considered necessarily valid. Even if the results are valid, the measures of association must be considered inflated. On the other hand, a potential bias may produce valid results if an association is found between a factor and an outcome despite the presence of negative bias. In that case, the measure of association obtained should be considered an underestimation of the true value.

Cahill: What is the direction here?

Tuuli: The potential bias in this study is likely positive, since the manner in which the cases were selected would tend to increase the association between encephalopathy and metabolic acidosis and patterns of EFM. The selection bias would result in cases that are more closely associated with patterns of EFM than are controls. Thus, the strength of the association is increased if a true association exists—or even if a spurious association was inadvertently created. In this study, selection bias, if present, might have inflated a real association between patterns of EFM and encephalopathy and metabolic acidosis. Or worse, it might have produced an association when there is none.

Cahill: How important is it to consider neonatal neurologic outcomes when studying EFM?

Shanks: It is imperative to consider neonatal neurologic outcomes when studying EFM. The very goals of fetal heart rate assessment are to recognize fetuses where timely intervention would prevent death and avoid adverse neurologic injury.

In this study, patients were classified as abnormal, normal, and intermediate, based on umbilical artery gases and whether they had signs of neurologic encephalopathy. The smallest number of patients was categorized as “abnormal,” given the low incidence of this condition. Again, patients in this group had umbilical artery base deficits exceeding 12 mmol/L, and they developed neurologic signs of encephalopathy in the early neonatal period. The authors culled their patients from reported cases in hospital series (n = 40) and medicolegal files (n = 20). The selection bias that results from this may be problematic. Patients with good outcomes despite poor tracings may not be reported, and therefore the authors may have overstated the association between poor tracings and adverse neonatal outcomes.

The American College of Obstetricians and Gynecologists (ACOG) defines an acute intrapartum hypoxic event as sufficient to cause cerebral palsy if it has all 4 of the following criteria: evidence of metabolic acidosis (pH <7 and base deficit >12 mmol/L); early onset of severe or moderate neonatal encephalopathy in infants born at 34 or more weeks of gestation; cerebral palsy of the spastic quadriplegic or dyskinetic type; and exclusion of other identifiable etiologies, such as trauma, infection, and genetic disorders. The authors did a reasonable job of matching their definition of encephalopathy with the criteria set forth by ACOG. Babies with congenital malformations and inborn errors of metabolism were appropriately omitted. However, they did not report pH at delivery or account for infectious etiologies.

Cahill: What are the strengths and weaknesses of the study design?

Shanks: This is essentially a case-control study, which has inherent strengths and weaknesses. It permits us to study conditions where the outcome of interest is rare, and the retrospective nature allows us to accomplish this in an inexpensive and efficient way. Conversely, a potential weakness is the bias and confounding that can result from sampling 2 different populations.