Alex C. Vidaeff, MD, MPH

Gyamfi-Bannerman and Son report their secondary analysis of a randomized controlled trial (RCT) conducted through the Eunice Kennedy Shriver National Institute of Child Health and Human Development’s Maternal-Fetal Medicine Units Network. The aim of the parent RCT was to determine whether antenatal administration of magnesium sulfate decreases the rate of cerebral palsy or death in children delivered preterm. More than 80% of the women enrolled in this RCT had preterm PROM. Of these women, 98% were given antenatal corticosteroids for fetal maturation, and 9% received two courses. This aspect of the RCT provided an opportunity for Gyamfi-Bannerman and Son to study the comparative effects of one course versus two courses of ACS in the setting of preterm PROM.

Background of the study
Concern about the declining efficacy of ACS when the interval between administration and delivery exceeds 7 days prompted several randomized trials exploring the safety and efficacy of multiple ACS courses. Multiple courses were ultimately deemed to be inadvisable because of an association with reduced birth weight and neonatal head circumference.^1,2 However, the unfavorable effects of ACS on anthropometrics was observed when more than three courses were administered, leaving open the possibility of giving only one additional course when needed as a “rescue” dose. Indeed, the use of a single rescue course of ACS in women with intact membranes had a favorable impact on neonatal respiratory function in two RCTs.^3,4

The best available evidence (Level 1) demonstrates that the use of a single course of ACS in preterm PROM does not increase the risk of neonatal or maternal infection even in the setting of prolonged rupture of membranes.⁵ However, pregnant women with preterm PROM were excluded from trials of repetitive ACS dosing because earlier observational studies had suggested that they would experience a substantially increased risk of infectious morbidity when three or more courses of ACS are given.^6–8 What remained debatable on a scientific level was whether a single rescue dose of ACS in women with preterm PROM would be safe and beneficial.

Findings of the analysis
Compared with a single course of ACS, exposure to two courses did not influence the rate of neonatal sepsis or chorioamnionitis. As reassuring as that finding may be, the study found no benefit for the additional course, although it was powered to do so. There was no difference in the rates of respiratory distress syndrome between the study groups.

The findings of Gyamfi-Bannerman and Son replicate those of a subgroup analysis of women in an RCT comparing weekly and single-course ACS.⁹ In that study, weekly courses of ACS in women with preterm PROM did not improve neonatal outcomes beyond what was achieved with single-course therapy. Similar findings have been reported by the Cochrane database.¹⁰

Ruptured versus intact membranes: When is the benefit of ACS greater?
The improvement in neonatal outcomes observed with ACS in pregnancies complicated by preterm PROM is not as pronounced as it is in gestations with intact membranes. In preterm PROM, fetuses reportedly are stressed by the presence of intrauterine inflammation or infection, or both, which accelerates lung maturity by encouraging the secretion of endogenous corticosteroids, resulting in the production of surfactant and eliminating the potential benefit of exogenous ACS.¹¹ This theoretical consideration has not been verified in a systematic analysis of accumulated data,⁵ and the administration of a single course of ACS in preterm PROM now has been shown clearly to improve neonatal outcomes.¹² As Gyamfi-Bannerman and Son demonstrate, the same cannot be said about the administration of a rescue dose of ACS.

Strengths and limitations of the trial
Gyamfi-Bannerman and Son did not address the same interaction as the parent study. The exposure of interest was ACS, an event that occurred in a manner unrelated to the randomization for magnesium sulfate administration. Therefore, the outcome data are no longer randomized in nature, and the study becomes a retrospective cohort analysis.

Gyamfi-Bannerman and Son recognize the limitations of such a study, especially the lack of standardization in the intervention (exact timing of intervention or type of formulation). As with any nonrandomized experiment, the potential for unintended systematic bias is present.

This study, in particular, was subject to “survivor bias,” as reflected in the significantly different intervals between membrane rupture and delivery in the two groups.  

What this evidence means for practice
We must always remain cautious about basing policy or clinical decisions on cohort studies. It has been argued that basing a change in clinical practice on the findings of subgroup analyses is a deviation from fundamental scientific truth.¹³ Such findings should be regarded as hypothesis testing only and consi­dered exploratory in nature.

This study’s abstract conclusion that there is a lack of association between a second dose of ACS and neonatal sepsis should not be regarded as justification to use a rescue course of ACS in women with preterm PROM. We recommend that such a practice be avoided outside the context of an ­approved research protocol.
— Kathleen M. Antony, MD, and Alex C. Vidaeff, MD, MPH

Share your thoughts on this article! Send your Letter to the Editor to [email protected]. Please include your name and the city and state in which you practice.

Gyamfi-Bannerman and Son report their secondary analysis of a randomized controlled trial (RCT) conducted through the Eunice Kennedy Shriver National Institute of Child Health and Human Development’s Maternal-Fetal Medicine Units Network. The aim of the parent RCT was to determine whether antenatal administration of magnesium sulfate decreases the rate of cerebral palsy or death in children delivered preterm. More than 80% of the women enrolled in this RCT had preterm PROM. Of these women, 98% were given antenatal corticosteroids for fetal maturation, and 9% received two courses. This aspect of the RCT provided an opportunity for Gyamfi-Bannerman and Son to study the comparative effects of one course versus two courses of ACS in the setting of preterm PROM.

Background of the study
Concern about the declining efficacy of ACS when the interval between administration and delivery exceeds 7 days prompted several randomized trials exploring the safety and efficacy of multiple ACS courses. Multiple courses were ultimately deemed to be inadvisable because of an association with reduced birth weight and neonatal head circumference.^1,2 However, the unfavorable effects of ACS on anthropometrics was observed when more than three courses were administered, leaving open the possibility of giving only one additional course when needed as a “rescue” dose. Indeed, the use of a single rescue course of ACS in women with intact membranes had a favorable impact on neonatal respiratory function in two RCTs.^3,4

The best available evidence (Level 1) demonstrates that the use of a single course of ACS in preterm PROM does not increase the risk of neonatal or maternal infection even in the setting of prolonged rupture of membranes.⁵ However, pregnant women with preterm PROM were excluded from trials of repetitive ACS dosing because earlier observational studies had suggested that they would experience a substantially increased risk of infectious morbidity when three or more courses of ACS are given.^6–8 What remained debatable on a scientific level was whether a single rescue dose of ACS in women with preterm PROM would be safe and beneficial.

Findings of the analysis
Compared with a single course of ACS, exposure to two courses did not influence the rate of neonatal sepsis or chorioamnionitis. As reassuring as that finding may be, the study found no benefit for the additional course, although it was powered to do so. There was no difference in the rates of respiratory distress syndrome between the study groups.

The findings of Gyamfi-Bannerman and Son replicate those of a subgroup analysis of women in an RCT comparing weekly and single-course ACS.⁹ In that study, weekly courses of ACS in women with preterm PROM did not improve neonatal outcomes beyond what was achieved with single-course therapy. Similar findings have been reported by the Cochrane database.¹⁰

Ruptured versus intact membranes: When is the benefit of ACS greater?
The improvement in neonatal outcomes observed with ACS in pregnancies complicated by preterm PROM is not as pronounced as it is in gestations with intact membranes. In preterm PROM, fetuses reportedly are stressed by the presence of intrauterine inflammation or infection, or both, which accelerates lung maturity by encouraging the secretion of endogenous corticosteroids, resulting in the production of surfactant and eliminating the potential benefit of exogenous ACS.¹¹ This theoretical consideration has not been verified in a systematic analysis of accumulated data,⁵ and the administration of a single course of ACS in preterm PROM now has been shown clearly to improve neonatal outcomes.¹² As Gyamfi-Bannerman and Son demonstrate, the same cannot be said about the administration of a rescue dose of ACS.

Strengths and limitations of the trial
Gyamfi-Bannerman and Son did not address the same interaction as the parent study. The exposure of interest was ACS, an event that occurred in a manner unrelated to the randomization for magnesium sulfate administration. Therefore, the outcome data are no longer randomized in nature, and the study becomes a retrospective cohort analysis.

Gyamfi-Bannerman and Son recognize the limitations of such a study, especially the lack of standardization in the intervention (exact timing of intervention or type of formulation). As with any nonrandomized experiment, the potential for unintended systematic bias is present.

This study, in particular, was subject to “survivor bias,” as reflected in the significantly different intervals between membrane rupture and delivery in the two groups.  

What this evidence means for practice
We must always remain cautious about basing policy or clinical decisions on cohort studies. It has been argued that basing a change in clinical practice on the findings of subgroup analyses is a deviation from fundamental scientific truth.¹³ Such findings should be regarded as hypothesis testing only and consi­dered exploratory in nature.

This study’s abstract conclusion that there is a lack of association between a second dose of ACS and neonatal sepsis should not be regarded as justification to use a rescue course of ACS in women with preterm PROM. We recommend that such a practice be avoided outside the context of an ­approved research protocol.
— Kathleen M. Antony, MD, and Alex C. Vidaeff, MD, MPH

Share your thoughts on this article! Send your Letter to the Editor to [email protected]. Please include your name and the city and state in which you practice.

Gyamfi-Bannerman and Son report their secondary analysis of a randomized controlled trial (RCT) conducted through the Eunice Kennedy Shriver National Institute of Child Health and Human Development’s Maternal-Fetal Medicine Units Network. The aim of the parent RCT was to determine whether antenatal administration of magnesium sulfate decreases the rate of cerebral palsy or death in children delivered preterm. More than 80% of the women enrolled in this RCT had preterm PROM. Of these women, 98% were given antenatal corticosteroids for fetal maturation, and 9% received two courses. This aspect of the RCT provided an opportunity for Gyamfi-Bannerman and Son to study the comparative effects of one course versus two courses of ACS in the setting of preterm PROM.

Background of the study
Concern about the declining efficacy of ACS when the interval between administration and delivery exceeds 7 days prompted several randomized trials exploring the safety and efficacy of multiple ACS courses. Multiple courses were ultimately deemed to be inadvisable because of an association with reduced birth weight and neonatal head circumference.^1,2 However, the unfavorable effects of ACS on anthropometrics was observed when more than three courses were administered, leaving open the possibility of giving only one additional course when needed as a “rescue” dose. Indeed, the use of a single rescue course of ACS in women with intact membranes had a favorable impact on neonatal respiratory function in two RCTs.^3,4

The best available evidence (Level 1) demonstrates that the use of a single course of ACS in preterm PROM does not increase the risk of neonatal or maternal infection even in the setting of prolonged rupture of membranes.⁵ However, pregnant women with preterm PROM were excluded from trials of repetitive ACS dosing because earlier observational studies had suggested that they would experience a substantially increased risk of infectious morbidity when three or more courses of ACS are given.^6–8 What remained debatable on a scientific level was whether a single rescue dose of ACS in women with preterm PROM would be safe and beneficial.

Findings of the analysis
Compared with a single course of ACS, exposure to two courses did not influence the rate of neonatal sepsis or chorioamnionitis. As reassuring as that finding may be, the study found no benefit for the additional course, although it was powered to do so. There was no difference in the rates of respiratory distress syndrome between the study groups.

The findings of Gyamfi-Bannerman and Son replicate those of a subgroup analysis of women in an RCT comparing weekly and single-course ACS.⁹ In that study, weekly courses of ACS in women with preterm PROM did not improve neonatal outcomes beyond what was achieved with single-course therapy. Similar findings have been reported by the Cochrane database.¹⁰

Ruptured versus intact membranes: When is the benefit of ACS greater?
The improvement in neonatal outcomes observed with ACS in pregnancies complicated by preterm PROM is not as pronounced as it is in gestations with intact membranes. In preterm PROM, fetuses reportedly are stressed by the presence of intrauterine inflammation or infection, or both, which accelerates lung maturity by encouraging the secretion of endogenous corticosteroids, resulting in the production of surfactant and eliminating the potential benefit of exogenous ACS.¹¹ This theoretical consideration has not been verified in a systematic analysis of accumulated data,⁵ and the administration of a single course of ACS in preterm PROM now has been shown clearly to improve neonatal outcomes.¹² As Gyamfi-Bannerman and Son demonstrate, the same cannot be said about the administration of a rescue dose of ACS.

Strengths and limitations of the trial
Gyamfi-Bannerman and Son did not address the same interaction as the parent study. The exposure of interest was ACS, an event that occurred in a manner unrelated to the randomization for magnesium sulfate administration. Therefore, the outcome data are no longer randomized in nature, and the study becomes a retrospective cohort analysis.

Gyamfi-Bannerman and Son recognize the limitations of such a study, especially the lack of standardization in the intervention (exact timing of intervention or type of formulation). As with any nonrandomized experiment, the potential for unintended systematic bias is present.

This study, in particular, was subject to “survivor bias,” as reflected in the significantly different intervals between membrane rupture and delivery in the two groups.  

What this evidence means for practice
We must always remain cautious about basing policy or clinical decisions on cohort studies. It has been argued that basing a change in clinical practice on the findings of subgroup analyses is a deviation from fundamental scientific truth.¹³ Such findings should be regarded as hypothesis testing only and consi­dered exploratory in nature.

This study’s abstract conclusion that there is a lack of association between a second dose of ACS and neonatal sepsis should not be regarded as justification to use a rescue course of ACS in women with preterm PROM. We recommend that such a practice be avoided outside the context of an ­approved research protocol.
— Kathleen M. Antony, MD, and Alex C. Vidaeff, MD, MPH

Share your thoughts on this article! Send your Letter to the Editor to [email protected]. Please include your name and the city and state in which you practice.

In contemporary medical practice, we expect our clinical actions to reflect the best and most current evidence. In many cases, however, the evidence available to us is weak or irrelevant. In their investigation, Berghella and colleagues set out to assess the quality of evidence in the setting of transvaginal cervical cerclage by reviewing the published data on selected perioperative strategies. They elected to perform a systematic review, as opposed to a narrative review (a simple opinion piece), because this approach follows an explicit process designed to limit bias and random error in the interpretation of scientific research.

The studies they analyzed vary from observational investigations to randomized trials, generating considerable heterogeneity in the data. Therefore, it would not have been feasible or appropriate for them to combine the results in a quantitative review (ie, meta-analysis). Their solution: to limit the analysis to a qualitative systematic review.

The term “systematic review” implies that investigators have an accurate and ­comprehensive understanding of existent data, with each study representing one contribution to a much larger body of knowledge. Over the years, Berghella and colleagues have contributed extensively to the literature on cervical cerclage and are well qualified to provide an analytic framework for the flood of published information on this practice. Although they focused primarily on how to perform cerclage, a discussion of when to ­perform cerclage cannot be separated from any consideration of efficacy.

When, exactly, is cerclage indicated?
The original indication for cerclage, established more than 50 years ago, required both a history of second-trimester loss and asymptomatic cervical changes in the current pregnancy. Since then, many cerclages have been performed on the basis of history alone or on current cervical changes regardless of history. However, the most recent professional guidelines reconfirm that any cerclage procedure should be supported by both historical and contemporaneous findings.^1,2 

Investigators have demonstrated that the measurement of cervical length by ­transvaginal ultrasound should generally be an integral part of clinical evaluation for asymptomatic cervical changes. Indeed, sonographic assessment has emerged as a tool capable of reducing “overcall” and unnecessary intervention.³ On the other hand, a meta-analysis of four randomized trials of ultrasound-indicated cerclage found it to be beneficial in women with a short cervix only if they also had a history of preterm delivery.⁴

In addition, randomized trials have documented a benefit for cerclage in two other clinical contexts:

• Results from a secondary analysis of data from a large randomized trial published in 1993 suggest that elective cerclage can be based on history alone in women with three or more second-trimester losses or preterm births.⁵
• Authors of a randomized trial published in 2003 ­suggested that women with advanced cervical changes, such as dilatation of the external os with exposure of the fetal membranes, may benefit from “emergency” cerclage even in the absence of a prior preterm delivery.⁶

How the data were analyzed
The data included in the review were analyzed separately, according to three widely accepted indications for cerclage:

• history-indicated: a history of three or more second-trimester losses and/or preterm births
• ultrasound-indicated: ultrasonographic detection of a cervical length of less than 25 mm, as measured by transvaginal ultrasound, in a woman with a history of second-trimester loss or preterm birth
• physical-examination–indicated: physical examination (manual or with a speculum) that confirms a dilated cervix.

Granted, this terminology can be confusing, as in the case of ultrasound-indicated cerclage, which includes aspects of the patient’s history. Moreover, I doubt that the studies included in this analysis always adhered to these definitions. The heterogeneity of the study population and the ambiguity of these definitions may limit the applicability of findings. In fact, they constitute the major (albeit practically unavoidable) limitation of this review.

The optimal approach to subclinical infection is unclear
Although there is a consensus that overt intra-amniotic infection is an absolute contraindication to cerclage, the implications of subclinical intra-amniotic infection in asymptomatic women are unclear. About 50% of women considered for emergency cerclage are likely to have intra-amniotic infection.⁷ An ongoing randomized trial is expected to elucidate the benefit of precerclage amniocentesis in such cases.

The sonographic detection of sludge in the amniotic fluid also has been associated with intra-amniotic infection. However, after analyzing the data, Berghella and colleagues did not find adequate justification for amniocentesis in this setting. A more practical question might be whether cerclage is advisable at all when sludge is present. Data from a recently reported abstract suggest that the presence of sludge increases the likelihood of early preterm birth independent of cervical length.⁸

Other gray areas
Another absolute contraindication to cerclage is the presence of painful uterine contractions in a woman exhibiting cervical change. The study findings seemed to imply that when uterine contractions are detected via tocodynamometric monitoring but are not experienced by the patient, cerclage may be appropriate. In my opinion, this issue represents another open clinical question.

On at least one occasion, I have discovered that the patient scheduled for cerclage has not undergone preoperative screening for genital tract infections. In such a scenario, it is unclear whether it is best to cancel the procedure or to proceed. Berghella and colleagues concluded that evidence is insufficient to support routine, universal screening for genital tract infection prior to cerclage. In the absence of risk factors for infection, it may be appropriate to proceed with cerclage.

The issue of combined use of cerclage and 17α hydroxyprogesterone caproate also was addressed in this review. Berghella and colleagues recommended that progestin supplementation be continued if cerclage is subsequently performed. I, too, follow this approach, although the two interventions (progestin followed by cerclage) have not been studied in this sequence.

Related article: A stepwise approach to cervical cerclage (June 2012)

Data are limited, and conflicting, on the use of 17α hydroxyprogesterone caproate after cerclage. A small randomized trial suggested benefit,⁹ while a retrospective cohort study¹⁰ and a secondary analysis of data from a randomized trial of cerclage¹¹ found no benefit for the addition of 17α hydroxyprogesterone caproate.

Berghella and colleagues advocated against reinforcing (second or repeat) cerclage, an intervention that may be associated with a higher incidence of preterm birth.¹² Accordingly, continuing sonographic assessment of cervical length after cerclage may be futile.   What this evidence means for practice
Increasing evidence suggests that cervical shortening is not limited to an innate or acquired cervical weakness but represents an early, asymptomatic phase on the pathway to preterm birth. Nevertheless, cerclage continues to be practiced widely and has been shown to be beneficial in selected populations.

We clinicians often face complex cases of presumed cervical insufficiency for which there are no simple or unequivocal recommendations. This appraisal by Berghella and colleagues will help us separate evidence-validated approaches from misconceptions in the technical aspects of cerclage.
--Alex C. Vidaeff, MD, MPH

We want to hear from you! Tell us what you think.

In contemporary medical practice, we expect our clinical actions to reflect the best and most current evidence. In many cases, however, the evidence available to us is weak or irrelevant. In their investigation, Berghella and colleagues set out to assess the quality of evidence in the setting of transvaginal cervical cerclage by reviewing the published data on selected perioperative strategies. They elected to perform a systematic review, as opposed to a narrative review (a simple opinion piece), because this approach follows an explicit process designed to limit bias and random error in the interpretation of scientific research.

The studies they analyzed vary from observational investigations to randomized trials, generating considerable heterogeneity in the data. Therefore, it would not have been feasible or appropriate for them to combine the results in a quantitative review (ie, meta-analysis). Their solution: to limit the analysis to a qualitative systematic review.

The term “systematic review” implies that investigators have an accurate and ­comprehensive understanding of existent data, with each study representing one contribution to a much larger body of knowledge. Over the years, Berghella and colleagues have contributed extensively to the literature on cervical cerclage and are well qualified to provide an analytic framework for the flood of published information on this practice. Although they focused primarily on how to perform cerclage, a discussion of when to ­perform cerclage cannot be separated from any consideration of efficacy.

When, exactly, is cerclage indicated?
The original indication for cerclage, established more than 50 years ago, required both a history of second-trimester loss and asymptomatic cervical changes in the current pregnancy. Since then, many cerclages have been performed on the basis of history alone or on current cervical changes regardless of history. However, the most recent professional guidelines reconfirm that any cerclage procedure should be supported by both historical and contemporaneous findings.^1,2 

Investigators have demonstrated that the measurement of cervical length by ­transvaginal ultrasound should generally be an integral part of clinical evaluation for asymptomatic cervical changes. Indeed, sonographic assessment has emerged as a tool capable of reducing “overcall” and unnecessary intervention.³ On the other hand, a meta-analysis of four randomized trials of ultrasound-indicated cerclage found it to be beneficial in women with a short cervix only if they also had a history of preterm delivery.⁴

In addition, randomized trials have documented a benefit for cerclage in two other clinical contexts:

• Results from a secondary analysis of data from a large randomized trial published in 1993 suggest that elective cerclage can be based on history alone in women with three or more second-trimester losses or preterm births.⁵
• Authors of a randomized trial published in 2003 ­suggested that women with advanced cervical changes, such as dilatation of the external os with exposure of the fetal membranes, may benefit from “emergency” cerclage even in the absence of a prior preterm delivery.⁶

How the data were analyzed
The data included in the review were analyzed separately, according to three widely accepted indications for cerclage:

• history-indicated: a history of three or more second-trimester losses and/or preterm births
• ultrasound-indicated: ultrasonographic detection of a cervical length of less than 25 mm, as measured by transvaginal ultrasound, in a woman with a history of second-trimester loss or preterm birth
• physical-examination–indicated: physical examination (manual or with a speculum) that confirms a dilated cervix.

Granted, this terminology can be confusing, as in the case of ultrasound-indicated cerclage, which includes aspects of the patient’s history. Moreover, I doubt that the studies included in this analysis always adhered to these definitions. The heterogeneity of the study population and the ambiguity of these definitions may limit the applicability of findings. In fact, they constitute the major (albeit practically unavoidable) limitation of this review.

The optimal approach to subclinical infection is unclear
Although there is a consensus that overt intra-amniotic infection is an absolute contraindication to cerclage, the implications of subclinical intra-amniotic infection in asymptomatic women are unclear. About 50% of women considered for emergency cerclage are likely to have intra-amniotic infection.⁷ An ongoing randomized trial is expected to elucidate the benefit of precerclage amniocentesis in such cases.

The sonographic detection of sludge in the amniotic fluid also has been associated with intra-amniotic infection. However, after analyzing the data, Berghella and colleagues did not find adequate justification for amniocentesis in this setting. A more practical question might be whether cerclage is advisable at all when sludge is present. Data from a recently reported abstract suggest that the presence of sludge increases the likelihood of early preterm birth independent of cervical length.⁸

Other gray areas
Another absolute contraindication to cerclage is the presence of painful uterine contractions in a woman exhibiting cervical change. The study findings seemed to imply that when uterine contractions are detected via tocodynamometric monitoring but are not experienced by the patient, cerclage may be appropriate. In my opinion, this issue represents another open clinical question.

On at least one occasion, I have discovered that the patient scheduled for cerclage has not undergone preoperative screening for genital tract infections. In such a scenario, it is unclear whether it is best to cancel the procedure or to proceed. Berghella and colleagues concluded that evidence is insufficient to support routine, universal screening for genital tract infection prior to cerclage. In the absence of risk factors for infection, it may be appropriate to proceed with cerclage.

The issue of combined use of cerclage and 17α hydroxyprogesterone caproate also was addressed in this review. Berghella and colleagues recommended that progestin supplementation be continued if cerclage is subsequently performed. I, too, follow this approach, although the two interventions (progestin followed by cerclage) have not been studied in this sequence.

Related article: A stepwise approach to cervical cerclage (June 2012)

Data are limited, and conflicting, on the use of 17α hydroxyprogesterone caproate after cerclage. A small randomized trial suggested benefit,⁹ while a retrospective cohort study¹⁰ and a secondary analysis of data from a randomized trial of cerclage¹¹ found no benefit for the addition of 17α hydroxyprogesterone caproate.

Berghella and colleagues advocated against reinforcing (second or repeat) cerclage, an intervention that may be associated with a higher incidence of preterm birth.¹² Accordingly, continuing sonographic assessment of cervical length after cerclage may be futile.   What this evidence means for practice
Increasing evidence suggests that cervical shortening is not limited to an innate or acquired cervical weakness but represents an early, asymptomatic phase on the pathway to preterm birth. Nevertheless, cerclage continues to be practiced widely and has been shown to be beneficial in selected populations.

We clinicians often face complex cases of presumed cervical insufficiency for which there are no simple or unequivocal recommendations. This appraisal by Berghella and colleagues will help us separate evidence-validated approaches from misconceptions in the technical aspects of cerclage.
--Alex C. Vidaeff, MD, MPH

We want to hear from you! Tell us what you think.

In contemporary medical practice, we expect our clinical actions to reflect the best and most current evidence. In many cases, however, the evidence available to us is weak or irrelevant. In their investigation, Berghella and colleagues set out to assess the quality of evidence in the setting of transvaginal cervical cerclage by reviewing the published data on selected perioperative strategies. They elected to perform a systematic review, as opposed to a narrative review (a simple opinion piece), because this approach follows an explicit process designed to limit bias and random error in the interpretation of scientific research.

The studies they analyzed vary from observational investigations to randomized trials, generating considerable heterogeneity in the data. Therefore, it would not have been feasible or appropriate for them to combine the results in a quantitative review (ie, meta-analysis). Their solution: to limit the analysis to a qualitative systematic review.

The term “systematic review” implies that investigators have an accurate and ­comprehensive understanding of existent data, with each study representing one contribution to a much larger body of knowledge. Over the years, Berghella and colleagues have contributed extensively to the literature on cervical cerclage and are well qualified to provide an analytic framework for the flood of published information on this practice. Although they focused primarily on how to perform cerclage, a discussion of when to ­perform cerclage cannot be separated from any consideration of efficacy.

When, exactly, is cerclage indicated?
The original indication for cerclage, established more than 50 years ago, required both a history of second-trimester loss and asymptomatic cervical changes in the current pregnancy. Since then, many cerclages have been performed on the basis of history alone or on current cervical changes regardless of history. However, the most recent professional guidelines reconfirm that any cerclage procedure should be supported by both historical and contemporaneous findings.^1,2 

Investigators have demonstrated that the measurement of cervical length by ­transvaginal ultrasound should generally be an integral part of clinical evaluation for asymptomatic cervical changes. Indeed, sonographic assessment has emerged as a tool capable of reducing “overcall” and unnecessary intervention.³ On the other hand, a meta-analysis of four randomized trials of ultrasound-indicated cerclage found it to be beneficial in women with a short cervix only if they also had a history of preterm delivery.⁴

In addition, randomized trials have documented a benefit for cerclage in two other clinical contexts:

• Results from a secondary analysis of data from a large randomized trial published in 1993 suggest that elective cerclage can be based on history alone in women with three or more second-trimester losses or preterm births.⁵
• Authors of a randomized trial published in 2003 ­suggested that women with advanced cervical changes, such as dilatation of the external os with exposure of the fetal membranes, may benefit from “emergency” cerclage even in the absence of a prior preterm delivery.⁶

How the data were analyzed
The data included in the review were analyzed separately, according to three widely accepted indications for cerclage:

• history-indicated: a history of three or more second-trimester losses and/or preterm births
• ultrasound-indicated: ultrasonographic detection of a cervical length of less than 25 mm, as measured by transvaginal ultrasound, in a woman with a history of second-trimester loss or preterm birth
• physical-examination–indicated: physical examination (manual or with a speculum) that confirms a dilated cervix.

Granted, this terminology can be confusing, as in the case of ultrasound-indicated cerclage, which includes aspects of the patient’s history. Moreover, I doubt that the studies included in this analysis always adhered to these definitions. The heterogeneity of the study population and the ambiguity of these definitions may limit the applicability of findings. In fact, they constitute the major (albeit practically unavoidable) limitation of this review.

The optimal approach to subclinical infection is unclear
Although there is a consensus that overt intra-amniotic infection is an absolute contraindication to cerclage, the implications of subclinical intra-amniotic infection in asymptomatic women are unclear. About 50% of women considered for emergency cerclage are likely to have intra-amniotic infection.⁷ An ongoing randomized trial is expected to elucidate the benefit of precerclage amniocentesis in such cases.

The sonographic detection of sludge in the amniotic fluid also has been associated with intra-amniotic infection. However, after analyzing the data, Berghella and colleagues did not find adequate justification for amniocentesis in this setting. A more practical question might be whether cerclage is advisable at all when sludge is present. Data from a recently reported abstract suggest that the presence of sludge increases the likelihood of early preterm birth independent of cervical length.⁸

Other gray areas
Another absolute contraindication to cerclage is the presence of painful uterine contractions in a woman exhibiting cervical change. The study findings seemed to imply that when uterine contractions are detected via tocodynamometric monitoring but are not experienced by the patient, cerclage may be appropriate. In my opinion, this issue represents another open clinical question.

On at least one occasion, I have discovered that the patient scheduled for cerclage has not undergone preoperative screening for genital tract infections. In such a scenario, it is unclear whether it is best to cancel the procedure or to proceed. Berghella and colleagues concluded that evidence is insufficient to support routine, universal screening for genital tract infection prior to cerclage. In the absence of risk factors for infection, it may be appropriate to proceed with cerclage.

The issue of combined use of cerclage and 17α hydroxyprogesterone caproate also was addressed in this review. Berghella and colleagues recommended that progestin supplementation be continued if cerclage is subsequently performed. I, too, follow this approach, although the two interventions (progestin followed by cerclage) have not been studied in this sequence.

Related article: A stepwise approach to cervical cerclage (June 2012)

Data are limited, and conflicting, on the use of 17α hydroxyprogesterone caproate after cerclage. A small randomized trial suggested benefit,⁹ while a retrospective cohort study¹⁰ and a secondary analysis of data from a randomized trial of cerclage¹¹ found no benefit for the addition of 17α hydroxyprogesterone caproate.

Berghella and colleagues advocated against reinforcing (second or repeat) cerclage, an intervention that may be associated with a higher incidence of preterm birth.¹² Accordingly, continuing sonographic assessment of cervical length after cerclage may be futile.   What this evidence means for practice
Increasing evidence suggests that cervical shortening is not limited to an innate or acquired cervical weakness but represents an early, asymptomatic phase on the pathway to preterm birth. Nevertheless, cerclage continues to be practiced widely and has been shown to be beneficial in selected populations.

We clinicians often face complex cases of presumed cervical insufficiency for which there are no simple or unequivocal recommendations. This appraisal by Berghella and colleagues will help us separate evidence-validated approaches from misconceptions in the technical aspects of cerclage.
--Alex C. Vidaeff, MD, MPH

We want to hear from you! Tell us what you think.

Approximately 13 million preterm births occur annually worldwide.¹ Depending on the geographic locale, PPROM is responsible for 16% to 40% of these births.²

The clinical approach to PPROM is one of the most contentious issues in obstetrics, with disagreement on virtually every aspect of it. Under debate are the lower and upper limits of the gestational age range at which intervention is warranted, as well as the use of ancillary interventions such as corticosteroids and antibiotics. Briery and colleagues add to the scientific debate now by asking whether 17P would be effective as cotreatment (with antibiotics) to prolong latency after PPROM.

According to their findings, the answer to this question is “No.”

Details of the trial

Briery and colleagues conducted a placebo-controlled, double-blind, randomized clinical trial of women with a singleton gestation complicated by PPROM. Excluded from the study were women whose pregnancy involved additional fetal or placental complications.

All women included in the study received antibiotics according to a protocol from the National Institutes of Health; they also were given betamethasone for fetal maturation. Tocolytics were not used. Because randomization did not occur until after each woman was transferred from the labor and delivery unit to the high-risk floor, we can assume that no participants were manifesting uterine contractions.

Women received weekly injections of 17P or placebo until 34 weeks’ gestation or delivery. The primary outcome was the interval from study entry to delivery.

One woman had a pregnancy of 23.5 weeks’ duration at randomization; the remainder had gestations that were 24 weeks or older. There were no other differences in demographics, cervical dilatation, gestational age at study entry, or reasons for delivery between the two study groups.

Study design may have been unrealistic

The authors calculated that they needed a sample size of 56 patients to detect a 50% increase in latency, based on population data from their institution showing that 80% of patients who have PPROM deliver within 7 days. Such a calculation may have set an unrealistic—albeit logistically convenient—goal, rendering the study underpowered to detect smaller effects. Note, for example, that when antibiotics are given to women who have PPROM, prolongation of the latency period is only 33% (pooled effect from a recent meta-analysis).³ Even so, given the findings of Briery and colleagues, latency improvement after 17P administration would appear to be unlikely even in a larger study. There was not even a trend toward a longer interval to delivery (mean of 11.2 days with 17P vs 14.5 days with placebo).

Only secondary prevention of preterm birth is effective

The indications for progesterone supplementation in pregnancy are still evolving as part of a sustained scientific effort to prevent preterm labor and delivery. Strategies to prevent preterm delivery can be categorized as primary, secondary, or tertiary, as can strategies for other public health concerns.

Because any number of variables—known and unknown—may trigger preterm labor, identifying them and providing primary preventive strategies in the entire pregnant population remain elusive tasks.

Tertiary prevention—i.e., treatment given to already symptomatic individuals—is also notoriously ineffective. There are no data supporting the use of progesterone as primary prevention (in low-risk women) or tertiary prevention (e.g., tocolytic). ACOG made note of this in 2003, and its conclusions remain valid today.⁴ According to a 2010 Cochrane review, there is insufficient evidence to advocate progestational agents as tocolytic agents for women who present with threatened or established preterm labor.⁵

In light of these data, the results reported by Briery and colleagues are hardly surprising. In women who may have already entered the irreversible phase of parturition (manifesting uterine contractions; presenting with advanced, painless cervical dilatation; or after PPROM), progesterone will remain ineffective. The only applicable use of prophylactic progesterone in pregnancy is as secondary prevention.⁴ In contrast to primary and tertiary prevention, the secondary level of prevention—i.e., an intervention aimed at minimizing the risk of preterm birth in women who are identified as having an elevated risk—is supported by several systematic reviews of randomized, controlled trials.^6,7 According to these reviews, progesterone certainly is effective in high-risk pregnant women who have a short cervix or a history of spontaneous preterm birth. The same cannot be said about women who have PPROM.

We want to hear from you!  Tell us what you think.

Approximately 13 million preterm births occur annually worldwide.¹ Depending on the geographic locale, PPROM is responsible for 16% to 40% of these births.²

The clinical approach to PPROM is one of the most contentious issues in obstetrics, with disagreement on virtually every aspect of it. Under debate are the lower and upper limits of the gestational age range at which intervention is warranted, as well as the use of ancillary interventions such as corticosteroids and antibiotics. Briery and colleagues add to the scientific debate now by asking whether 17P would be effective as cotreatment (with antibiotics) to prolong latency after PPROM.

According to their findings, the answer to this question is “No.”

Details of the trial

Briery and colleagues conducted a placebo-controlled, double-blind, randomized clinical trial of women with a singleton gestation complicated by PPROM. Excluded from the study were women whose pregnancy involved additional fetal or placental complications.

All women included in the study received antibiotics according to a protocol from the National Institutes of Health; they also were given betamethasone for fetal maturation. Tocolytics were not used. Because randomization did not occur until after each woman was transferred from the labor and delivery unit to the high-risk floor, we can assume that no participants were manifesting uterine contractions.

Women received weekly injections of 17P or placebo until 34 weeks’ gestation or delivery. The primary outcome was the interval from study entry to delivery.

One woman had a pregnancy of 23.5 weeks’ duration at randomization; the remainder had gestations that were 24 weeks or older. There were no other differences in demographics, cervical dilatation, gestational age at study entry, or reasons for delivery between the two study groups.

Study design may have been unrealistic

The authors calculated that they needed a sample size of 56 patients to detect a 50% increase in latency, based on population data from their institution showing that 80% of patients who have PPROM deliver within 7 days. Such a calculation may have set an unrealistic—albeit logistically convenient—goal, rendering the study underpowered to detect smaller effects. Note, for example, that when antibiotics are given to women who have PPROM, prolongation of the latency period is only 33% (pooled effect from a recent meta-analysis).³ Even so, given the findings of Briery and colleagues, latency improvement after 17P administration would appear to be unlikely even in a larger study. There was not even a trend toward a longer interval to delivery (mean of 11.2 days with 17P vs 14.5 days with placebo).

Only secondary prevention of preterm birth is effective

The indications for progesterone supplementation in pregnancy are still evolving as part of a sustained scientific effort to prevent preterm labor and delivery. Strategies to prevent preterm delivery can be categorized as primary, secondary, or tertiary, as can strategies for other public health concerns.

Because any number of variables—known and unknown—may trigger preterm labor, identifying them and providing primary preventive strategies in the entire pregnant population remain elusive tasks.

Tertiary prevention—i.e., treatment given to already symptomatic individuals—is also notoriously ineffective. There are no data supporting the use of progesterone as primary prevention (in low-risk women) or tertiary prevention (e.g., tocolytic). ACOG made note of this in 2003, and its conclusions remain valid today.⁴ According to a 2010 Cochrane review, there is insufficient evidence to advocate progestational agents as tocolytic agents for women who present with threatened or established preterm labor.⁵

In light of these data, the results reported by Briery and colleagues are hardly surprising. In women who may have already entered the irreversible phase of parturition (manifesting uterine contractions; presenting with advanced, painless cervical dilatation; or after PPROM), progesterone will remain ineffective. The only applicable use of prophylactic progesterone in pregnancy is as secondary prevention.⁴ In contrast to primary and tertiary prevention, the secondary level of prevention—i.e., an intervention aimed at minimizing the risk of preterm birth in women who are identified as having an elevated risk—is supported by several systematic reviews of randomized, controlled trials.^6,7 According to these reviews, progesterone certainly is effective in high-risk pregnant women who have a short cervix or a history of spontaneous preterm birth. The same cannot be said about women who have PPROM.

We want to hear from you!  Tell us what you think.

Approximately 13 million preterm births occur annually worldwide.¹ Depending on the geographic locale, PPROM is responsible for 16% to 40% of these births.²

The clinical approach to PPROM is one of the most contentious issues in obstetrics, with disagreement on virtually every aspect of it. Under debate are the lower and upper limits of the gestational age range at which intervention is warranted, as well as the use of ancillary interventions such as corticosteroids and antibiotics. Briery and colleagues add to the scientific debate now by asking whether 17P would be effective as cotreatment (with antibiotics) to prolong latency after PPROM.

According to their findings, the answer to this question is “No.”

Details of the trial

Briery and colleagues conducted a placebo-controlled, double-blind, randomized clinical trial of women with a singleton gestation complicated by PPROM. Excluded from the study were women whose pregnancy involved additional fetal or placental complications.

All women included in the study received antibiotics according to a protocol from the National Institutes of Health; they also were given betamethasone for fetal maturation. Tocolytics were not used. Because randomization did not occur until after each woman was transferred from the labor and delivery unit to the high-risk floor, we can assume that no participants were manifesting uterine contractions.

Women received weekly injections of 17P or placebo until 34 weeks’ gestation or delivery. The primary outcome was the interval from study entry to delivery.

One woman had a pregnancy of 23.5 weeks’ duration at randomization; the remainder had gestations that were 24 weeks or older. There were no other differences in demographics, cervical dilatation, gestational age at study entry, or reasons for delivery between the two study groups.

Study design may have been unrealistic

The authors calculated that they needed a sample size of 56 patients to detect a 50% increase in latency, based on population data from their institution showing that 80% of patients who have PPROM deliver within 7 days. Such a calculation may have set an unrealistic—albeit logistically convenient—goal, rendering the study underpowered to detect smaller effects. Note, for example, that when antibiotics are given to women who have PPROM, prolongation of the latency period is only 33% (pooled effect from a recent meta-analysis).³ Even so, given the findings of Briery and colleagues, latency improvement after 17P administration would appear to be unlikely even in a larger study. There was not even a trend toward a longer interval to delivery (mean of 11.2 days with 17P vs 14.5 days with placebo).

Only secondary prevention of preterm birth is effective

The indications for progesterone supplementation in pregnancy are still evolving as part of a sustained scientific effort to prevent preterm labor and delivery. Strategies to prevent preterm delivery can be categorized as primary, secondary, or tertiary, as can strategies for other public health concerns.

Because any number of variables—known and unknown—may trigger preterm labor, identifying them and providing primary preventive strategies in the entire pregnant population remain elusive tasks.

Tertiary prevention—i.e., treatment given to already symptomatic individuals—is also notoriously ineffective. There are no data supporting the use of progesterone as primary prevention (in low-risk women) or tertiary prevention (e.g., tocolytic). ACOG made note of this in 2003, and its conclusions remain valid today.⁴ According to a 2010 Cochrane review, there is insufficient evidence to advocate progestational agents as tocolytic agents for women who present with threatened or established preterm labor.⁵

In light of these data, the results reported by Briery and colleagues are hardly surprising. In women who may have already entered the irreversible phase of parturition (manifesting uterine contractions; presenting with advanced, painless cervical dilatation; or after PPROM), progesterone will remain ineffective. The only applicable use of prophylactic progesterone in pregnancy is as secondary prevention.⁴ In contrast to primary and tertiary prevention, the secondary level of prevention—i.e., an intervention aimed at minimizing the risk of preterm birth in women who are identified as having an elevated risk—is supported by several systematic reviews of randomized, controlled trials.^6,7 According to these reviews, progesterone certainly is effective in high-risk pregnant women who have a short cervix or a history of spontaneous preterm birth. The same cannot be said about women who have PPROM.

We want to hear from you!  Tell us what you think.

• 1) a history of second-trimester loss involving painless cervical dilatation in the absence of infection, bleeding, amniorrhexis, and fetal demise
  
• 2) asymptomatic cervical changes in the current pregnancy.
  

Although our understanding of cervical insufficiency has undergone many revisions and reinterpretations in the intervening years, we still lack an accepted diagnostic test or proven criteria for diagnosis. Cerclage is placed in 1% of all pregnancies in the United States, but there is no consensus on indications, and the effectiveness is still a matter of debate.¹ Cerclage placement based on ultrasonographic (US) measurement of cervical length has been proposed as the solution to this clinical quagmire, in the wake of evidence suggesting that cervical length may act as a surrogate for cervical competence.

A patient-level meta-analysis of four randomized trials of cervical cerclage, published in 2005, reconfirmed the original indication for cerclage. In women who had a cervical length below 25 mm, cerclage reduced the rate of preterm birth at less than 35 weeks’ gestation only if they had a history of preterm birth.² This finding prompted the question: Would such women represent a truly homogeneous population in terms of therapeutic response to cerclage?

The Owen trial attempts to answer this specific question.

Details of the trial

Women who had a cervix shorter than 25 mm and a history of preterm birth and who were pregnant with a singleton gestation were eligible. Candidates for elective cerclage based on history, or for emergency cerclage based on cervical dilatation of at least 2 cm with visible membranes, were excluded from the study—possibly reducing the generalizability of the findings.

For the remaining 302 participants, cerclage appeared to have an overall benefit when survival analysis took the duration of gestation into consideration. But only women who had a cervical length below 15 mm had a significant reduction in the primary outcome (preterm birth at less than 35 weeks’ gestation) with cerclage. These results are somewhat reminiscent of the findings of a randomized comparison of cerclage and 17α-hydroxyprogesterone caproate in women who had a short cervix (measured by US), in which cerclage proved to be superior only when cervical length was less than 15 mm.³

Why the 15-mm cutoff isn’t definitive

Despite these findings, the 15-mm measurement cannot be assumed to be completely prescriptive because it was selected somewhat arbitrarily. Furthermore, it may be inadvisable to wait for cervical length to decrease below 15 mm. In pregnancies in which cervical length is below 5 mm, there is a significantly higher expression of intra-amniotic inflammation than in those in which cervical length is 6 to 25 mm, according to another recent study.⁴ Women who have a very short cervix may be far along the inflammatory cascade and may have already entered the irreversible phase of parturition, reducing the efficacy and even the advisability of cerclage. A positive fetal fibronectin test (as a marker of inflammation and choriodecidual disruption) and an increased level of interleukin-8 in cervical mucus reportedly identified a subgroup of women with a short cervix who would not benefit from cerclage—and who might even be harmed by it.^5,6

Because preterm birth is such a complex disorder, it is unlikely that one intervention will be effective in all women—even within a certain stratum of cervical length. Rather, it may be necessary to identify subsets of pregnant women amenable to targeted or tailored intervention.

• 1) a history of second-trimester loss involving painless cervical dilatation in the absence of infection, bleeding, amniorrhexis, and fetal demise
  
• 2) asymptomatic cervical changes in the current pregnancy.
  

Although our understanding of cervical insufficiency has undergone many revisions and reinterpretations in the intervening years, we still lack an accepted diagnostic test or proven criteria for diagnosis. Cerclage is placed in 1% of all pregnancies in the United States, but there is no consensus on indications, and the effectiveness is still a matter of debate.¹ Cerclage placement based on ultrasonographic (US) measurement of cervical length has been proposed as the solution to this clinical quagmire, in the wake of evidence suggesting that cervical length may act as a surrogate for cervical competence.

A patient-level meta-analysis of four randomized trials of cervical cerclage, published in 2005, reconfirmed the original indication for cerclage. In women who had a cervical length below 25 mm, cerclage reduced the rate of preterm birth at less than 35 weeks’ gestation only if they had a history of preterm birth.² This finding prompted the question: Would such women represent a truly homogeneous population in terms of therapeutic response to cerclage?

The Owen trial attempts to answer this specific question.

Details of the trial

Women who had a cervix shorter than 25 mm and a history of preterm birth and who were pregnant with a singleton gestation were eligible. Candidates for elective cerclage based on history, or for emergency cerclage based on cervical dilatation of at least 2 cm with visible membranes, were excluded from the study—possibly reducing the generalizability of the findings.

For the remaining 302 participants, cerclage appeared to have an overall benefit when survival analysis took the duration of gestation into consideration. But only women who had a cervical length below 15 mm had a significant reduction in the primary outcome (preterm birth at less than 35 weeks’ gestation) with cerclage. These results are somewhat reminiscent of the findings of a randomized comparison of cerclage and 17α-hydroxyprogesterone caproate in women who had a short cervix (measured by US), in which cerclage proved to be superior only when cervical length was less than 15 mm.³

Why the 15-mm cutoff isn’t definitive

Despite these findings, the 15-mm measurement cannot be assumed to be completely prescriptive because it was selected somewhat arbitrarily. Furthermore, it may be inadvisable to wait for cervical length to decrease below 15 mm. In pregnancies in which cervical length is below 5 mm, there is a significantly higher expression of intra-amniotic inflammation than in those in which cervical length is 6 to 25 mm, according to another recent study.⁴ Women who have a very short cervix may be far along the inflammatory cascade and may have already entered the irreversible phase of parturition, reducing the efficacy and even the advisability of cerclage. A positive fetal fibronectin test (as a marker of inflammation and choriodecidual disruption) and an increased level of interleukin-8 in cervical mucus reportedly identified a subgroup of women with a short cervix who would not benefit from cerclage—and who might even be harmed by it.^5,6

Because preterm birth is such a complex disorder, it is unlikely that one intervention will be effective in all women—even within a certain stratum of cervical length. Rather, it may be necessary to identify subsets of pregnant women amenable to targeted or tailored intervention.

• 1) a history of second-trimester loss involving painless cervical dilatation in the absence of infection, bleeding, amniorrhexis, and fetal demise
  
• 2) asymptomatic cervical changes in the current pregnancy.
  

Although our understanding of cervical insufficiency has undergone many revisions and reinterpretations in the intervening years, we still lack an accepted diagnostic test or proven criteria for diagnosis. Cerclage is placed in 1% of all pregnancies in the United States, but there is no consensus on indications, and the effectiveness is still a matter of debate.¹ Cerclage placement based on ultrasonographic (US) measurement of cervical length has been proposed as the solution to this clinical quagmire, in the wake of evidence suggesting that cervical length may act as a surrogate for cervical competence.

A patient-level meta-analysis of four randomized trials of cervical cerclage, published in 2005, reconfirmed the original indication for cerclage. In women who had a cervical length below 25 mm, cerclage reduced the rate of preterm birth at less than 35 weeks’ gestation only if they had a history of preterm birth.² This finding prompted the question: Would such women represent a truly homogeneous population in terms of therapeutic response to cerclage?

The Owen trial attempts to answer this specific question.

Details of the trial

Women who had a cervix shorter than 25 mm and a history of preterm birth and who were pregnant with a singleton gestation were eligible. Candidates for elective cerclage based on history, or for emergency cerclage based on cervical dilatation of at least 2 cm with visible membranes, were excluded from the study—possibly reducing the generalizability of the findings.

For the remaining 302 participants, cerclage appeared to have an overall benefit when survival analysis took the duration of gestation into consideration. But only women who had a cervical length below 15 mm had a significant reduction in the primary outcome (preterm birth at less than 35 weeks’ gestation) with cerclage. These results are somewhat reminiscent of the findings of a randomized comparison of cerclage and 17α-hydroxyprogesterone caproate in women who had a short cervix (measured by US), in which cerclage proved to be superior only when cervical length was less than 15 mm.³

Why the 15-mm cutoff isn’t definitive

Despite these findings, the 15-mm measurement cannot be assumed to be completely prescriptive because it was selected somewhat arbitrarily. Furthermore, it may be inadvisable to wait for cervical length to decrease below 15 mm. In pregnancies in which cervical length is below 5 mm, there is a significantly higher expression of intra-amniotic inflammation than in those in which cervical length is 6 to 25 mm, according to another recent study.⁴ Women who have a very short cervix may be far along the inflammatory cascade and may have already entered the irreversible phase of parturition, reducing the efficacy and even the advisability of cerclage. A positive fetal fibronectin test (as a marker of inflammation and choriodecidual disruption) and an increased level of interleukin-8 in cervical mucus reportedly identified a subgroup of women with a short cervix who would not benefit from cerclage—and who might even be harmed by it.^5,6

Because preterm birth is such a complex disorder, it is unlikely that one intervention will be effective in all women—even within a certain stratum of cervical length. Rather, it may be necessary to identify subsets of pregnant women amenable to targeted or tailored intervention.

When a teenage nullipara underwent labor induction for preeclampsia at 37 weeks, she was given epidural analgesia and seizure prophylaxis with magnesium sulfate. Her electronic fetal heart rate (FHR) tracing was initially reassuring, with only occasional variable decelerations, but subsequently revealed a baseline of 140 beats per minute (bpm), minimal to absent variability, no accelerations, and variable decelerations to 90 bpm with rapid return to baseline.

The tracing was interpreted as nonreassuring, and a fetal pulse oximeter was inserted. It revealed a fetal oxygen saturation rate between 45% and 50%, and labor was allowed to continue. After 3.5 hours in the second stage, the patient was delivered by outlet forceps. Her infant had Apgar scores of 8 at 1 minute and 9 at 5 minutes. The umbilical arterial pH was 7.25, and base excess was–4.9.

Fetal pulse oximetry made it possible to manage this case without resorting to emergent cesarean. But is this noninvasive technology truly a step forward in intrapartum assessment of fetal well-being?

We describe what the evidence (a single randomized study and a number of observational studies) reveals about these questions:

1. How accurately does fetal pulse oximetry reflect the fetal condition?
2. What is the critical threshold for fetal oxygen desaturation?
3. Is a single reading reliable?
4. Does oximetry correlate with acid-base status?
5. Does the combination of oximetry and electronic monitoring improve accuracy?
6. Will fetal pulse oximetry improve neonatal outcomes?
7. How precise is it?
8. Is it easy to use?

Needed: Effective adjunct to electronic monitoring

Except in the chronically hypoxic fetus (which is affected by the time labor begins), the pathophysiology of acute intrapartum events is a continuum, from hypoxemia to respiratory acidosis to metabolic acidosis and, ultimately, clinical impairment. The goal of intrapartum surveillance is to detect fetal hypoxemia before it progresses to asphyxia and perinatal mortality or long-term morbidity.

Although it is approved as an adjunct to electronic fetal monitoring (EFM), fetal pulse oximetry has gained only sporadic use since it became available in the United States in 2000—even though EFM has proved disappointing as a tool for predicting fetal hypoxia. Only about 10% of US obstetrical units had fetal pulse oximetry technology as of 2002.¹

Clinicians began questioning the reliability of subjective interpretation of fetal heart tracings soon after EFM went into general use. Thirty years later, a meta-analysis of 12 randomized clinical trials involving 58,855 gravidas cast doubt on the benefits of EFM,² which is associated with an increase in operative deliveries as a result of high sensitivity but low specificity in predicting fetal hypoxia and acidosis.

FDA approval was based on sole randomized trial

The only commercially available fetal oximetry sensor, the Nellcor N-400 (Nellcor, Pleasanton, Calif), obtained US Food and Drug Administration (FDA) approval as an adjunct to EFM when the latter indicates a nonreassuring FHR pattern. That approval was based on the only randomized study³ of fetal pulse oximetry conducted, which involved 1,010 women with predefined nonreassuring FHR patterns in labor.

Goal: Reduced cesarean rate with comparable outcomes. Investigators hypothesized that adjunctive fetal oximetry would improve assessment and reduce the cesarean rate without altering neonatal outcome. Indeed, in the oximetry group, the rate of cesarean delivery performed for a nonreassuring FHR tracing (4.5% versus 10.2%; P = .007) was significantly reduced. Other findings:

• Same neonatal outcomes, with no significant differences between the 2 groups.
• Higher cesarean rate for dystocia in the intervention group, offsetting any advantage in the overall cesarean delivery rate (29% versus 26%). This unexpected increase in cesarean deliveries raises several possibilities:
• Given the unblinded design, it is possible that clinicians, circumspect of the pulse oximetry, continued to perform cesareans for nonreassuring FHR, but labeled the indication for surgery differently. The validity of the dystocia diagnosis was discredited by a subsequent partogram analysis that showed a similar rate of arrested labor in both groups.
• A nonreassuring FHR in conditions of normal fetal oxygenation is predictive of dystocia. Previous randomized studies of EFM have suggested the same thing.⁴
• Dystocia is the consequence of the device itself. Anecdotal observations suggest a higher rate of persistent occiput posterior positions with fetal oximetry.

Other trials underway. The ongoing Fetal Oximetry (FOX) trial of the National Institute of Child Health and Human Development Maternal-Fetal Medicine Units Network, involving 10,000 nulliparous participants, is comparing cesarean delivery rates and safety outcomes in patients monitored for FHR plus pulse oximetry with a group in which the clinicians are blinded to the pulse oximetry readings. Another randomized controlled trial of fetal pulse oximetry is underway in Australia.

Potential for increased costs. The American College of Obstetricians and Gynecologists (ACOG) has raised concerns about the potential increase in costs without demonstrable improvement in outcome.⁵ ACOG has not endorsed fetal pulse oximetry for general practice.

Question 1How accurately does pulse oximetry reflect the fetal condition?

It yields only indirect information on the partial pressure of oxygen in the blood and no data on perfusion or acid-base status.

In other clinical settings, oxygen saturation is not an acceptable substitute for arterial blood gas analysis. The pulse oximeter is not a hemoximeter—only that device directly and reliably determines blood oxygen saturation by spectrophotometry.⁶ Even the calculated oxygen saturation values provided automatically by modern blood gas analyzers are inaccurate.⁷

Studies report varying results. In a comparison⁸ of fetal oxygen saturation by hemoximetry in a fetal scalp blood (FSB) sample and fetal arterial oxyhemoglobin saturation (FSpO₂) by pulse oximetry immediately before the blood sampling, the FSpO₂ medians were always higher than the FSB hemoximetry saturation—which led to false-negative results in hypoxic babies.

In animal studies, pulse oximetry correlated well with simultaneously measured arterial oxygen saturation (r = 0.98, P = .01),⁹ but data from human studies are inconsistent. While McNamara et al¹⁰ reported good correlation between FSpO₂ measurements and umbilical artery blood oxygen saturation at birth (r = 0.59, P <.001), Langer et al¹¹ found no relationship between FSpO₂ levels determined during pushing efforts and oxygen saturation in umbilical vein blood at birth.

Possible reasons for the ambiguous findings:

• differences in practice, such as use of umbilical venous versus arterial blood, or measurement during pushing versus between pushes,
• different intervals from FSpO₂ reading to umbilical blood sampling, or
• incomparable groups, such as all women in labor versus those with abnormal FHR.

Limitations. Fetal pulse oximetry measures arterial oxygen saturation during the systolic pulse wave in the skin microcirculation at head level. In the fetus, this is part of the preductal circulation, with oxygen saturation levels somewhere between umbilical arterial and umbilical venous blood oxygen saturation.

Theoretically, FSpO₂ should be closer to FSB than to umbilical blood. Although FSB samples consist of capillary blood, which is not exactly central arterial blood, the differences are small, at least in the neonate.¹² In the intrapartum period, however, several variables with unknown effect may weaken relationships:

• different intervals between the last oximetry signal and blood sampling after delivery
• differences in local tissue perfusion status¹³
• perfusion changes during fetal compromise, as the fetus centralizes its blood flow, with vasoconstriction in the skin circulation

Question 2What is the critical threshold for fetal oxygen desaturation?

Human studies indicate that an FSpO₂ of 33% is approximately the 10th percentile on the normal distribution, and an FSpO₂ of 29% to 30% represents the third to fifth percentiles in normal-outcome labor.¹⁴ Studies in catheterized fetal sheep suggest that the level below which metabolic acidosis can be anticipated is an FSpO₂ of about 30%.¹⁵

The 30% threshold also is supported by prospective human data from a multicenter trial.¹⁶ According to those data, an FSpO₂ of less than 30% has 100% sensitivity in predicting an FSB pH below 7.20. FSpO₂ of less than 30% also correlated with a lack of variability on the FHR tracing.¹⁷

The cutoff of 30% should not be interpreted as an indication of fetal distress, however. Rather, it represents a threshold below which increasing fetal acidosis will be encountered ( FIGURE 1). Oxygen saturation is a dynamic biologic parameter with broad variation.

FIGURE 1 Tracking fetal arterial oxyhemoglobin saturation

Question 3Is a single reading reliable?

The normal fetus has a remarkable capacity to compensate for transient episodes of desaturation. Thus, a single reading cannot reflect the fetal condition; the trend in FSpO₂ must be taken into account. Research indicates only FSpO₂ levels below 30% for more than 2 minutes¹⁸ or more than 10 minutes¹⁹ are likely to be associated with intrapartum acidosis.

ACOG has raised concerns about the potential increase in costs without demonstrable improvement in outcome.

Gorenberg et al²⁰ retrospectively correlated FSpO₂ with umbilical artery pH and found that neither the 30% threshold alone nor the duration of FSpO₂ below 30% correlated with fetal acidemia (pH below 7.20). Rather, the repetition of such episodes was more predictive. The authors concluded that more than 10 episodes of FSpO₂below 30% would overcome the ability of the fetus to compensate.

The study was underpowered to detect a significant difference in acidemia, and did not allow sufficient observation time to detect the natural progression of hypoxia to metabolic acidosis, a better indicator of fetal compromise. Additional research is needed.

Question 4Does oximetry correlate with acid-base status?

Many of the studies mentioned here assumed a correlation. Whenever oxygen saturation in the umbilical artery is 30% or more, acidosis (pH below 7.13) in the same blood is rare—only 1%.²¹ However, the correlation between fetal pulse oximetry values and acid-base status is much weaker.⁸.

Leszczynska-Gorzelak et al²² found no relationship between FSpO₂ levels in the first or second stage of labor and pH or partial pressure of oxygen in umbilical vein blood at delivery. Other investigators concluded similarly, considering intrapartum FSpO₂ of limited use for predicting acidosis at birth, irrespective of FSpO₂ cutoff.^23,24

Rijnders et al²⁴ found no significant correlation between fetal scalp or umbilical artery blood pH and mean FSpO₂ for the last 30 minutes before sampling (r = 0.02, P = .9). Even the lowest FSpO₂ level did not correlate with arterial pH (r = .04, P = .84). None of the study’s 3 cases of umbilical pH below 7.05 would have been detected using the mean FSpO₂ before delivery, and only 1 would have been detected using the lowest FSpO₂.

In another multicenter study involving the Nellcor system in 164 cases with abnormal FHR, a correlation between oximetry and FSB sampling (r = 0.29, P < .01) was noted in the first stage of labor, but second-stage FSpO₂ readings did not correlate with oxygen saturation, partial pressure of oxygen, pH, or bicarbonate level in the umbilical artery at birth.²⁵

An observational series²⁶ of 128 fetuses with nonreassuring FHR patterns concluded that fetal distress was insufficiently identified by oximetry. Only 2 of the 11 cases with umbilical artery pH below 7.20 were detected by pulse oximetry recordings below 30% during the last 30 minutes of the second stage, and out of 5 cases with hypoxic readings in the second stage, only 2 were acidotic at birth. The calculated sensitivity was 18%, specificity 92%, positive predictive value (PPV) 40%, and negative predictive value (NPV) 80%. A low Apgar score was never predicted by fetal pulse oximetry.

Others used the same Nellcor system over the final 30 minutes of labor and a cutoff for umbilical blood acidemia of pH below 7.13 and reported similar numbers: sensitivity 28%, specificity 94%, PPV 40%, and NPV 80%.²³

Vitoratos et al²⁷ analyzed FSpO₂ readings in active labor (not limited to the last 30 minutes before delivery) and obtained somewhat better values: sensitivity 72%, specificity 93%, PPV 61.5%, and NPV 95.8% for an umbilical artery blood pH below 7.15.

The impression that the validity of fetal pulse oximetry is higher in earlier labor than in the second stage is supported by data from Stiller et al. ²⁸ Leszczynska-Gorzelak et al²⁹ found a significant decrease in mean FSpO₂ from the first stage to the second stage of normal labor (51.9% versus 43.8%, P < .001), and Dildy et al¹⁴ noted a similar difference upon analyzing 160 normal labors (59% versus 53%), but other studies failed to verify such differences.^25,30

Observational studies had unrealistic pH cutoff. All the evidence presented thus far on the validity of fetal pulse oximetry in predicting acidemia is based on observational data. A common deficiency is the unrealistic cutoff for pathologic fetal acidemia—a pH of less than 7.13 to 7.20—when it is widely accepted that “pathologic fetal acidemia” reflects an umbilical artery blood pH below 7.³¹ Even in this group, two thirds of neonates are unaffected by morbidity.

Need to identify metabolic acidosis. It also is accepted that the presence of a metabolic component to fetal acidemia may be as important—if not more important—than a single pH cutoff.³¹ Only a few human studies of pulse oximetry have distinguished between respiratory and metabolic acidemia. When they did, intrapartum fetal pulse oximetry was unable to predict umbilical artery base excess.^23,25

The only randomized study failed to determine whether clinical decisions can be based solely on fetal pulse oximetry. ³ The investigators did suggest that sensitivity and specificity for metabolic acidemia was improved in the intervention group—a promising appraisal, in contrast with previous observational data.

In the study, 7 neonates (3 in the intervention group and 4 controls) had umbilical artery blood pH below 7. All 4 controls had vaginal delivery. There also were 6 cases of elevated base excess (ie, -16 mEq/L or below) among controls. None were recorded in the intervention group, and the 3 cases of acidemia were recognized antepartum and led to cesareans.

Unfortunately, the study design did not guarantee that patient management was based exclusively on EFM with or without fetal pulse oximetry. Vibro-acoustic stimulation or FSB sampling was required before proceeding to cesarean delivery in both groups.

It appears that the negative predictive value of fetal oximetry is of greater practical value than other attributes.

When FSpO₂ was less than 30% for the entire interval between 2 contractions, or was unobtainable, the physician was supposed to revert to interpretation of EFM. When that was persistently nonreassuring, the physician was given the option of scalp stimulation or FSB sampling. Thus, it was not determined whether clinical decisions can be based exclusively on fetal pulse oximetry. Schmidt et al²⁶ suggested that such exclusive application of fetal pulse oximetry might actually jeopardize fetal health.

Question 5 Does the combination of oximetry and EFM improve accuracy?

Fetal pulse oximetry was not used independently in any of the studies discussed here, but in association with EFM, which has a sensitivity for fetal acidosis of 93%, specificity of 29%, PPV of 2.6%, and NPV of 99.5%.³²

From a statistical point of view, whenever 2 evaluation methods with the same endpoint (fetal acidosis) are combined, sensitivity decreases while specificity increases, theoretically resulting in less unnecessary intervention. That is exactly what investigators have reported: sensitivity as low as 18%²⁶ for fetal oximetry, and specificity as high as 94%.²³ However, the value of this new technology might not be so much the prediction of acidosis but identification of the well-oxygenated fetus so that labor may be safely continued in the presence of a concerning—but not ominous—FHR tracing.

Question 6Will it improve neonatal outcomes?

Neonatal outcome is the ultimate endpoint in obstetrical care. In the randomized trial by Garite et al,³ there was no difference in neonatal outcome between the groups using or not using fetal pulse oximetry. According to Chua et al,³³ FSpO₂ levels measured even 10 minutes before delivery have no relation to neonatal outcome.

Leszczynska-Gorzelak et al²² believe FSpO₂ is more predictive of neonatal outcome in the first stage than the second. However, Apgar score had no relationship with FSpO₂readings in the first or second stage. Butterwegge³⁴ reported 6 cases of FSpO₂ below 30% for more than 30 minutes, all with good neonatal outcome, and Alshimmiri et al²³ noted that only normal FSpO₂ correlates with fetal well-being. Thus, it appears that the NPV of fetal oximetry is of greater practical value than other attributes.

Question 7How precise is it?

The Nellcor system monitors the quality of FSpO₂ measurement; no value is displayed if the signal lacks the characteristics of a fetal arterial plethysmographic curve or if contact between sensor and skin is insufficient. Because of fetal movements and other artifacts, posting time is always less than 100%.

In the French multicenter study,²⁵ the mean reliable signal time in the first stage of labor was only 64.7%—even less in the second stage (54%). Signal retention was 67% in the randomized trial by Garite et al. ³

Many artifacts may impede signal acquisition and impact the reliability of a reading:

• The sensor’s position on the fetal head. For example, the difference in FSpO₂ readings between the forehead and occiput may be as much as 13.4%. (The sensor is designed to go against the fetal cheek, but may move around.)
• Incomplete sensor-to-skin contact, such as with high fetal head station, -2 or above.
• Marked caput formation.
• Increased intrauterine pressure accompanying contractions, especially at presentation stations of +2 or below (FIGURE 2). FSpO₂ monitoring requires detection of fetal pulses, which may be undetectable when the surrounding pressure is high, resulting in a loss of signal.
• Interposition of vernix or fetal hair.
• Presence of meconium, which behaves like a red-light filter, altering the ratio of red to infrared light and resulting in artificially low values.³⁵ This theoretical concern is rejected by Yam et al,³⁶ who did not observe any effect of meconium on FSpO₂ values. (When the amniotic fluid is meconium-stained, Carbonne et al³⁷ showed that fetal oximetry is a better predictor of meconium aspiration syndrome than FSB sampling.) The data on the influence of meconium on FSpO₂ readings remain contradictory.

All these conditions may impair precision and contribute to poor sensitivity.

FIGURE 2 A weakening signal during pushing

Question 8Is it easy to use?

An Australian survey³⁸ assessed clinicians’ perceptions during placement of the oximetry sensor. Ease of placement was rated as good or excellent in 71% of cases, and the patient’s comfort was rated as good or excellent in 90% of cases. Chua et al³⁹ reported a mean insertion time of 90 seconds, with a reliable signal obtained within 5 minutes in 87% of placements. The French multicenter study²⁵ mentioned earlier concluded that the procedure is satisfactory and easier than FSB sampling. The device itself was harmless to both mother and fetus.⁴⁰

Potential research directions

Fetal pulse oximetry may be an effective tool in clinical scenarios such as:

• fetal arrhythmias with uninterpretable FHR tracing
• fetal tachycardia associated with maternal fever, thyrotoxicosis, or fetal supraventricular tachycardia, when distinguishing other contributions to tachycardia may be difficult
• fetal bradycardia caused by a complete heart block, which may render EFM undecipherable
• when amnioinfusion is attempted for variable decelerations and it is necessary to differentiate a nonreassuring FHR tracing related to transient in utero stress (eg, umbilical cord compressions) from ominous tracings

Another area not yet addressed is cost-effectiveness beyond the immediate direct costs (approximately $11,000 for the monitor and $150 for each disposable sensor). Also uncertain is whether laboring women will accept the device (how disturbing or invasive it is perceived to be) and how acceptable or applicable it is outside tertiary institutions.

Dr. Vidaeff reports no financial relationships relevant to this article. Dr. Ramin reports grant support from the US National Institutes of Health.

When a teenage nullipara underwent labor induction for preeclampsia at 37 weeks, she was given epidural analgesia and seizure prophylaxis with magnesium sulfate. Her electronic fetal heart rate (FHR) tracing was initially reassuring, with only occasional variable decelerations, but subsequently revealed a baseline of 140 beats per minute (bpm), minimal to absent variability, no accelerations, and variable decelerations to 90 bpm with rapid return to baseline.

The tracing was interpreted as nonreassuring, and a fetal pulse oximeter was inserted. It revealed a fetal oxygen saturation rate between 45% and 50%, and labor was allowed to continue. After 3.5 hours in the second stage, the patient was delivered by outlet forceps. Her infant had Apgar scores of 8 at 1 minute and 9 at 5 minutes. The umbilical arterial pH was 7.25, and base excess was–4.9.

Fetal pulse oximetry made it possible to manage this case without resorting to emergent cesarean. But is this noninvasive technology truly a step forward in intrapartum assessment of fetal well-being?

We describe what the evidence (a single randomized study and a number of observational studies) reveals about these questions:

1. How accurately does fetal pulse oximetry reflect the fetal condition?
2. What is the critical threshold for fetal oxygen desaturation?
3. Is a single reading reliable?
4. Does oximetry correlate with acid-base status?
5. Does the combination of oximetry and electronic monitoring improve accuracy?
6. Will fetal pulse oximetry improve neonatal outcomes?
7. How precise is it?
8. Is it easy to use?

Needed: Effective adjunct to electronic monitoring

Except in the chronically hypoxic fetus (which is affected by the time labor begins), the pathophysiology of acute intrapartum events is a continuum, from hypoxemia to respiratory acidosis to metabolic acidosis and, ultimately, clinical impairment. The goal of intrapartum surveillance is to detect fetal hypoxemia before it progresses to asphyxia and perinatal mortality or long-term morbidity.

Although it is approved as an adjunct to electronic fetal monitoring (EFM), fetal pulse oximetry has gained only sporadic use since it became available in the United States in 2000—even though EFM has proved disappointing as a tool for predicting fetal hypoxia. Only about 10% of US obstetrical units had fetal pulse oximetry technology as of 2002.¹

Clinicians began questioning the reliability of subjective interpretation of fetal heart tracings soon after EFM went into general use. Thirty years later, a meta-analysis of 12 randomized clinical trials involving 58,855 gravidas cast doubt on the benefits of EFM,² which is associated with an increase in operative deliveries as a result of high sensitivity but low specificity in predicting fetal hypoxia and acidosis.

FDA approval was based on sole randomized trial

The only commercially available fetal oximetry sensor, the Nellcor N-400 (Nellcor, Pleasanton, Calif), obtained US Food and Drug Administration (FDA) approval as an adjunct to EFM when the latter indicates a nonreassuring FHR pattern. That approval was based on the only randomized study³ of fetal pulse oximetry conducted, which involved 1,010 women with predefined nonreassuring FHR patterns in labor.

Goal: Reduced cesarean rate with comparable outcomes. Investigators hypothesized that adjunctive fetal oximetry would improve assessment and reduce the cesarean rate without altering neonatal outcome. Indeed, in the oximetry group, the rate of cesarean delivery performed for a nonreassuring FHR tracing (4.5% versus 10.2%; P = .007) was significantly reduced. Other findings:

• Same neonatal outcomes, with no significant differences between the 2 groups.
• Higher cesarean rate for dystocia in the intervention group, offsetting any advantage in the overall cesarean delivery rate (29% versus 26%). This unexpected increase in cesarean deliveries raises several possibilities:
• Given the unblinded design, it is possible that clinicians, circumspect of the pulse oximetry, continued to perform cesareans for nonreassuring FHR, but labeled the indication for surgery differently. The validity of the dystocia diagnosis was discredited by a subsequent partogram analysis that showed a similar rate of arrested labor in both groups.
• A nonreassuring FHR in conditions of normal fetal oxygenation is predictive of dystocia. Previous randomized studies of EFM have suggested the same thing.⁴
• Dystocia is the consequence of the device itself. Anecdotal observations suggest a higher rate of persistent occiput posterior positions with fetal oximetry.

Other trials underway. The ongoing Fetal Oximetry (FOX) trial of the National Institute of Child Health and Human Development Maternal-Fetal Medicine Units Network, involving 10,000 nulliparous participants, is comparing cesarean delivery rates and safety outcomes in patients monitored for FHR plus pulse oximetry with a group in which the clinicians are blinded to the pulse oximetry readings. Another randomized controlled trial of fetal pulse oximetry is underway in Australia.

Potential for increased costs. The American College of Obstetricians and Gynecologists (ACOG) has raised concerns about the potential increase in costs without demonstrable improvement in outcome.⁵ ACOG has not endorsed fetal pulse oximetry for general practice.

Question 1How accurately does pulse oximetry reflect the fetal condition?

It yields only indirect information on the partial pressure of oxygen in the blood and no data on perfusion or acid-base status.

In other clinical settings, oxygen saturation is not an acceptable substitute for arterial blood gas analysis. The pulse oximeter is not a hemoximeter—only that device directly and reliably determines blood oxygen saturation by spectrophotometry.⁶ Even the calculated oxygen saturation values provided automatically by modern blood gas analyzers are inaccurate.⁷

Studies report varying results. In a comparison⁸ of fetal oxygen saturation by hemoximetry in a fetal scalp blood (FSB) sample and fetal arterial oxyhemoglobin saturation (FSpO₂) by pulse oximetry immediately before the blood sampling, the FSpO₂ medians were always higher than the FSB hemoximetry saturation—which led to false-negative results in hypoxic babies.

In animal studies, pulse oximetry correlated well with simultaneously measured arterial oxygen saturation (r = 0.98, P = .01),⁹ but data from human studies are inconsistent. While McNamara et al¹⁰ reported good correlation between FSpO₂ measurements and umbilical artery blood oxygen saturation at birth (r = 0.59, P <.001), Langer et al¹¹ found no relationship between FSpO₂ levels determined during pushing efforts and oxygen saturation in umbilical vein blood at birth.

Possible reasons for the ambiguous findings:

• differences in practice, such as use of umbilical venous versus arterial blood, or measurement during pushing versus between pushes,
• different intervals from FSpO₂ reading to umbilical blood sampling, or
• incomparable groups, such as all women in labor versus those with abnormal FHR.

Limitations. Fetal pulse oximetry measures arterial oxygen saturation during the systolic pulse wave in the skin microcirculation at head level. In the fetus, this is part of the preductal circulation, with oxygen saturation levels somewhere between umbilical arterial and umbilical venous blood oxygen saturation.

Theoretically, FSpO₂ should be closer to FSB than to umbilical blood. Although FSB samples consist of capillary blood, which is not exactly central arterial blood, the differences are small, at least in the neonate.¹² In the intrapartum period, however, several variables with unknown effect may weaken relationships:

• different intervals between the last oximetry signal and blood sampling after delivery
• differences in local tissue perfusion status¹³
• perfusion changes during fetal compromise, as the fetus centralizes its blood flow, with vasoconstriction in the skin circulation

Question 2What is the critical threshold for fetal oxygen desaturation?

Human studies indicate that an FSpO₂ of 33% is approximately the 10th percentile on the normal distribution, and an FSpO₂ of 29% to 30% represents the third to fifth percentiles in normal-outcome labor.¹⁴ Studies in catheterized fetal sheep suggest that the level below which metabolic acidosis can be anticipated is an FSpO₂ of about 30%.¹⁵

The 30% threshold also is supported by prospective human data from a multicenter trial.¹⁶ According to those data, an FSpO₂ of less than 30% has 100% sensitivity in predicting an FSB pH below 7.20. FSpO₂ of less than 30% also correlated with a lack of variability on the FHR tracing.¹⁷

The cutoff of 30% should not be interpreted as an indication of fetal distress, however. Rather, it represents a threshold below which increasing fetal acidosis will be encountered ( FIGURE 1). Oxygen saturation is a dynamic biologic parameter with broad variation.

FIGURE 1 Tracking fetal arterial oxyhemoglobin saturation

Question 3Is a single reading reliable?

The normal fetus has a remarkable capacity to compensate for transient episodes of desaturation. Thus, a single reading cannot reflect the fetal condition; the trend in FSpO₂ must be taken into account. Research indicates only FSpO₂ levels below 30% for more than 2 minutes¹⁸ or more than 10 minutes¹⁹ are likely to be associated with intrapartum acidosis.

ACOG has raised concerns about the potential increase in costs without demonstrable improvement in outcome.

Gorenberg et al²⁰ retrospectively correlated FSpO₂ with umbilical artery pH and found that neither the 30% threshold alone nor the duration of FSpO₂ below 30% correlated with fetal acidemia (pH below 7.20). Rather, the repetition of such episodes was more predictive. The authors concluded that more than 10 episodes of FSpO₂below 30% would overcome the ability of the fetus to compensate.

The study was underpowered to detect a significant difference in acidemia, and did not allow sufficient observation time to detect the natural progression of hypoxia to metabolic acidosis, a better indicator of fetal compromise. Additional research is needed.

Question 4Does oximetry correlate with acid-base status?

Many of the studies mentioned here assumed a correlation. Whenever oxygen saturation in the umbilical artery is 30% or more, acidosis (pH below 7.13) in the same blood is rare—only 1%.²¹ However, the correlation between fetal pulse oximetry values and acid-base status is much weaker.⁸.

Leszczynska-Gorzelak et al²² found no relationship between FSpO₂ levels in the first or second stage of labor and pH or partial pressure of oxygen in umbilical vein blood at delivery. Other investigators concluded similarly, considering intrapartum FSpO₂ of limited use for predicting acidosis at birth, irrespective of FSpO₂ cutoff.^23,24

Rijnders et al²⁴ found no significant correlation between fetal scalp or umbilical artery blood pH and mean FSpO₂ for the last 30 minutes before sampling (r = 0.02, P = .9). Even the lowest FSpO₂ level did not correlate with arterial pH (r = .04, P = .84). None of the study’s 3 cases of umbilical pH below 7.05 would have been detected using the mean FSpO₂ before delivery, and only 1 would have been detected using the lowest FSpO₂.

In another multicenter study involving the Nellcor system in 164 cases with abnormal FHR, a correlation between oximetry and FSB sampling (r = 0.29, P < .01) was noted in the first stage of labor, but second-stage FSpO₂ readings did not correlate with oxygen saturation, partial pressure of oxygen, pH, or bicarbonate level in the umbilical artery at birth.²⁵

An observational series²⁶ of 128 fetuses with nonreassuring FHR patterns concluded that fetal distress was insufficiently identified by oximetry. Only 2 of the 11 cases with umbilical artery pH below 7.20 were detected by pulse oximetry recordings below 30% during the last 30 minutes of the second stage, and out of 5 cases with hypoxic readings in the second stage, only 2 were acidotic at birth. The calculated sensitivity was 18%, specificity 92%, positive predictive value (PPV) 40%, and negative predictive value (NPV) 80%. A low Apgar score was never predicted by fetal pulse oximetry.

Others used the same Nellcor system over the final 30 minutes of labor and a cutoff for umbilical blood acidemia of pH below 7.13 and reported similar numbers: sensitivity 28%, specificity 94%, PPV 40%, and NPV 80%.²³

Vitoratos et al²⁷ analyzed FSpO₂ readings in active labor (not limited to the last 30 minutes before delivery) and obtained somewhat better values: sensitivity 72%, specificity 93%, PPV 61.5%, and NPV 95.8% for an umbilical artery blood pH below 7.15.

The impression that the validity of fetal pulse oximetry is higher in earlier labor than in the second stage is supported by data from Stiller et al. ²⁸ Leszczynska-Gorzelak et al²⁹ found a significant decrease in mean FSpO₂ from the first stage to the second stage of normal labor (51.9% versus 43.8%, P < .001), and Dildy et al¹⁴ noted a similar difference upon analyzing 160 normal labors (59% versus 53%), but other studies failed to verify such differences.^25,30

Observational studies had unrealistic pH cutoff. All the evidence presented thus far on the validity of fetal pulse oximetry in predicting acidemia is based on observational data. A common deficiency is the unrealistic cutoff for pathologic fetal acidemia—a pH of less than 7.13 to 7.20—when it is widely accepted that “pathologic fetal acidemia” reflects an umbilical artery blood pH below 7.³¹ Even in this group, two thirds of neonates are unaffected by morbidity.

Need to identify metabolic acidosis. It also is accepted that the presence of a metabolic component to fetal acidemia may be as important—if not more important—than a single pH cutoff.³¹ Only a few human studies of pulse oximetry have distinguished between respiratory and metabolic acidemia. When they did, intrapartum fetal pulse oximetry was unable to predict umbilical artery base excess.^23,25

The only randomized study failed to determine whether clinical decisions can be based solely on fetal pulse oximetry. ³ The investigators did suggest that sensitivity and specificity for metabolic acidemia was improved in the intervention group—a promising appraisal, in contrast with previous observational data.

In the study, 7 neonates (3 in the intervention group and 4 controls) had umbilical artery blood pH below 7. All 4 controls had vaginal delivery. There also were 6 cases of elevated base excess (ie, -16 mEq/L or below) among controls. None were recorded in the intervention group, and the 3 cases of acidemia were recognized antepartum and led to cesareans.

Unfortunately, the study design did not guarantee that patient management was based exclusively on EFM with or without fetal pulse oximetry. Vibro-acoustic stimulation or FSB sampling was required before proceeding to cesarean delivery in both groups.

It appears that the negative predictive value of fetal oximetry is of greater practical value than other attributes.

When FSpO₂ was less than 30% for the entire interval between 2 contractions, or was unobtainable, the physician was supposed to revert to interpretation of EFM. When that was persistently nonreassuring, the physician was given the option of scalp stimulation or FSB sampling. Thus, it was not determined whether clinical decisions can be based exclusively on fetal pulse oximetry. Schmidt et al²⁶ suggested that such exclusive application of fetal pulse oximetry might actually jeopardize fetal health.

Question 5 Does the combination of oximetry and EFM improve accuracy?

Fetal pulse oximetry was not used independently in any of the studies discussed here, but in association with EFM, which has a sensitivity for fetal acidosis of 93%, specificity of 29%, PPV of 2.6%, and NPV of 99.5%.³²

From a statistical point of view, whenever 2 evaluation methods with the same endpoint (fetal acidosis) are combined, sensitivity decreases while specificity increases, theoretically resulting in less unnecessary intervention. That is exactly what investigators have reported: sensitivity as low as 18%²⁶ for fetal oximetry, and specificity as high as 94%.²³ However, the value of this new technology might not be so much the prediction of acidosis but identification of the well-oxygenated fetus so that labor may be safely continued in the presence of a concerning—but not ominous—FHR tracing.

Question 6Will it improve neonatal outcomes?

Neonatal outcome is the ultimate endpoint in obstetrical care. In the randomized trial by Garite et al,³ there was no difference in neonatal outcome between the groups using or not using fetal pulse oximetry. According to Chua et al,³³ FSpO₂ levels measured even 10 minutes before delivery have no relation to neonatal outcome.

Leszczynska-Gorzelak et al²² believe FSpO₂ is more predictive of neonatal outcome in the first stage than the second. However, Apgar score had no relationship with FSpO₂readings in the first or second stage. Butterwegge³⁴ reported 6 cases of FSpO₂ below 30% for more than 30 minutes, all with good neonatal outcome, and Alshimmiri et al²³ noted that only normal FSpO₂ correlates with fetal well-being. Thus, it appears that the NPV of fetal oximetry is of greater practical value than other attributes.

Question 7How precise is it?

The Nellcor system monitors the quality of FSpO₂ measurement; no value is displayed if the signal lacks the characteristics of a fetal arterial plethysmographic curve or if contact between sensor and skin is insufficient. Because of fetal movements and other artifacts, posting time is always less than 100%.

In the French multicenter study,²⁵ the mean reliable signal time in the first stage of labor was only 64.7%—even less in the second stage (54%). Signal retention was 67% in the randomized trial by Garite et al. ³

Many artifacts may impede signal acquisition and impact the reliability of a reading:

• The sensor’s position on the fetal head. For example, the difference in FSpO₂ readings between the forehead and occiput may be as much as 13.4%. (The sensor is designed to go against the fetal cheek, but may move around.)
• Incomplete sensor-to-skin contact, such as with high fetal head station, -2 or above.
• Marked caput formation.
• Increased intrauterine pressure accompanying contractions, especially at presentation stations of +2 or below (FIGURE 2). FSpO₂ monitoring requires detection of fetal pulses, which may be undetectable when the surrounding pressure is high, resulting in a loss of signal.
• Interposition of vernix or fetal hair.
• Presence of meconium, which behaves like a red-light filter, altering the ratio of red to infrared light and resulting in artificially low values.³⁵ This theoretical concern is rejected by Yam et al,³⁶ who did not observe any effect of meconium on FSpO₂ values. (When the amniotic fluid is meconium-stained, Carbonne et al³⁷ showed that fetal oximetry is a better predictor of meconium aspiration syndrome than FSB sampling.) The data on the influence of meconium on FSpO₂ readings remain contradictory.

All these conditions may impair precision and contribute to poor sensitivity.

FIGURE 2 A weakening signal during pushing

Question 8Is it easy to use?

An Australian survey³⁸ assessed clinicians’ perceptions during placement of the oximetry sensor. Ease of placement was rated as good or excellent in 71% of cases, and the patient’s comfort was rated as good or excellent in 90% of cases. Chua et al³⁹ reported a mean insertion time of 90 seconds, with a reliable signal obtained within 5 minutes in 87% of placements. The French multicenter study²⁵ mentioned earlier concluded that the procedure is satisfactory and easier than FSB sampling. The device itself was harmless to both mother and fetus.⁴⁰

Potential research directions

Fetal pulse oximetry may be an effective tool in clinical scenarios such as:

• fetal arrhythmias with uninterpretable FHR tracing
• fetal tachycardia associated with maternal fever, thyrotoxicosis, or fetal supraventricular tachycardia, when distinguishing other contributions to tachycardia may be difficult
• fetal bradycardia caused by a complete heart block, which may render EFM undecipherable
• when amnioinfusion is attempted for variable decelerations and it is necessary to differentiate a nonreassuring FHR tracing related to transient in utero stress (eg, umbilical cord compressions) from ominous tracings

Another area not yet addressed is cost-effectiveness beyond the immediate direct costs (approximately $11,000 for the monitor and $150 for each disposable sensor). Also uncertain is whether laboring women will accept the device (how disturbing or invasive it is perceived to be) and how acceptable or applicable it is outside tertiary institutions.

Dr. Vidaeff reports no financial relationships relevant to this article. Dr. Ramin reports grant support from the US National Institutes of Health.

When a teenage nullipara underwent labor induction for preeclampsia at 37 weeks, she was given epidural analgesia and seizure prophylaxis with magnesium sulfate. Her electronic fetal heart rate (FHR) tracing was initially reassuring, with only occasional variable decelerations, but subsequently revealed a baseline of 140 beats per minute (bpm), minimal to absent variability, no accelerations, and variable decelerations to 90 bpm with rapid return to baseline.

The tracing was interpreted as nonreassuring, and a fetal pulse oximeter was inserted. It revealed a fetal oxygen saturation rate between 45% and 50%, and labor was allowed to continue. After 3.5 hours in the second stage, the patient was delivered by outlet forceps. Her infant had Apgar scores of 8 at 1 minute and 9 at 5 minutes. The umbilical arterial pH was 7.25, and base excess was–4.9.

Fetal pulse oximetry made it possible to manage this case without resorting to emergent cesarean. But is this noninvasive technology truly a step forward in intrapartum assessment of fetal well-being?

We describe what the evidence (a single randomized study and a number of observational studies) reveals about these questions:

1. How accurately does fetal pulse oximetry reflect the fetal condition?
2. What is the critical threshold for fetal oxygen desaturation?
3. Is a single reading reliable?
4. Does oximetry correlate with acid-base status?
5. Does the combination of oximetry and electronic monitoring improve accuracy?
6. Will fetal pulse oximetry improve neonatal outcomes?
7. How precise is it?
8. Is it easy to use?

Needed: Effective adjunct to electronic monitoring

Except in the chronically hypoxic fetus (which is affected by the time labor begins), the pathophysiology of acute intrapartum events is a continuum, from hypoxemia to respiratory acidosis to metabolic acidosis and, ultimately, clinical impairment. The goal of intrapartum surveillance is to detect fetal hypoxemia before it progresses to asphyxia and perinatal mortality or long-term morbidity.

Although it is approved as an adjunct to electronic fetal monitoring (EFM), fetal pulse oximetry has gained only sporadic use since it became available in the United States in 2000—even though EFM has proved disappointing as a tool for predicting fetal hypoxia. Only about 10% of US obstetrical units had fetal pulse oximetry technology as of 2002.¹

Clinicians began questioning the reliability of subjective interpretation of fetal heart tracings soon after EFM went into general use. Thirty years later, a meta-analysis of 12 randomized clinical trials involving 58,855 gravidas cast doubt on the benefits of EFM,² which is associated with an increase in operative deliveries as a result of high sensitivity but low specificity in predicting fetal hypoxia and acidosis.

FDA approval was based on sole randomized trial

The only commercially available fetal oximetry sensor, the Nellcor N-400 (Nellcor, Pleasanton, Calif), obtained US Food and Drug Administration (FDA) approval as an adjunct to EFM when the latter indicates a nonreassuring FHR pattern. That approval was based on the only randomized study³ of fetal pulse oximetry conducted, which involved 1,010 women with predefined nonreassuring FHR patterns in labor.

Goal: Reduced cesarean rate with comparable outcomes. Investigators hypothesized that adjunctive fetal oximetry would improve assessment and reduce the cesarean rate without altering neonatal outcome. Indeed, in the oximetry group, the rate of cesarean delivery performed for a nonreassuring FHR tracing (4.5% versus 10.2%; P = .007) was significantly reduced. Other findings:

• Same neonatal outcomes, with no significant differences between the 2 groups.
• Higher cesarean rate for dystocia in the intervention group, offsetting any advantage in the overall cesarean delivery rate (29% versus 26%). This unexpected increase in cesarean deliveries raises several possibilities:
• Given the unblinded design, it is possible that clinicians, circumspect of the pulse oximetry, continued to perform cesareans for nonreassuring FHR, but labeled the indication for surgery differently. The validity of the dystocia diagnosis was discredited by a subsequent partogram analysis that showed a similar rate of arrested labor in both groups.
• A nonreassuring FHR in conditions of normal fetal oxygenation is predictive of dystocia. Previous randomized studies of EFM have suggested the same thing.⁴
• Dystocia is the consequence of the device itself. Anecdotal observations suggest a higher rate of persistent occiput posterior positions with fetal oximetry.

Other trials underway. The ongoing Fetal Oximetry (FOX) trial of the National Institute of Child Health and Human Development Maternal-Fetal Medicine Units Network, involving 10,000 nulliparous participants, is comparing cesarean delivery rates and safety outcomes in patients monitored for FHR plus pulse oximetry with a group in which the clinicians are blinded to the pulse oximetry readings. Another randomized controlled trial of fetal pulse oximetry is underway in Australia.

Potential for increased costs. The American College of Obstetricians and Gynecologists (ACOG) has raised concerns about the potential increase in costs without demonstrable improvement in outcome.⁵ ACOG has not endorsed fetal pulse oximetry for general practice.

Question 1How accurately does pulse oximetry reflect the fetal condition?

It yields only indirect information on the partial pressure of oxygen in the blood and no data on perfusion or acid-base status.

In other clinical settings, oxygen saturation is not an acceptable substitute for arterial blood gas analysis. The pulse oximeter is not a hemoximeter—only that device directly and reliably determines blood oxygen saturation by spectrophotometry.⁶ Even the calculated oxygen saturation values provided automatically by modern blood gas analyzers are inaccurate.⁷

Studies report varying results. In a comparison⁸ of fetal oxygen saturation by hemoximetry in a fetal scalp blood (FSB) sample and fetal arterial oxyhemoglobin saturation (FSpO₂) by pulse oximetry immediately before the blood sampling, the FSpO₂ medians were always higher than the FSB hemoximetry saturation—which led to false-negative results in hypoxic babies.

In animal studies, pulse oximetry correlated well with simultaneously measured arterial oxygen saturation (r = 0.98, P = .01),⁹ but data from human studies are inconsistent. While McNamara et al¹⁰ reported good correlation between FSpO₂ measurements and umbilical artery blood oxygen saturation at birth (r = 0.59, P <.001), Langer et al¹¹ found no relationship between FSpO₂ levels determined during pushing efforts and oxygen saturation in umbilical vein blood at birth.

Possible reasons for the ambiguous findings:

• differences in practice, such as use of umbilical venous versus arterial blood, or measurement during pushing versus between pushes,
• different intervals from FSpO₂ reading to umbilical blood sampling, or
• incomparable groups, such as all women in labor versus those with abnormal FHR.

Limitations. Fetal pulse oximetry measures arterial oxygen saturation during the systolic pulse wave in the skin microcirculation at head level. In the fetus, this is part of the preductal circulation, with oxygen saturation levels somewhere between umbilical arterial and umbilical venous blood oxygen saturation.

Theoretically, FSpO₂ should be closer to FSB than to umbilical blood. Although FSB samples consist of capillary blood, which is not exactly central arterial blood, the differences are small, at least in the neonate.¹² In the intrapartum period, however, several variables with unknown effect may weaken relationships:

• different intervals between the last oximetry signal and blood sampling after delivery
• differences in local tissue perfusion status¹³
• perfusion changes during fetal compromise, as the fetus centralizes its blood flow, with vasoconstriction in the skin circulation

Question 2What is the critical threshold for fetal oxygen desaturation?

Human studies indicate that an FSpO₂ of 33% is approximately the 10th percentile on the normal distribution, and an FSpO₂ of 29% to 30% represents the third to fifth percentiles in normal-outcome labor.¹⁴ Studies in catheterized fetal sheep suggest that the level below which metabolic acidosis can be anticipated is an FSpO₂ of about 30%.¹⁵

The 30% threshold also is supported by prospective human data from a multicenter trial.¹⁶ According to those data, an FSpO₂ of less than 30% has 100% sensitivity in predicting an FSB pH below 7.20. FSpO₂ of less than 30% also correlated with a lack of variability on the FHR tracing.¹⁷

The cutoff of 30% should not be interpreted as an indication of fetal distress, however. Rather, it represents a threshold below which increasing fetal acidosis will be encountered ( FIGURE 1). Oxygen saturation is a dynamic biologic parameter with broad variation.

FIGURE 1 Tracking fetal arterial oxyhemoglobin saturation

Question 3Is a single reading reliable?

The normal fetus has a remarkable capacity to compensate for transient episodes of desaturation. Thus, a single reading cannot reflect the fetal condition; the trend in FSpO₂ must be taken into account. Research indicates only FSpO₂ levels below 30% for more than 2 minutes¹⁸ or more than 10 minutes¹⁹ are likely to be associated with intrapartum acidosis.

ACOG has raised concerns about the potential increase in costs without demonstrable improvement in outcome.

Gorenberg et al²⁰ retrospectively correlated FSpO₂ with umbilical artery pH and found that neither the 30% threshold alone nor the duration of FSpO₂ below 30% correlated with fetal acidemia (pH below 7.20). Rather, the repetition of such episodes was more predictive. The authors concluded that more than 10 episodes of FSpO₂below 30% would overcome the ability of the fetus to compensate.

The study was underpowered to detect a significant difference in acidemia, and did not allow sufficient observation time to detect the natural progression of hypoxia to metabolic acidosis, a better indicator of fetal compromise. Additional research is needed.

Question 4Does oximetry correlate with acid-base status?

Many of the studies mentioned here assumed a correlation. Whenever oxygen saturation in the umbilical artery is 30% or more, acidosis (pH below 7.13) in the same blood is rare—only 1%.²¹ However, the correlation between fetal pulse oximetry values and acid-base status is much weaker.⁸.

Leszczynska-Gorzelak et al²² found no relationship between FSpO₂ levels in the first or second stage of labor and pH or partial pressure of oxygen in umbilical vein blood at delivery. Other investigators concluded similarly, considering intrapartum FSpO₂ of limited use for predicting acidosis at birth, irrespective of FSpO₂ cutoff.^23,24

Rijnders et al²⁴ found no significant correlation between fetal scalp or umbilical artery blood pH and mean FSpO₂ for the last 30 minutes before sampling (r = 0.02, P = .9). Even the lowest FSpO₂ level did not correlate with arterial pH (r = .04, P = .84). None of the study’s 3 cases of umbilical pH below 7.05 would have been detected using the mean FSpO₂ before delivery, and only 1 would have been detected using the lowest FSpO₂.

In another multicenter study involving the Nellcor system in 164 cases with abnormal FHR, a correlation between oximetry and FSB sampling (r = 0.29, P < .01) was noted in the first stage of labor, but second-stage FSpO₂ readings did not correlate with oxygen saturation, partial pressure of oxygen, pH, or bicarbonate level in the umbilical artery at birth.²⁵

An observational series²⁶ of 128 fetuses with nonreassuring FHR patterns concluded that fetal distress was insufficiently identified by oximetry. Only 2 of the 11 cases with umbilical artery pH below 7.20 were detected by pulse oximetry recordings below 30% during the last 30 minutes of the second stage, and out of 5 cases with hypoxic readings in the second stage, only 2 were acidotic at birth. The calculated sensitivity was 18%, specificity 92%, positive predictive value (PPV) 40%, and negative predictive value (NPV) 80%. A low Apgar score was never predicted by fetal pulse oximetry.

Others used the same Nellcor system over the final 30 minutes of labor and a cutoff for umbilical blood acidemia of pH below 7.13 and reported similar numbers: sensitivity 28%, specificity 94%, PPV 40%, and NPV 80%.²³

Vitoratos et al²⁷ analyzed FSpO₂ readings in active labor (not limited to the last 30 minutes before delivery) and obtained somewhat better values: sensitivity 72%, specificity 93%, PPV 61.5%, and NPV 95.8% for an umbilical artery blood pH below 7.15.

The impression that the validity of fetal pulse oximetry is higher in earlier labor than in the second stage is supported by data from Stiller et al. ²⁸ Leszczynska-Gorzelak et al²⁹ found a significant decrease in mean FSpO₂ from the first stage to the second stage of normal labor (51.9% versus 43.8%, P < .001), and Dildy et al¹⁴ noted a similar difference upon analyzing 160 normal labors (59% versus 53%), but other studies failed to verify such differences.^25,30

Observational studies had unrealistic pH cutoff. All the evidence presented thus far on the validity of fetal pulse oximetry in predicting acidemia is based on observational data. A common deficiency is the unrealistic cutoff for pathologic fetal acidemia—a pH of less than 7.13 to 7.20—when it is widely accepted that “pathologic fetal acidemia” reflects an umbilical artery blood pH below 7.³¹ Even in this group, two thirds of neonates are unaffected by morbidity.

Need to identify metabolic acidosis. It also is accepted that the presence of a metabolic component to fetal acidemia may be as important—if not more important—than a single pH cutoff.³¹ Only a few human studies of pulse oximetry have distinguished between respiratory and metabolic acidemia. When they did, intrapartum fetal pulse oximetry was unable to predict umbilical artery base excess.^23,25

The only randomized study failed to determine whether clinical decisions can be based solely on fetal pulse oximetry. ³ The investigators did suggest that sensitivity and specificity for metabolic acidemia was improved in the intervention group—a promising appraisal, in contrast with previous observational data.

In the study, 7 neonates (3 in the intervention group and 4 controls) had umbilical artery blood pH below 7. All 4 controls had vaginal delivery. There also were 6 cases of elevated base excess (ie, -16 mEq/L or below) among controls. None were recorded in the intervention group, and the 3 cases of acidemia were recognized antepartum and led to cesareans.

Unfortunately, the study design did not guarantee that patient management was based exclusively on EFM with or without fetal pulse oximetry. Vibro-acoustic stimulation or FSB sampling was required before proceeding to cesarean delivery in both groups.

It appears that the negative predictive value of fetal oximetry is of greater practical value than other attributes.

When FSpO₂ was less than 30% for the entire interval between 2 contractions, or was unobtainable, the physician was supposed to revert to interpretation of EFM. When that was persistently nonreassuring, the physician was given the option of scalp stimulation or FSB sampling. Thus, it was not determined whether clinical decisions can be based exclusively on fetal pulse oximetry. Schmidt et al²⁶ suggested that such exclusive application of fetal pulse oximetry might actually jeopardize fetal health.

Question 5 Does the combination of oximetry and EFM improve accuracy?

Fetal pulse oximetry was not used independently in any of the studies discussed here, but in association with EFM, which has a sensitivity for fetal acidosis of 93%, specificity of 29%, PPV of 2.6%, and NPV of 99.5%.³²

From a statistical point of view, whenever 2 evaluation methods with the same endpoint (fetal acidosis) are combined, sensitivity decreases while specificity increases, theoretically resulting in less unnecessary intervention. That is exactly what investigators have reported: sensitivity as low as 18%²⁶ for fetal oximetry, and specificity as high as 94%.²³ However, the value of this new technology might not be so much the prediction of acidosis but identification of the well-oxygenated fetus so that labor may be safely continued in the presence of a concerning—but not ominous—FHR tracing.

Question 6Will it improve neonatal outcomes?

Neonatal outcome is the ultimate endpoint in obstetrical care. In the randomized trial by Garite et al,³ there was no difference in neonatal outcome between the groups using or not using fetal pulse oximetry. According to Chua et al,³³ FSpO₂ levels measured even 10 minutes before delivery have no relation to neonatal outcome.

Leszczynska-Gorzelak et al²² believe FSpO₂ is more predictive of neonatal outcome in the first stage than the second. However, Apgar score had no relationship with FSpO₂readings in the first or second stage. Butterwegge³⁴ reported 6 cases of FSpO₂ below 30% for more than 30 minutes, all with good neonatal outcome, and Alshimmiri et al²³ noted that only normal FSpO₂ correlates with fetal well-being. Thus, it appears that the NPV of fetal oximetry is of greater practical value than other attributes.

Question 7How precise is it?

The Nellcor system monitors the quality of FSpO₂ measurement; no value is displayed if the signal lacks the characteristics of a fetal arterial plethysmographic curve or if contact between sensor and skin is insufficient. Because of fetal movements and other artifacts, posting time is always less than 100%.

In the French multicenter study,²⁵ the mean reliable signal time in the first stage of labor was only 64.7%—even less in the second stage (54%). Signal retention was 67% in the randomized trial by Garite et al. ³

Many artifacts may impede signal acquisition and impact the reliability of a reading:

• The sensor’s position on the fetal head. For example, the difference in FSpO₂ readings between the forehead and occiput may be as much as 13.4%. (The sensor is designed to go against the fetal cheek, but may move around.)
• Incomplete sensor-to-skin contact, such as with high fetal head station, -2 or above.
• Marked caput formation.
• Increased intrauterine pressure accompanying contractions, especially at presentation stations of +2 or below (FIGURE 2). FSpO₂ monitoring requires detection of fetal pulses, which may be undetectable when the surrounding pressure is high, resulting in a loss of signal.
• Interposition of vernix or fetal hair.
• Presence of meconium, which behaves like a red-light filter, altering the ratio of red to infrared light and resulting in artificially low values.³⁵ This theoretical concern is rejected by Yam et al,³⁶ who did not observe any effect of meconium on FSpO₂ values. (When the amniotic fluid is meconium-stained, Carbonne et al³⁷ showed that fetal oximetry is a better predictor of meconium aspiration syndrome than FSB sampling.) The data on the influence of meconium on FSpO₂ readings remain contradictory.

All these conditions may impair precision and contribute to poor sensitivity.

FIGURE 2 A weakening signal during pushing

Question 8Is it easy to use?

An Australian survey³⁸ assessed clinicians’ perceptions during placement of the oximetry sensor. Ease of placement was rated as good or excellent in 71% of cases, and the patient’s comfort was rated as good or excellent in 90% of cases. Chua et al³⁹ reported a mean insertion time of 90 seconds, with a reliable signal obtained within 5 minutes in 87% of placements. The French multicenter study²⁵ mentioned earlier concluded that the procedure is satisfactory and easier than FSB sampling. The device itself was harmless to both mother and fetus.⁴⁰

Potential research directions

Fetal pulse oximetry may be an effective tool in clinical scenarios such as:

• fetal arrhythmias with uninterpretable FHR tracing
• fetal tachycardia associated with maternal fever, thyrotoxicosis, or fetal supraventricular tachycardia, when distinguishing other contributions to tachycardia may be difficult
• fetal bradycardia caused by a complete heart block, which may render EFM undecipherable
• when amnioinfusion is attempted for variable decelerations and it is necessary to differentiate a nonreassuring FHR tracing related to transient in utero stress (eg, umbilical cord compressions) from ominous tracings

Another area not yet addressed is cost-effectiveness beyond the immediate direct costs (approximately $11,000 for the monitor and $150 for each disposable sensor). Also uncertain is whether laboring women will accept the device (how disturbing or invasive it is perceived to be) and how acceptable or applicable it is outside tertiary institutions.

Dr. Vidaeff reports no financial relationships relevant to this article. Dr. Ramin reports grant support from the US National Institutes of Health.