A C TA Obstetricia et Gynecologica
AOGS LE TT E R TO THE EDIT O R
On the evidence for intrapartum fetal monitoring with ECG-ST analysis
Sir We appreciate the two commentaries by Øian and Blix (1,2) on our reviews of the evidence supporting the electrocardiogram ST interval analysis (STAN) concept as reflected in the randomized controlled trials (RCTs) (3), and the critical appraisal of the published meta-analyses of the STAN system (4). Øian and Blix state that the criteria used for quality assessment of the five RCTs are unclear (1). We respectfully disagree, as our review (3) explicitly identifies each element considered for such assessment: power calculations of outcome variables, pre-trial training, inclusion criteria, randomization and recruitment pace, management protocols during labor, intrapartum interventions, outcomes of metabolic acidosis, and neonatal intensive care admissions. Øian and Blix (1) consider the RCT to be the reference standard for clinical trials, but remark that it is often conducted in a selected population sample studied during highly controlled conditions not typical of routine clinical care. To demonstrate the external validity of ST analysis, observational studies conducted in different clinical settings (5–8) have shown benefits similar to those found in the highest quality RCTs. We agree with the statement that “the STAN technology is not just a machine, it is a concept” that requires education, training and interpretation (1). In Table 1 we have compiled data (9,10) to illustrate the importance of training and experience for the STAN system. It was mainly in the second half of the RCTs that the benefits of STAN monitoring were achieved. The French RCT (11) was excluded from our meta-analysis (4) because the entry criteria violated the STAN clinical guidelines. Blix and Øian (2) argue that “including the French trial [in
the meta-analysis] would provide a better estimate of the actual effectiveness of STAN, as deviations from STAN guidelines are frequent in clinical practice”. However, deviations in the clinical use of a method must be distinguished from deviations in the entry criteria of an RCT. The requirement that STAN monitoring must commence before deterioration of the fetal condition occurs was not unanimously secured in the French RCT. To establish a causal relation between an intervention and an outcome, a nonambiguous temporal precedence must exist, where the intervention occurs before the outcome (12). Although cases with non-reassuring cardiotograms were allowed to enter other RCTs, our reviews (3,4) showed that patients enrolled in the French RCT were at considerably higher risk for failure of the STAN system to detect ST interval abnormalities. Øian and Blix (1) are surprised that we did not discuss in more detail the limited evidence regarding the effects on hard endpoints and the lack of long-term outcome data. Our critique of each RCT addresses their first concern (3). Long-term outcome data are obviously important, but they were not the primary aims of the RCTs. Blix and Øian (2) consider metabolic acidosis to be a risk factor for sequelae rather than a hard endpoint reflecting intrapartum management. To support their judgment, they cite a study by Hafstr€ om et al. (13), showing that 89.4% of term newborns with metabolic acidosis did not develop such sequelae. However, they disregard the finding that 10.6% of acidotic newborns exhibited a neurodevelopmental diagnosis or were dead at the age of 6.5 years, compared with 3.3% of these outcomes in the control group (13). The comprehensive systematic review and meta-analysis by Malin et al. (14) confirms that umbilical cord
Table 1. Relative risk with 95% CI for metabolic acidosis, umbilical cord arterial blood pH < 7.05, admission to neonatal intensive care unit and composite adverse neonatal outcome in the groups monitored with fetal ECG-ST analysis in adjunct to CTG compared with the groups monitored with CTG alone. Total IPD meta-analysis RR (95% CI) Metabolic acidosis Cord artery pH < 7.05 Admission NICU Composite adverse outcome
0.76 0.87 0.92 0.80
IPD meta-analysis of second half of RCTs RR (95% CI)
(0.53–1.10) (0.70–1.09) (0.78–1.09) (0.62–1.05)
0.50 0.67 0.76 0.60
(0.29–0.85) (0.49–0.92) (0.59–0.97) (0.40–0.89)
Relative risk (RR) with 95% confidence interval (95% CI) for metabolic acidosis in extracellular fluid, umbilical cord arterial blood pH < 7.05, admission to the neonatal intensive care unit (NICU) and composite adverse neonatal outcome (intrapartum fetal death, neonatal death, Apgar score ≤3 at 5 min, neonatal seizures, cord artery pH ≤ 7.05 and base deficit in extracellular fluid ≥12.0 mmol/L, intubation for ventilation at delivery, and/or neonatal encephalopathy) in the groups monitored with fetal ECG-ST analysis in adjunct to cardiotography (CTG) compared with the groups monitored with CTG alone in randomized controlled trials. Data are from the Individual Participant Data meta-analysis (9) with supplementary information (10). An RR
AOGS LE TT E R TO THE EDIT O R
On the evidence for intrapartum fetal monitoring with ECG-ST analysis
Sir We appreciate the two commentaries by Øian and Blix (1,2) on our reviews of the evidence supporting the electrocardiogram ST interval analysis (STAN) concept as reflected in the randomized controlled trials (RCTs) (3), and the critical appraisal of the published meta-analyses of the STAN system (4). Øian and Blix state that the criteria used for quality assessment of the five RCTs are unclear (1). We respectfully disagree, as our review (3) explicitly identifies each element considered for such assessment: power calculations of outcome variables, pre-trial training, inclusion criteria, randomization and recruitment pace, management protocols during labor, intrapartum interventions, outcomes of metabolic acidosis, and neonatal intensive care admissions. Øian and Blix (1) consider the RCT to be the reference standard for clinical trials, but remark that it is often conducted in a selected population sample studied during highly controlled conditions not typical of routine clinical care. To demonstrate the external validity of ST analysis, observational studies conducted in different clinical settings (5–8) have shown benefits similar to those found in the highest quality RCTs. We agree with the statement that “the STAN technology is not just a machine, it is a concept” that requires education, training and interpretation (1). In Table 1 we have compiled data (9,10) to illustrate the importance of training and experience for the STAN system. It was mainly in the second half of the RCTs that the benefits of STAN monitoring were achieved. The French RCT (11) was excluded from our meta-analysis (4) because the entry criteria violated the STAN clinical guidelines. Blix and Øian (2) argue that “including the French trial [in
the meta-analysis] would provide a better estimate of the actual effectiveness of STAN, as deviations from STAN guidelines are frequent in clinical practice”. However, deviations in the clinical use of a method must be distinguished from deviations in the entry criteria of an RCT. The requirement that STAN monitoring must commence before deterioration of the fetal condition occurs was not unanimously secured in the French RCT. To establish a causal relation between an intervention and an outcome, a nonambiguous temporal precedence must exist, where the intervention occurs before the outcome (12). Although cases with non-reassuring cardiotograms were allowed to enter other RCTs, our reviews (3,4) showed that patients enrolled in the French RCT were at considerably higher risk for failure of the STAN system to detect ST interval abnormalities. Øian and Blix (1) are surprised that we did not discuss in more detail the limited evidence regarding the effects on hard endpoints and the lack of long-term outcome data. Our critique of each RCT addresses their first concern (3). Long-term outcome data are obviously important, but they were not the primary aims of the RCTs. Blix and Øian (2) consider metabolic acidosis to be a risk factor for sequelae rather than a hard endpoint reflecting intrapartum management. To support their judgment, they cite a study by Hafstr€ om et al. (13), showing that 89.4% of term newborns with metabolic acidosis did not develop such sequelae. However, they disregard the finding that 10.6% of acidotic newborns exhibited a neurodevelopmental diagnosis or were dead at the age of 6.5 years, compared with 3.3% of these outcomes in the control group (13). The comprehensive systematic review and meta-analysis by Malin et al. (14) confirms that umbilical cord
Table 1. Relative risk with 95% CI for metabolic acidosis, umbilical cord arterial blood pH < 7.05, admission to neonatal intensive care unit and composite adverse neonatal outcome in the groups monitored with fetal ECG-ST analysis in adjunct to CTG compared with the groups monitored with CTG alone. Total IPD meta-analysis RR (95% CI) Metabolic acidosis Cord artery pH < 7.05 Admission NICU Composite adverse outcome
0.76 0.87 0.92 0.80
IPD meta-analysis of second half of RCTs RR (95% CI)
(0.53–1.10) (0.70–1.09) (0.78–1.09) (0.62–1.05)
0.50 0.67 0.76 0.60
(0.29–0.85) (0.49–0.92) (0.59–0.97) (0.40–0.89)
Relative risk (RR) with 95% confidence interval (95% CI) for metabolic acidosis in extracellular fluid, umbilical cord arterial blood pH < 7.05, admission to the neonatal intensive care unit (NICU) and composite adverse neonatal outcome (intrapartum fetal death, neonatal death, Apgar score ≤3 at 5 min, neonatal seizures, cord artery pH ≤ 7.05 and base deficit in extracellular fluid ≥12.0 mmol/L, intubation for ventilation at delivery, and/or neonatal encephalopathy) in the groups monitored with fetal ECG-ST analysis in adjunct to cardiotography (CTG) compared with the groups monitored with CTG alone in randomized controlled trials. Data are from the Individual Participant Data meta-analysis (9) with supplementary information (10). An RR
