- Email: [email protected]
ST segment analysis of the fetal electrocardiogram plus electronic fetal heart rate monitoring in labor and its relationship to umbilical cord arterial blood gases
American Journal of Obstetrics and Gynecology (2004) 191, 879e84
www.elsevier.com/locate/ajog
ST segment analysis of the fetal electrocardiogram plus electronic fetal heart rate monitoring in labor and its relationship to umbilical cord arterial blood gases Kristina L. Dervaitis, MD,a Monica Poole, MSc,a Gail Schmidt, MA,a Deborah Penava, MD,a Renato Natale, MD,a Robert Gagnon, MD,a,b,* Department of Obstetrics and Gynaecologya and Physiology,b University of Western Ontario, London, Ontario, Canada
KEY WORDS Fetal electrocardiogram Intrapartum surveillance Metabolic acidosis
Objective: This study was undertaken to determine the ability of intrapartum electronic fetal heart rate monitoring (EFM) plus fetal electrocardiogram (ECG) ST segment automated ANalysis (STAN, Neoventa Medical, Go¨teborg, Sweden) monitoring to predict metabolic acidemia (defined as an umbilical cord artery pH ! 7.15 and base deficit R 12 mmol/L) at birth. Study design: Women with singleton, term pregnancies who had a clinical indication for internal EFM with a fetal scalp electrode were included in the study. Attending physicians were blinded to the ST analysis information, only using available EFM as per current clinical practice. After delivery, 2 trained observers blinded to neonatal outcome and ST analysis information performed visual classification of the EFM tracing in 10-minute epochs according to FIGO guidelines. ST events automatically detected by the STAN S21 monitor (Neoventa Medical) were combined with the visual EFM classification as per STAN clinical guidelines (Neoventa Medical). Results: When applying STAN clinical guidelines from a sample of 143 women, our data indicated a sensitivity of 43%, specificity of 74%, negative predictive value of 96%, and a positive predictive value of 8% for metabolic acidemia at birth. Poor ECG quality, despite good EFM tracings (ECG signal loss), occurred 11% of the tracing time. Conclusion: The STAN clinical guidelines have a poor positive predictive value and a sensitivity of less than 50% for metabolic acidemia at birth. Ó 2004 Elsevier Inc. All rights reserved.
This project received financial assistance from the physicians of Ontario through the Physician Services Incorporated Foundation. Presented at the Twenty-Fourth Annual Meeting of the Society for Maternal Fetal Medicine, New Orleans, La, February 2-7, 2004. The STAN S21 monitors were provided by Neoventa Medical, Go¨teborg, Sweden, during the study. * Reprint request: Robert Gagnon, MD, Department of Obstetrics and Gynaecology, St. Joseph’s Health Care Centre, 268 Grosvenor St, London, Ontario, Canada N6A 4L6. E-mail: [email protected] 0002-9378/$ - see front matter Ó 2004 Elsevier Inc. All rights reserved. doi:10.1016/j.ajog.2004.05.059
Visual interpretation of electronic fetal heart rate monitoring (EFM) tracings has been associated with limited interobserver and intraobserver reliability.1 Because the presence of nonreassuring fetal heart rate (FHR) tracings has a low positive predictive value for fetal metabolic acidemia, abnormalities in FHR tracings during labor have led to an increase in the number of unnecessary operative interventions without improvement in perinatal outcomes.2,3 Quantification of fetal electrocardiogram (ECG) ST segment waveform
880 changes that occur during acute and chronic fetal hypoxia by measuring the ratio between the height of the T wave and the QRS complex amplitude (T/QRS ratio) using the STAN S21 monitor (Neoventa Medical, Go¨teborg, Sweden) combined with visual analysis of EFM has been suggested as a useful adjunct to detect fetal metabolic acidosis.4-6 In a retrospective, observational, multicenter study, Amer-Wahlin et al4 found that all 15 cases of intrapartum acidosis, defined as umbilical cord artery pH less than 7.05 and umbilical cord artery base deficit in the extracellular fluid (BDecf) 12 mmol/L or greater were identified by current STAN S21 clinical guidelines (Neoventa Medical).4 Neilson7 further concluded that the use of fetal ST waveform analysis when a decision has been made to undertake EFM may be of benefit, based on 2 randomized clinical trials with the use of the STAN technology (Neoventa Medical) to guide patient management that resulted in a reduction in the incidence of metabolic acidosis at birth and operative deliveries.5,8 Because there is no study in North America that used the STAN technology, we proposed to determine the ability of combined intrapartum EFM plus fetal ECG ST segment automated ANalysis (STAN) monitoring to predict metabolic acidemia by using a prospective observational cohort design.
Patients and methods The study protocol was approved by the University of Western Ontario Research Ethics Board for Health Sciences Research Involving Human Subjects (protocol no. 8667). Recruitment took place from March 2002 to November 2003 in the birthing unit at St. Joseph’s Health Care London, a regional tertiary referral center. Inclusion criteria were term laboring pregnant women (>36 completed weeks) with a singleton, vertex presentation, who had a clinical indication for internal EFM using a fetal scalp electrode, according to the standard clinical practice of the attending obstetrician, independent of study criteria or protocol. After informed consent was obtained, the standard FHR monitor was replaced with the STAN S21 monitor (Neoventa Medical). Once the fetal ECG signal was of good quality as per visual inspection on the screen, all ST-related information on the digital screen was covered so that fetal ECG waveform and ST events information could not be used for clinical management. Any woman with a FHR tracing recording time of less than 20 minutes’ duration, or with an interruption in the FHR tracing for more than 20 minutes immediately before delivery was excluded. FHR was recorded by using a standard spiral scalp electrode (Model 15133E, Philips Medical Systems, Andover, Mass) and a skin plate applied on the inner
Dervaitis et al thigh; monitoring was continued until delivery with the use of the STAN S21 monitor (Neoventa Medical). Changes in the ST segment that STAN automatically detects with the ST Log function that are used within the current STAN clinical guidelines have been described elsewhere.4,5 In addition, the number of minutes of fetal ECG of poor signal quality is automatically recorded and saved in the list of the ST Log events. In this study, only the traditional FHR tracing was visible on the STAN S21 monitor screen and simultaneously recorded at a paper speed of 3 cm/min, and was available to the attending obstetrician for intrapartum fetal health surveillance and management according to their standard clinical practice. Therefore, nursing and obstetric staff were blinded to the fetal ECG and ST Log data, which was recorded and stored automatically for analysis after delivery. All FHR tracings were collected and analyzed after delivery by 2 trained study investigators (K.D. and R.G.), blinded to patient name and neonatal outcome. FHR tracings were reviewed with the STAN viewer software (Neoventa Medical, release R2C) in 10-minute epochs at 3 cm/min without any ST segment analysis information. Each 10-minute epoch was then classified independently by the 2 study investigators (K.D. and R.G.) into normal, intermediary, abnormal, or preterminal according to the STAN clinical guidelines, which are based on the FIGO guidelines for visual EFM interpretation.9 The current STAN S21 fetal monitor does not have internal logic for automated computerized FHR analysis to classify the tracing as normal, intermediary, abnormal, or preterminal. This classification is achieved on the basis of subjective visual analysis of the FHR tracing on the monitor screen. Interobserver agreement for visual EFM interpretation of 10-minute epochs was 92% of the time with the remaining 8% interpreted after agreement by consensus. Once visual EFM classification was completed for each 10-minute epoch and entered into the database, timed fetal ECG ST events automatically analyzed and recorded in the ST Log event were entered into the database by a different study investigator (M.P.), blinded to patient name and neonatal outcome at the time of data entry. The database was designed so that by combining visual FHR analysis classification and ST events, the precise times at which the STAN S21 simplified clinical guidelines dictated intervention could be automatically identified. According to these guidelines, an intervention would have normally included immediate delivery or alleviation of a cause of fetal distress.4 Because it has been suggested that intervention should occur immediately if STAN S21 dictated intervention,9 the first time STAN S21 dictated intervention was recorded and used for grouping into the tracings where STAN S21 clinical guidelines dictated either intervention or no intervention as previously
Dervaitis et al described.4 The overall percentage of signal loss was calculated on the basis of the number of minutes for each 10-minute epoch during which a poor fetal ECG signal was detected and recorded within the automated ST Log event. After delivery, the umbilical cord was doubly clamped and umbilical cord blood artery and vein gases were measured. Blood was taken first from the artery and then from the vein, by using 3-mL preheparinized plastic syringes. Sampling was performed within 10 minutes of birth, with cord blood then placed on ice. Subsequent analysis was achieved within 45 minutes of delivery as previously described.10 An ABL-500 blood gas analyzer (Radiometer, Copenhagen, Denmark) was used throughout the study. The cutoff for metabolic acidemia requiring clinical intervention or ‘‘intervention line’’ according to the Society of Obstetricians and Gynecologists of Canada clinical guidelines and defined as an umbilical artery cord pH less than 7.15 and base deficit 12 mmol/L or greater at birth was used.11 Umbilical cord blood artery BDecf was also calculated by using the Siggaard-Andersen acid base chart algorithm.12 The following variables were calculated: the number of STAN events, the number of times STAN S21 dictated intervention, the overall percentage fetal ECG poor signal and complete umbilical cord blood artery and vein gases. The following variables were also recorded: indication for internal EFM using fetal scalp electrode, maternal age, gestational age, parity, fetal presentation, presence of any antenatal complication, presence of meconium, presence of cord complications, time of birth, birth weight, placental weight, type of labor, type of delivery, and reason for any operative delivery. Secondary outcomes included 1- and 5-minute Apgar scores, need for infant resuscitation, need for neonatal intensive care unit admission. Although calculating sample size for validity testing may not be necessary, our calculation of the sample size was estimated to have a power = 0.80 with an error a = .05 to detect a 3.24-fold difference between a group in which STAN clinical guidelines dictate intervention (n = 42 cases needed) and a group in which STAN clinical guidelines dictate no intervention (n = 102 cases needed). This estimated sample size was based on the following assumptions: an overall incidence of 2.95% of metabolic acidosis observed at St. Joseph’s Health Care London as defined per our primary outcome, an incidence of 26.9% of metabolic acidosis (umbilical cord artery pH ! 7.15) in the study reported by Arukulmaran et al13 in women in which a fetal scalp electrode is applied and ST segment changes are observed, and an incidence of 8.3% of metabolic acidosis in which a fetal scalp electrode is applied as reported by Westgate et al.8 Therefore, an estimated 144 patients was needed in total. Analysis of the primary outcome, being an
881 umbilical artery metabolic acidemia defined as pH less than 7.15 and base deficit 12 mmol/L or greater, was performed with c2 and Fisher exact tests for determination of sensitivity, specificity, and positive and negative predictive values of this diagnostic test with the use of SPSS software version 10.0 (SPSS Inc, Chicago, Ill). Unpaired t test was also used when indicated. A P value of less than .05 was considered statistically significant.
Results A total of 191 women agreed to participate in the study. Fifteen (7.9%) women dropped out from the study because of technical difficulties. Subsequent software updates corrected some of the problems, but we were not able to obtain even a satisfactory FHR tracing with the STAN S21 monitor in 8 of 191 women (4.2%). In all 15 women, although STAN monitoring was not feasible, a satisfactory FHR tracing was obtained with a standard FHR monitor. Our prospective cohort consisted of the remaining 176 women who met study criteria. Patient characteristics and details of their pregnancy and birth are shown in Table I. Umbilical artery and vein cord blood pH, pCO2, base deficit, and BDecf are described in Table II. Thirty-two women were subsequently excluded from the analysis because there was an interruption in the FHR tracing for longer than 20 minutes immediately before delivery. Of these 32, no infant was found to have metabolic acidemia at birth. The reason for the interruption in the FHR tracing was not documented in each case, but possible reasons include: accidental removal of the fetal scalp electrode before delivery, need to move the patient to a different room for operative intervention, delay between time of scalp clip removal, and delivery by cesarean section (Table I). Umbilical cord blood gas data was missing in 1 case. Primary outcome analysis was thus performed on the remaining 143 participants. Although 7 of 143 (4.9%) neonates met the criteria for metabolic acidemia, there was no case of newborn hypoxic-ischemic encephalopathy. STAN clinical guidelines dictated intervention on 1 or more occasions in 38 of 143 patients (27%). The number of neonates with metabolic acidemia at birth was almost equally distributed between those in whom STAN clinical guidelines dictated intervention and those in whom STAN clinical guidelines did not dictate intervention (Table III). Our data indicated a sensitivity of 43%, specificity of 74%, negative predictive value of 96%, and a positive predictive value of 8% of the STAN clinical guidelines for predicting metabolic acidemia at birth (Table III). There was no significant difference in the incidence of metabolic acidemia whether STAN clinical guidelines dictated intervention or not (c2 = 0.32; P = .57). The presence of meconium was not predicted by the STAN clinical guidelines (c2 = 0.19; P = .66).
882
Dervaitis et al
Table I
Patient characteristics
Gestational age at delivery (wk, mean G SD) Birth weight (g, mean G SD) 1-min Apgar score (median [range]) 5-min Apgar score (median [range]) Intrapartum fever (%) Epidural (%) Meconium (%) Induced labor (%) Hypertensive disorder (%) Late FHR decelerations (%) Variable FHR decelerations (%) Fetal bradycardia (%) Fetal tachycardia (%) Decreased FHR variability (%) Oligohydramnios (%) NICU admission (%) Umbilical cord complicationy (%) Shoulder dystocia (%) Forceps (%) Cesarean section (%) Hypoxic ischemic encephalopathy (%)
All patients (n = 176)
Included (n = 144)
Excluded (n = 32)
39.6 G 1.2 3467 G 511 8 (1-10) 9 (3-10) 8.0 93.2 27.3 67.0 17.0 10.8 81.3 29.0 10.8 6.3 4.5 9.1 40.3 2.8 17.6 29.5 0
39.6 G 1.2 3462 G 528 8 (1-10) 9 (3-10) 5.6 93.1 23.6 66.7 15.3 12.5 81.9 30.6 10.4 6.9 4.2 9.0 43.7 2.8 17.4 22.9 0
39.8 G 1.4 3490 G 435 8 (2-9) 9 (5-10) 18.8* 93.8 43.8* 68.8 25.1 3.1 78.1 21.9 12.5 3.1 6.3 9.4 25.0 3.1 18.7 53.1z 0
* P ! .05 vs included group. z P ! .001 vs included group. y Includes 1 or more of: nuchal cord, true knot, cord around body, cord prolapse, velamentous insertion, other.
Table II
Umbilical cord artery and vein pH, pCO2, and base deficit
Umbilical artery pH (mean G SD) PCO2 (mm Hg, mean G SD) Base deficit (mmol/L, mean G SD) Base deficitecf (mmol/L, mean G SD) Umbilical vein pH (mean G SD) PCO2 (mm Hg, mean G SD) Base deficit (mmol/L, mean G SD) Base deficitecf (mmol/L, mean G SD)
All patients (n = 176)
Included (n = 143)
Excluded (n = 32)
7.23 G 0.07 52.6 G 8.9 6.7 G 3.2 4.8 G 2.9
7.23 G 0.07 53.0 G 9.2 6.5 G 3.1 4.5 G 2.8
7.22 G 0.08 50.7 G 7.6 7.7 G 3.9 5.8 G 3.3*
7.30 G 0.06 42.0 G 6.7 5.9 G 2.4 5.0 G 2.1
7.30 G 0.06 42.2 G 6.9 5.7 G 2.2 4.9 G 2.0
7.30 G 0.06 41.1 G 5.4 6.5 G 3.0 5.6 G 2.5
* P = .04 vs included group.
A total of 5435 FHR tracing 10-minute epochs were analyzed for these 143 women of which 60.0% were classified as normal, 18.3% intermediary, 21.2% abnormal, and 0.5% preterminal. Of 143 women, 116 (81.1%) had more than 1 FHR category within the FHR tracing. There was a higher proportion of preterminal FHR epochs in the acidemic group when compared with the nonacidemic group (2.2% vs 0.4%; P ! .001). Poor ECG signal quality occurred 11% of the recording time. When STAN guidelines dictated intervention, the fetal ECG signal loss was 12% in the acidemic group and 10% in the nonacidemic group (P = .37). When STAN guidelines did not dictate intervention, the fetal ECG signal loss was 18% in the acidemic group and 13% in
the nonacidemic group (P = .27). The mean umbilical cord blood arterial pH was 7.07 G 0.06 in the acidemic group compared with 7.24 G 0.05 in the nonacidemic group (P ! .001). Although there was no significant difference in the incidence of low 1-minute Apgar score, a low 5-minute Apgar score of 7 or less was more frequent in the acidemic group compared with the nonacidemic group (c2 = 5.37; P = .02). When compared with the nonacidemic group, there was a significantly higher incidence of meconium (71.4% vs 21.3%, P ! .01) and intrauterine growth restriction (14.3% vs 0.7%, P ! .01) in the acidemic group. None of the other variables of neonatal morbidity reached statistical significance, and all neonates were discharged home in
Dervaitis et al good condition. There was no significant difference in the number of STAN events (P = .54), the number of times STAN dictated intervention (P = .59), and the time interval between the first-time STAN clinical guidelines dictated intervention and the time of birth (P = .55) between the acidotic and the nonacidotic groups of fetuses.
Comment This study demonstrated that with the use of STAN S21 clinical guidelines, the probability of metabolic acidemia at birth as defined as ‘‘intervention line’’ per Canadian guidelines is low (!5%) when intervention is not required. The positive predictive value for metabolic acidemia when STAN clinical guidelines dictated to intervene was relatively low at less than 10% because of a high frequency of times when STAN clinical guidelines indicated intervention whether metabolic acidemia was present or not at birth. The presence of metabolic acidemia was associated with a higher incidence of meconium, intrauterine growth restriction, and mild but statistically significant neonatal depression at birth as demonstrated by a higher incidence of low 5-minute Apgar score when compared with the nonacidemic group. Our results are different from those of Amer-Wahlin et al.4 In their study, the positive predictive value for an umbilical artery cord blood pH less than 7.15, when the STAN S21 clinical guidelines indicated to intervene, was 79%; this represents a marked difference when compared with 8% as in the current study. One of the major differences between the 2 studies is that we blinded attending obstetric staff to ST segment and fetal ECG information. Because it is well established that during visual analysis of FHR tracings, the clinician tends to categorize an abnormal FHR tracing as being normal both with antepartum FHR tracings1 and intrapartum FHR tracings,14 it is possible that the staff obstetrician may not have paid attention to an abnormal FHR tracing resulting in the delivery of an acidemic fetus. The sudden appearance of an ST event on the EFM tracing may also raise the awareness of the obstetrician to a potentially abnormal FHR tracing. It is also possible that because automated ST segment analysis and subjective visual FHR analysis are used concurrently to decide to intervene or not, there might be over- and underreading of the FHR tracing, depending on the obstetrician bias about the fetal condition. As a result, the simultaneous use of both modalities, one objective and automated such as ST segment analysis and the other subjective on the basis of visual analysis of FHR tracing, might have numerous confounding effects on the other that deserve further investigation. Our definition used for metabolic acidemia was different in that we used the standard umbilical artery
883 Table III Occurrence of metabolic acidosis at birth in relation to STAN clinical guidelines
Acidemic NOT Acidemic Total
STAN dictates intervention
STAN does NOT dictate intervention
Total
3 35 38
4 101 105
7 136 143
Sensitivity = 3/7 (43%); specificity = 101/136 (74%); negative predictive value = 101/105 (96%); positive predictive value = 3/38 (8%). c2 = 0.32.
cord blood base deficit instead of BDecf as in the Swedish trial.5,12 However, when recalculating our data to reflect the BDecf, no case would have met the criteria of the more severe metabolic acidemia of an umbilical artery cord blood pH less than 7.05 and BDecf greater than 12 mmol/L as defined in the Swedish trial. As a result we could not have calculated the positive predictive value for acidemia by using this definition and the specificity remained low at 73.4% (data not shown). It is also becoming increasingly evident that the definitions of ‘‘metabolic acidemia’’ have varied among studies.15 For example, there is a significant inverse curvilinear relationship between umbilical artery cord blood pH and base deficit values, and the incidence of adverse neonatal outcomes such as low 5-minute Apgar score, intraventricular hemorrhage, and respiratory difficulties becoming significant at an umbilical arterial cord pH less than 7.20 in preterm neonates.15 Therefore, the cutoff used for umbilical artery cord blood pH that would be beneficial to the neonate if detected effectively during labor may be higher than previously thought. The lack of relationship between ST segment analysis of the fetal ECG plus EFM in labor observed in the current study does not negate the value of the Swedish randomized clinical trial.5 However, our data suggest that although the clinical value of the STAN clinical guidelines has already been demonstrated by reducing the number of neonates with metabolic acidemia at birth5 and neonatal hypoxic-ischemic encephalopathy,9 the performance of this new technology to identify metabolic acidosis is markedly reduced if visual assessment of the FHR tracing is performed independently from the ST segment information and significant ST segment events as performed in the current study. In this blinded analysis, there were unexpected limitations that we have identified, particularly in regard to the quality of the fetal ECG signal; in addition to an overall incidence of poor fetal ECG signal of 11% likely caused by a low amplitude fetal ECG voltage signal in comparison to the maternal ECG signal, we were not able to obtain even a satisfactory FHR tracing in 4.2% of patients. It is possible that more suitable fetal scalp electrodes may be
884 required than the ones currently commercially available in North America to reduce the fetal ECG signal loss. These technical issues need to be addressed if this new technology is to have clinical application. We conclude that in this observational prospective cohort study, the sequential analysis of FHR tracings and associated ST Log events used to identify times when STAN clinical guidelines dictated to intervene or not to intervene has a low positive predictive value and a sensitivity of less than 50% for metabolic acidemia at birth.
Dervaitis et al
7. 8.
9.
10.
References 1. Gagnon R, Campbell MK, Hunse C. A comparison between visual and computer analysis of antepartum fetal heart rate tracings. Am J Obstet Gynecol 1993;168:842-7. 2. Thacker SB, Stroup DF, Peterson HB. Efficacy and safety of intrapartum electronic fetal monitoring: an update. Obstet Gynecol 1995;86:613-20. 3. Thacker SB, Stroup D, Chang M. Continuous electronic heart rate monitoring for fetal assessment during labor. Cochrane Database Syst Rev 2001;CD000063. 4. Amer-Wahlin I, Bordahl P, Eikeland T, Hellsten C, Noren H, Sornes T, et al. ST analysis of the fetal electrocardiogram during labor: nordic observational multicenter study. J Matern Fetal Neonatal Med 2002;12:260-6. 5. Amer-Wahlin I, Hellsten C, Noren H, Hagberg H, Herbst A, Kjellmer I, et al. Cardiotocography only versus cardiotocography plus ST analysis of fetal electrocardiogram for intrapartum fetal monitoring: a Swedish randomised controlled trial. Lancet 2001;358:534-8. 6. Luzietti R, Erkkola R, Hasbargen U, Mattsson LA, Thoulon JM, Rosen KG. European Community Multicenter Trial ‘‘Fetal ECG
11.
12.
13.
14.
15.
analysis during labor’’: ST plus CTG analysis. J Perinat Med 1999;27:431-40. Neilson JP. Fetal electrocardiogram (ECG) for fetal monitoring during labour. Cochrane Database Syst Rev 2003;CD000116. Westgate J, Harris M, Curnow JS, Greene KR. Plymouth randomized trial of cardiotocogram only versus ST waveform plus cardiotocogram for intrapartum monitoring in 2400 cases. Am J Obstet Gynecol 1993;169:1151-60. Noren H, Amer-Wahlin I, Hagberg H, Herbst A, Kjellmer I, Marsal K, et al. Fetal electrocardiography in labor and neonatal outcome: data from the Swedish randomized controlled trial on intrapartum fetal monitoring. Am J Obstet Gynecol 2003;188: 183-92. Richardson B, Nodwell A, Webster K, Alshimmiri M, Gagnon R, Natale R. Fetal oxygen saturation and fractional extraction at birth and the relationship to measures of acidosis. Am J Obstet Gynecol 1998;178:572-9. Liston R, Crane J, Hughes O, Kuling S, MacKinnon C, Milne K, et al. Fetal health surveillance in labour. J Obstet Gynaecol Can 2002;24:342-55. Siggaard-Andersen O. An acid-base chart for arterial blood with normal and pathophysiological reference areas. Scand J Clin Lab Invest 1971;27:239-45. Arulkumaran S, Lilja H, Lindecrantz K, Ratnam SS, Thavarasah AS, Rosen KG. Fetal ECG waveform analysis should improve fetal surveillance in labour. J Perinat Med 1990;18:13-22. Devoe L, Golde S, Kilman Y, Morton D, Shea K, Waller J. A comparison of visual analyses of intrapartum fetal heart rate tracings according to the new national institute of child health and human development guidelines with computer analyses by an automated fetal heart rate monitoring system. Am J Obstet Gynecol 2000;183:361-6. Victory R, Penava D, da Silva O, Natale R, Richardson B. Umbilical cord pH and base excess values in relation to neonatal morbidity for infants delivered preterm. Am J Obstet Gynecol 2003;189:803-7.
www.elsevier.com/locate/ajog
ST segment analysis of the fetal electrocardiogram plus electronic fetal heart rate monitoring in labor and its relationship to umbilical cord arterial blood gases Kristina L. Dervaitis, MD,a Monica Poole, MSc,a Gail Schmidt, MA,a Deborah Penava, MD,a Renato Natale, MD,a Robert Gagnon, MD,a,b,* Department of Obstetrics and Gynaecologya and Physiology,b University of Western Ontario, London, Ontario, Canada
KEY WORDS Fetal electrocardiogram Intrapartum surveillance Metabolic acidosis
Objective: This study was undertaken to determine the ability of intrapartum electronic fetal heart rate monitoring (EFM) plus fetal electrocardiogram (ECG) ST segment automated ANalysis (STAN, Neoventa Medical, Go¨teborg, Sweden) monitoring to predict metabolic acidemia (defined as an umbilical cord artery pH ! 7.15 and base deficit R 12 mmol/L) at birth. Study design: Women with singleton, term pregnancies who had a clinical indication for internal EFM with a fetal scalp electrode were included in the study. Attending physicians were blinded to the ST analysis information, only using available EFM as per current clinical practice. After delivery, 2 trained observers blinded to neonatal outcome and ST analysis information performed visual classification of the EFM tracing in 10-minute epochs according to FIGO guidelines. ST events automatically detected by the STAN S21 monitor (Neoventa Medical) were combined with the visual EFM classification as per STAN clinical guidelines (Neoventa Medical). Results: When applying STAN clinical guidelines from a sample of 143 women, our data indicated a sensitivity of 43%, specificity of 74%, negative predictive value of 96%, and a positive predictive value of 8% for metabolic acidemia at birth. Poor ECG quality, despite good EFM tracings (ECG signal loss), occurred 11% of the tracing time. Conclusion: The STAN clinical guidelines have a poor positive predictive value and a sensitivity of less than 50% for metabolic acidemia at birth. Ó 2004 Elsevier Inc. All rights reserved.
This project received financial assistance from the physicians of Ontario through the Physician Services Incorporated Foundation. Presented at the Twenty-Fourth Annual Meeting of the Society for Maternal Fetal Medicine, New Orleans, La, February 2-7, 2004. The STAN S21 monitors were provided by Neoventa Medical, Go¨teborg, Sweden, during the study. * Reprint request: Robert Gagnon, MD, Department of Obstetrics and Gynaecology, St. Joseph’s Health Care Centre, 268 Grosvenor St, London, Ontario, Canada N6A 4L6. E-mail: [email protected] 0002-9378/$ - see front matter Ó 2004 Elsevier Inc. All rights reserved. doi:10.1016/j.ajog.2004.05.059
Visual interpretation of electronic fetal heart rate monitoring (EFM) tracings has been associated with limited interobserver and intraobserver reliability.1 Because the presence of nonreassuring fetal heart rate (FHR) tracings has a low positive predictive value for fetal metabolic acidemia, abnormalities in FHR tracings during labor have led to an increase in the number of unnecessary operative interventions without improvement in perinatal outcomes.2,3 Quantification of fetal electrocardiogram (ECG) ST segment waveform
880 changes that occur during acute and chronic fetal hypoxia by measuring the ratio between the height of the T wave and the QRS complex amplitude (T/QRS ratio) using the STAN S21 monitor (Neoventa Medical, Go¨teborg, Sweden) combined with visual analysis of EFM has been suggested as a useful adjunct to detect fetal metabolic acidosis.4-6 In a retrospective, observational, multicenter study, Amer-Wahlin et al4 found that all 15 cases of intrapartum acidosis, defined as umbilical cord artery pH less than 7.05 and umbilical cord artery base deficit in the extracellular fluid (BDecf) 12 mmol/L or greater were identified by current STAN S21 clinical guidelines (Neoventa Medical).4 Neilson7 further concluded that the use of fetal ST waveform analysis when a decision has been made to undertake EFM may be of benefit, based on 2 randomized clinical trials with the use of the STAN technology (Neoventa Medical) to guide patient management that resulted in a reduction in the incidence of metabolic acidosis at birth and operative deliveries.5,8 Because there is no study in North America that used the STAN technology, we proposed to determine the ability of combined intrapartum EFM plus fetal ECG ST segment automated ANalysis (STAN) monitoring to predict metabolic acidemia by using a prospective observational cohort design.
Patients and methods The study protocol was approved by the University of Western Ontario Research Ethics Board for Health Sciences Research Involving Human Subjects (protocol no. 8667). Recruitment took place from March 2002 to November 2003 in the birthing unit at St. Joseph’s Health Care London, a regional tertiary referral center. Inclusion criteria were term laboring pregnant women (>36 completed weeks) with a singleton, vertex presentation, who had a clinical indication for internal EFM using a fetal scalp electrode, according to the standard clinical practice of the attending obstetrician, independent of study criteria or protocol. After informed consent was obtained, the standard FHR monitor was replaced with the STAN S21 monitor (Neoventa Medical). Once the fetal ECG signal was of good quality as per visual inspection on the screen, all ST-related information on the digital screen was covered so that fetal ECG waveform and ST events information could not be used for clinical management. Any woman with a FHR tracing recording time of less than 20 minutes’ duration, or with an interruption in the FHR tracing for more than 20 minutes immediately before delivery was excluded. FHR was recorded by using a standard spiral scalp electrode (Model 15133E, Philips Medical Systems, Andover, Mass) and a skin plate applied on the inner
Dervaitis et al thigh; monitoring was continued until delivery with the use of the STAN S21 monitor (Neoventa Medical). Changes in the ST segment that STAN automatically detects with the ST Log function that are used within the current STAN clinical guidelines have been described elsewhere.4,5 In addition, the number of minutes of fetal ECG of poor signal quality is automatically recorded and saved in the list of the ST Log events. In this study, only the traditional FHR tracing was visible on the STAN S21 monitor screen and simultaneously recorded at a paper speed of 3 cm/min, and was available to the attending obstetrician for intrapartum fetal health surveillance and management according to their standard clinical practice. Therefore, nursing and obstetric staff were blinded to the fetal ECG and ST Log data, which was recorded and stored automatically for analysis after delivery. All FHR tracings were collected and analyzed after delivery by 2 trained study investigators (K.D. and R.G.), blinded to patient name and neonatal outcome. FHR tracings were reviewed with the STAN viewer software (Neoventa Medical, release R2C) in 10-minute epochs at 3 cm/min without any ST segment analysis information. Each 10-minute epoch was then classified independently by the 2 study investigators (K.D. and R.G.) into normal, intermediary, abnormal, or preterminal according to the STAN clinical guidelines, which are based on the FIGO guidelines for visual EFM interpretation.9 The current STAN S21 fetal monitor does not have internal logic for automated computerized FHR analysis to classify the tracing as normal, intermediary, abnormal, or preterminal. This classification is achieved on the basis of subjective visual analysis of the FHR tracing on the monitor screen. Interobserver agreement for visual EFM interpretation of 10-minute epochs was 92% of the time with the remaining 8% interpreted after agreement by consensus. Once visual EFM classification was completed for each 10-minute epoch and entered into the database, timed fetal ECG ST events automatically analyzed and recorded in the ST Log event were entered into the database by a different study investigator (M.P.), blinded to patient name and neonatal outcome at the time of data entry. The database was designed so that by combining visual FHR analysis classification and ST events, the precise times at which the STAN S21 simplified clinical guidelines dictated intervention could be automatically identified. According to these guidelines, an intervention would have normally included immediate delivery or alleviation of a cause of fetal distress.4 Because it has been suggested that intervention should occur immediately if STAN S21 dictated intervention,9 the first time STAN S21 dictated intervention was recorded and used for grouping into the tracings where STAN S21 clinical guidelines dictated either intervention or no intervention as previously
Dervaitis et al described.4 The overall percentage of signal loss was calculated on the basis of the number of minutes for each 10-minute epoch during which a poor fetal ECG signal was detected and recorded within the automated ST Log event. After delivery, the umbilical cord was doubly clamped and umbilical cord blood artery and vein gases were measured. Blood was taken first from the artery and then from the vein, by using 3-mL preheparinized plastic syringes. Sampling was performed within 10 minutes of birth, with cord blood then placed on ice. Subsequent analysis was achieved within 45 minutes of delivery as previously described.10 An ABL-500 blood gas analyzer (Radiometer, Copenhagen, Denmark) was used throughout the study. The cutoff for metabolic acidemia requiring clinical intervention or ‘‘intervention line’’ according to the Society of Obstetricians and Gynecologists of Canada clinical guidelines and defined as an umbilical artery cord pH less than 7.15 and base deficit 12 mmol/L or greater at birth was used.11 Umbilical cord blood artery BDecf was also calculated by using the Siggaard-Andersen acid base chart algorithm.12 The following variables were calculated: the number of STAN events, the number of times STAN S21 dictated intervention, the overall percentage fetal ECG poor signal and complete umbilical cord blood artery and vein gases. The following variables were also recorded: indication for internal EFM using fetal scalp electrode, maternal age, gestational age, parity, fetal presentation, presence of any antenatal complication, presence of meconium, presence of cord complications, time of birth, birth weight, placental weight, type of labor, type of delivery, and reason for any operative delivery. Secondary outcomes included 1- and 5-minute Apgar scores, need for infant resuscitation, need for neonatal intensive care unit admission. Although calculating sample size for validity testing may not be necessary, our calculation of the sample size was estimated to have a power = 0.80 with an error a = .05 to detect a 3.24-fold difference between a group in which STAN clinical guidelines dictate intervention (n = 42 cases needed) and a group in which STAN clinical guidelines dictate no intervention (n = 102 cases needed). This estimated sample size was based on the following assumptions: an overall incidence of 2.95% of metabolic acidosis observed at St. Joseph’s Health Care London as defined per our primary outcome, an incidence of 26.9% of metabolic acidosis (umbilical cord artery pH ! 7.15) in the study reported by Arukulmaran et al13 in women in which a fetal scalp electrode is applied and ST segment changes are observed, and an incidence of 8.3% of metabolic acidosis in which a fetal scalp electrode is applied as reported by Westgate et al.8 Therefore, an estimated 144 patients was needed in total. Analysis of the primary outcome, being an
881 umbilical artery metabolic acidemia defined as pH less than 7.15 and base deficit 12 mmol/L or greater, was performed with c2 and Fisher exact tests for determination of sensitivity, specificity, and positive and negative predictive values of this diagnostic test with the use of SPSS software version 10.0 (SPSS Inc, Chicago, Ill). Unpaired t test was also used when indicated. A P value of less than .05 was considered statistically significant.
Results A total of 191 women agreed to participate in the study. Fifteen (7.9%) women dropped out from the study because of technical difficulties. Subsequent software updates corrected some of the problems, but we were not able to obtain even a satisfactory FHR tracing with the STAN S21 monitor in 8 of 191 women (4.2%). In all 15 women, although STAN monitoring was not feasible, a satisfactory FHR tracing was obtained with a standard FHR monitor. Our prospective cohort consisted of the remaining 176 women who met study criteria. Patient characteristics and details of their pregnancy and birth are shown in Table I. Umbilical artery and vein cord blood pH, pCO2, base deficit, and BDecf are described in Table II. Thirty-two women were subsequently excluded from the analysis because there was an interruption in the FHR tracing for longer than 20 minutes immediately before delivery. Of these 32, no infant was found to have metabolic acidemia at birth. The reason for the interruption in the FHR tracing was not documented in each case, but possible reasons include: accidental removal of the fetal scalp electrode before delivery, need to move the patient to a different room for operative intervention, delay between time of scalp clip removal, and delivery by cesarean section (Table I). Umbilical cord blood gas data was missing in 1 case. Primary outcome analysis was thus performed on the remaining 143 participants. Although 7 of 143 (4.9%) neonates met the criteria for metabolic acidemia, there was no case of newborn hypoxic-ischemic encephalopathy. STAN clinical guidelines dictated intervention on 1 or more occasions in 38 of 143 patients (27%). The number of neonates with metabolic acidemia at birth was almost equally distributed between those in whom STAN clinical guidelines dictated intervention and those in whom STAN clinical guidelines did not dictate intervention (Table III). Our data indicated a sensitivity of 43%, specificity of 74%, negative predictive value of 96%, and a positive predictive value of 8% of the STAN clinical guidelines for predicting metabolic acidemia at birth (Table III). There was no significant difference in the incidence of metabolic acidemia whether STAN clinical guidelines dictated intervention or not (c2 = 0.32; P = .57). The presence of meconium was not predicted by the STAN clinical guidelines (c2 = 0.19; P = .66).
882
Dervaitis et al
Table I
Patient characteristics
Gestational age at delivery (wk, mean G SD) Birth weight (g, mean G SD) 1-min Apgar score (median [range]) 5-min Apgar score (median [range]) Intrapartum fever (%) Epidural (%) Meconium (%) Induced labor (%) Hypertensive disorder (%) Late FHR decelerations (%) Variable FHR decelerations (%) Fetal bradycardia (%) Fetal tachycardia (%) Decreased FHR variability (%) Oligohydramnios (%) NICU admission (%) Umbilical cord complicationy (%) Shoulder dystocia (%) Forceps (%) Cesarean section (%) Hypoxic ischemic encephalopathy (%)
All patients (n = 176)
Included (n = 144)
Excluded (n = 32)
39.6 G 1.2 3467 G 511 8 (1-10) 9 (3-10) 8.0 93.2 27.3 67.0 17.0 10.8 81.3 29.0 10.8 6.3 4.5 9.1 40.3 2.8 17.6 29.5 0
39.6 G 1.2 3462 G 528 8 (1-10) 9 (3-10) 5.6 93.1 23.6 66.7 15.3 12.5 81.9 30.6 10.4 6.9 4.2 9.0 43.7 2.8 17.4 22.9 0
39.8 G 1.4 3490 G 435 8 (2-9) 9 (5-10) 18.8* 93.8 43.8* 68.8 25.1 3.1 78.1 21.9 12.5 3.1 6.3 9.4 25.0 3.1 18.7 53.1z 0
* P ! .05 vs included group. z P ! .001 vs included group. y Includes 1 or more of: nuchal cord, true knot, cord around body, cord prolapse, velamentous insertion, other.
Table II
Umbilical cord artery and vein pH, pCO2, and base deficit
Umbilical artery pH (mean G SD) PCO2 (mm Hg, mean G SD) Base deficit (mmol/L, mean G SD) Base deficitecf (mmol/L, mean G SD) Umbilical vein pH (mean G SD) PCO2 (mm Hg, mean G SD) Base deficit (mmol/L, mean G SD) Base deficitecf (mmol/L, mean G SD)
All patients (n = 176)
Included (n = 143)
Excluded (n = 32)
7.23 G 0.07 52.6 G 8.9 6.7 G 3.2 4.8 G 2.9
7.23 G 0.07 53.0 G 9.2 6.5 G 3.1 4.5 G 2.8
7.22 G 0.08 50.7 G 7.6 7.7 G 3.9 5.8 G 3.3*
7.30 G 0.06 42.0 G 6.7 5.9 G 2.4 5.0 G 2.1
7.30 G 0.06 42.2 G 6.9 5.7 G 2.2 4.9 G 2.0
7.30 G 0.06 41.1 G 5.4 6.5 G 3.0 5.6 G 2.5
* P = .04 vs included group.
A total of 5435 FHR tracing 10-minute epochs were analyzed for these 143 women of which 60.0% were classified as normal, 18.3% intermediary, 21.2% abnormal, and 0.5% preterminal. Of 143 women, 116 (81.1%) had more than 1 FHR category within the FHR tracing. There was a higher proportion of preterminal FHR epochs in the acidemic group when compared with the nonacidemic group (2.2% vs 0.4%; P ! .001). Poor ECG signal quality occurred 11% of the recording time. When STAN guidelines dictated intervention, the fetal ECG signal loss was 12% in the acidemic group and 10% in the nonacidemic group (P = .37). When STAN guidelines did not dictate intervention, the fetal ECG signal loss was 18% in the acidemic group and 13% in
the nonacidemic group (P = .27). The mean umbilical cord blood arterial pH was 7.07 G 0.06 in the acidemic group compared with 7.24 G 0.05 in the nonacidemic group (P ! .001). Although there was no significant difference in the incidence of low 1-minute Apgar score, a low 5-minute Apgar score of 7 or less was more frequent in the acidemic group compared with the nonacidemic group (c2 = 5.37; P = .02). When compared with the nonacidemic group, there was a significantly higher incidence of meconium (71.4% vs 21.3%, P ! .01) and intrauterine growth restriction (14.3% vs 0.7%, P ! .01) in the acidemic group. None of the other variables of neonatal morbidity reached statistical significance, and all neonates were discharged home in
Dervaitis et al good condition. There was no significant difference in the number of STAN events (P = .54), the number of times STAN dictated intervention (P = .59), and the time interval between the first-time STAN clinical guidelines dictated intervention and the time of birth (P = .55) between the acidotic and the nonacidotic groups of fetuses.
Comment This study demonstrated that with the use of STAN S21 clinical guidelines, the probability of metabolic acidemia at birth as defined as ‘‘intervention line’’ per Canadian guidelines is low (!5%) when intervention is not required. The positive predictive value for metabolic acidemia when STAN clinical guidelines dictated to intervene was relatively low at less than 10% because of a high frequency of times when STAN clinical guidelines indicated intervention whether metabolic acidemia was present or not at birth. The presence of metabolic acidemia was associated with a higher incidence of meconium, intrauterine growth restriction, and mild but statistically significant neonatal depression at birth as demonstrated by a higher incidence of low 5-minute Apgar score when compared with the nonacidemic group. Our results are different from those of Amer-Wahlin et al.4 In their study, the positive predictive value for an umbilical artery cord blood pH less than 7.15, when the STAN S21 clinical guidelines indicated to intervene, was 79%; this represents a marked difference when compared with 8% as in the current study. One of the major differences between the 2 studies is that we blinded attending obstetric staff to ST segment and fetal ECG information. Because it is well established that during visual analysis of FHR tracings, the clinician tends to categorize an abnormal FHR tracing as being normal both with antepartum FHR tracings1 and intrapartum FHR tracings,14 it is possible that the staff obstetrician may not have paid attention to an abnormal FHR tracing resulting in the delivery of an acidemic fetus. The sudden appearance of an ST event on the EFM tracing may also raise the awareness of the obstetrician to a potentially abnormal FHR tracing. It is also possible that because automated ST segment analysis and subjective visual FHR analysis are used concurrently to decide to intervene or not, there might be over- and underreading of the FHR tracing, depending on the obstetrician bias about the fetal condition. As a result, the simultaneous use of both modalities, one objective and automated such as ST segment analysis and the other subjective on the basis of visual analysis of FHR tracing, might have numerous confounding effects on the other that deserve further investigation. Our definition used for metabolic acidemia was different in that we used the standard umbilical artery
883 Table III Occurrence of metabolic acidosis at birth in relation to STAN clinical guidelines
Acidemic NOT Acidemic Total
STAN dictates intervention
STAN does NOT dictate intervention
Total
3 35 38
4 101 105
7 136 143
Sensitivity = 3/7 (43%); specificity = 101/136 (74%); negative predictive value = 101/105 (96%); positive predictive value = 3/38 (8%). c2 = 0.32.
cord blood base deficit instead of BDecf as in the Swedish trial.5,12 However, when recalculating our data to reflect the BDecf, no case would have met the criteria of the more severe metabolic acidemia of an umbilical artery cord blood pH less than 7.05 and BDecf greater than 12 mmol/L as defined in the Swedish trial. As a result we could not have calculated the positive predictive value for acidemia by using this definition and the specificity remained low at 73.4% (data not shown). It is also becoming increasingly evident that the definitions of ‘‘metabolic acidemia’’ have varied among studies.15 For example, there is a significant inverse curvilinear relationship between umbilical artery cord blood pH and base deficit values, and the incidence of adverse neonatal outcomes such as low 5-minute Apgar score, intraventricular hemorrhage, and respiratory difficulties becoming significant at an umbilical arterial cord pH less than 7.20 in preterm neonates.15 Therefore, the cutoff used for umbilical artery cord blood pH that would be beneficial to the neonate if detected effectively during labor may be higher than previously thought. The lack of relationship between ST segment analysis of the fetal ECG plus EFM in labor observed in the current study does not negate the value of the Swedish randomized clinical trial.5 However, our data suggest that although the clinical value of the STAN clinical guidelines has already been demonstrated by reducing the number of neonates with metabolic acidemia at birth5 and neonatal hypoxic-ischemic encephalopathy,9 the performance of this new technology to identify metabolic acidosis is markedly reduced if visual assessment of the FHR tracing is performed independently from the ST segment information and significant ST segment events as performed in the current study. In this blinded analysis, there were unexpected limitations that we have identified, particularly in regard to the quality of the fetal ECG signal; in addition to an overall incidence of poor fetal ECG signal of 11% likely caused by a low amplitude fetal ECG voltage signal in comparison to the maternal ECG signal, we were not able to obtain even a satisfactory FHR tracing in 4.2% of patients. It is possible that more suitable fetal scalp electrodes may be
884 required than the ones currently commercially available in North America to reduce the fetal ECG signal loss. These technical issues need to be addressed if this new technology is to have clinical application. We conclude that in this observational prospective cohort study, the sequential analysis of FHR tracings and associated ST Log events used to identify times when STAN clinical guidelines dictated to intervene or not to intervene has a low positive predictive value and a sensitivity of less than 50% for metabolic acidemia at birth.
Dervaitis et al
7. 8.
9.
10.
References 1. Gagnon R, Campbell MK, Hunse C. A comparison between visual and computer analysis of antepartum fetal heart rate tracings. Am J Obstet Gynecol 1993;168:842-7. 2. Thacker SB, Stroup DF, Peterson HB. Efficacy and safety of intrapartum electronic fetal monitoring: an update. Obstet Gynecol 1995;86:613-20. 3. Thacker SB, Stroup D, Chang M. Continuous electronic heart rate monitoring for fetal assessment during labor. Cochrane Database Syst Rev 2001;CD000063. 4. Amer-Wahlin I, Bordahl P, Eikeland T, Hellsten C, Noren H, Sornes T, et al. ST analysis of the fetal electrocardiogram during labor: nordic observational multicenter study. J Matern Fetal Neonatal Med 2002;12:260-6. 5. Amer-Wahlin I, Hellsten C, Noren H, Hagberg H, Herbst A, Kjellmer I, et al. Cardiotocography only versus cardiotocography plus ST analysis of fetal electrocardiogram for intrapartum fetal monitoring: a Swedish randomised controlled trial. Lancet 2001;358:534-8. 6. Luzietti R, Erkkola R, Hasbargen U, Mattsson LA, Thoulon JM, Rosen KG. European Community Multicenter Trial ‘‘Fetal ECG
11.
12.
13.
14.
15.
analysis during labor’’: ST plus CTG analysis. J Perinat Med 1999;27:431-40. Neilson JP. Fetal electrocardiogram (ECG) for fetal monitoring during labour. Cochrane Database Syst Rev 2003;CD000116. Westgate J, Harris M, Curnow JS, Greene KR. Plymouth randomized trial of cardiotocogram only versus ST waveform plus cardiotocogram for intrapartum monitoring in 2400 cases. Am J Obstet Gynecol 1993;169:1151-60. Noren H, Amer-Wahlin I, Hagberg H, Herbst A, Kjellmer I, Marsal K, et al. Fetal electrocardiography in labor and neonatal outcome: data from the Swedish randomized controlled trial on intrapartum fetal monitoring. Am J Obstet Gynecol 2003;188: 183-92. Richardson B, Nodwell A, Webster K, Alshimmiri M, Gagnon R, Natale R. Fetal oxygen saturation and fractional extraction at birth and the relationship to measures of acidosis. Am J Obstet Gynecol 1998;178:572-9. Liston R, Crane J, Hughes O, Kuling S, MacKinnon C, Milne K, et al. Fetal health surveillance in labour. J Obstet Gynaecol Can 2002;24:342-55. Siggaard-Andersen O. An acid-base chart for arterial blood with normal and pathophysiological reference areas. Scand J Clin Lab Invest 1971;27:239-45. Arulkumaran S, Lilja H, Lindecrantz K, Ratnam SS, Thavarasah AS, Rosen KG. Fetal ECG waveform analysis should improve fetal surveillance in labour. J Perinat Med 1990;18:13-22. Devoe L, Golde S, Kilman Y, Morton D, Shea K, Waller J. A comparison of visual analyses of intrapartum fetal heart rate tracings according to the new national institute of child health and human development guidelines with computer analyses by an automated fetal heart rate monitoring system. Am J Obstet Gynecol 2000;183:361-6. Victory R, Penava D, da Silva O, Natale R, Richardson B. Umbilical cord pH and base excess values in relation to neonatal morbidity for infants delivered preterm. Am J Obstet Gynecol 2003;189:803-7.