Dr CE Pennell, School of Women’s and Infants’ Health M550, The University of Western Australia, 35 Stirling Highway, Crawley, WA 6009, Australia. Email email@example.com
Please cite this paper as: White C, Doherty D, Kohan R, Newnham J, Pennell C. Evaluation of selection criteria for validating paired umbilical cord blood gas samples: an observational study. BJOG 2012;119:857–865.
Objective To compare six validation criteria for umbilical cord blood gas (UCBG) values in vigorous and nonvigorous neonates.
Design Retrospective cohort study.
Setting Single tertiary obstetric centre, King Edward Memorial Hospital (KEMH), Perth, Western Australia.
Sample A total of 37 763 consecutive deliveries at >23 weeks of gestation.
Methods Six validation criteria were compared to evaluate the proportion of deliveries with ‘valid’ UCBG data; and the proportion of vigorous and nonvigorous neonates with metabolic acidaemia.
Main outcomes Proportion of deliveries with ‘valid’ UCBG values; proportions of vigorous and nonvigorous neonates with normal, borderline and abnormal UCBG values.
Results The criteria based on KEMH 5th centile arteriovenous pH and Pco2 differences resulted in a higher proportion of neonates with ‘valid’ UCBG values than the previously described Westgate and Kro criteria. The increase in ‘valid’ UCBG values occurred across the entire study population including vigorous and nonvigorous neonates. Among neonates with short-term neonatal complications there was an increase in nonvigorous neonates with umbilical artery metabolic acidaemia. There was no corresponding increase in vigorous neonates diagnosed with abnormal UCBG values.
Conclusions Use of the KEMH criteria results in an increase in the proportion of nonvigorous term neonates with UCBG data considered ‘valid’ to aid clinicians in the management of the neonate shortly after delivery. This change occurs without increasing the rate of false-positive diagnoses of acidaemia in vigorous neonates. The KEMH ‘validation’ criteria were developed from an entire presenting population and provide a simple algorithm that can be universally applied to identify neonates with nonphysiological UCBG values.
There is a growing body of support for umbilical cord blood gas (UCBG) analysis to be undertaken on all deliveries to provide an objective marker of neonatal biochemical status at delivery1–6 that in turn can identify neonates that have been exposed to perinatal asphyxial insults. Although some authors have advocated sampling only the umbilical artery,7 the majority of authors and professional societies strongly advocate obtaining and analysing samples from both the umbilical artery and vein8–12 to ensure the biological validity of the blood gas values obtained.
Most previous studies using UCBG data have included all incidents where blood gas data were available. This approach does not take into account pre-analysis sampling errors that may result in the analysed samples not being a valid representation of paired umbilical arterial and venous blood. Pre-analysis sampling errors are primarily the result of inadvertent collection of mixed arterial and venous blood: less commonly, the same vessel can be sampled twice (usually the umbilical vein) or the samples may be mislabelled. To ensure the validity of paired UCBG values, two methods have been developed: the Westgate et al.11 criteria and the more recent Kro et al.13 criteria. The Westgate et al. criteria are constrained by the small sample size and the subpopulation from which they were developed. In contrast, the Kro et al. criteria were developed in a larger and more diverse cohort than the Westgate et al. criteria; however, some of the elements used in the Kro et al. criteria are complex and may be difficult to apply in clinical situations.
The ‘validation’ of accurate paired UCBG samples is unlikely to have clinically significant consequences for neonates delivered in good condition with uneventful neonatal periods. In contrast, for neonates born in poor clinical condition or those who have neonatal complications, the classification of paired UCBG samples as ‘valid’ (physiologically plausible) or ‘nonvalid’ (physiologically implausible) may have clinically significant consequences. This is of particular importance when considering the central role of UCBG values in the diagnosis of perinatal asphyxial injury10,14 and its resulting medical, social, financial and legal implications. Consequently, the development and evaluation of selection criteria for UCBG values must take into account not just the overall proportion of neonates with valid and nonvalid UCBG values but also, and perhaps more importantly, the proportions among neonates with adverse neonatal and clinical outcomes.
The aim of this study was to compare four new selection criteria with the Westgate et al. and Kro et al. criteria for identifying accurate paired UCBG samples in vigorous and nonvigorous neonates at delivery. In addition, this study includes the first evaluation of the effects of different selection criteria on the number and proportion of neonates with UCBG values available for interpretation among those that experience adverse clinical and neonatal outcomes.
The study was undertaken over a 7-year period (1 January 2003 to 31 December 2009) at King Edward Memorial Hospital (KEMH), Perth, Western Australia, the sole tertiary-level maternity unit for Western Australia. Paired arterial and venous UCBG and lactate samples were collected for all neonates delivered at ≥23 weeks of gestation apart from: (i) therapeutic abortions for fetal anomalies; and (ii) fetal deaths in utero diagnosed before the onset of labour. Detailed maternal, obstetric, intrapartum and neonatal information for all deliveries were extracted from the institutional electronic database.
As soon as possible after delivery, with the placenta in situ and ideally before the neonate’s first breath, an umbilical cord segment was isolated using cord clamps. Arterial and venous samples were collected using a 1-ml preheparinised plastic syringe (Rapidlyte™; Bayer Corporation, East Walpole, MA, USA) and a 21-gauge needle. Analysis was performed as soon as possible after collection, typically within 5–10 minutes after collection but occasionally extending up to 30 minutes.
Blood gas analysis was performed on paired arterial and venous samples using a Bayer Chiron Diagnostics Model 840 Blood Gas Analyser (Bayer Diagnostics, Leverkusen, Germany) located in the labour and delivery ward. Umbilical lactate was measured on the cord arterial sample using a Roche Accusport™ (Boehringer Mannheim, Mannheim, Germany). For the final 2 years of the study, a lactate electrode was installed in the blood gas analyser; however, the hand-held lactate meter was retained as a back-up mode of analysis. The blood gas analyser was serviced daily by hospital technical staff members and the lactate meters were calibrated with every 20 samples and serviced when necessary. Automatic one-point and two-point calibration of the blood gas analyser occurred every 30 and 120 minutes, respectively. The midwifery staff, usually those who collected the samples, undertook all analyses. All staff members were required to meet minimum competency standards in sampling and measurement techniques before performing UCBG sampling and analyses independently.
The six selection criteria for determining validated (biologically plausible) UCBG values were evaluated in this study (Table 1). The first two selection criteria are the original Westgate et al.11 criteria and Kro et al.13 criteria. The third is a modified version of the Westgate et al. criteria which uses the KEMH population fifth centile for arteriovenous (AV) pH difference (≥0.01 pH units) and the KEMH population tenth centile for AV Pco2 difference (≥3.4 mmHg) rather than the fifth and tenth centiles described in the original Plymouth study population. The KEMH criteria are similar to the Westgate and modified Westgate criteria: the KEMH fifth centile AV differences in pH (≥0.01 pH units) and Pco2 (≥0.2 mmHg) are used to confirm two-vessel sampling. The minimum information required for clinical interpretation of fetal acid–base status at delivery is a pH value, a Pco2 value and an indicator of metabolic acidaemia (either a base excess or a lactate value). These three measures are used to define the ‘minimal criteria’. The sixth and final criterion evaluated is based on having at least one cord-blood vessel lactate measure available at delivery: the ‘lactate criteria’.
Table 1. Definitions of various cord-blood gas validation criteria
KEMH, King Edward Memorial Hospital; ΔUAV, umbilical arteriovenous difference.
*Two standard deviations below the intrapartum mean maternal arterial Pco2.30,31
Arterial or venous base excess or lactate value present
Arterial or venous lactate value present
The definition of neonatal umbilical acidaemia is based on the International Cerebral Palsy Task Force, American College of Obstetricians and Gynaecologists and the American Academy of Paediatrics guidelines: an umbilical artery pH value <7.00 and an umbilical artery base excess value <−12 mmol/l.10,14 A nonvigorous neonate at delivery was defined as a neonate with a 5-minute Apgar score <7.
Descriptive data summaries were generated using frequency distributions (number and percentage). Comparisons of categorical outcomes were conducted using chi-square tests, whereas a continuous outcome comparison was conducted by one-way analysis of variance. spss for Windows statistical software was used for all data analysis (Version 15.0; SPSS Inc., Chicago, IL, USA). All statistical hypothesis tests were two-sided and P values < 0.05 were considered statistically significant.
During the 7-year study period, 38 404 neonates were delivered at 23 weeks of gestation or greater at KEMH, Perth, Western Australia. Six hundred and forty-one neonates were excluded because of therapeutic terminations for fetal anomaly or fetal deaths in utero diagnosed before the onset of labour. Detailed maternal, obstetric, intrapartum and neonatal information was available for all deliveries (Table 2). Across the study period there was an increase in maternal age (P < 0.001) as well as variation in maternal parity (P < 0.001). There was also an increase in operative deliveries with a corresponding decline in vaginal deliveries (P < 0.001). No significant change in plurality, birth weight or gestational age occurred over the study period (all P > 0.05).
Table 2. Maternal, obstetric, intrapartum and neonatal characteristics for KEMH cohort (n = 37 763)
Number (%) or median (interquartile range)
Maternal age(completed years)
16 222 (43.0)
34 978 (92.6)
28 833 (76.4)
Mode of delivery
20 827 (55.2)
12 596 (33.4)
Pre-existing maternal diabetes
Intrauterine growth restriction
Five-minute Apgar score
37 431 (99.1)
Neonatal intensive-care unit
Term neonatal intensive-care unit admissions
Umbilical artery blood gas values
−2.60 (−4.90 to −0.70)
Of the 37 763 neonates available for evaluation, umbilical cord blood sampling and analysis occurred for 35 030 neonates (92.8%). A variety of reasons were noted for missing data including: no sample being collected (n = 932; 2.5%); analyser-associated technical difficulties (n = 318; 0.8%); clotted samples (n = 291; 0.8%); physiological third-stage management (n = 235; 0.6%); inadequate sample volume (n = 210; 0.6%); and the neonate being born before arrival at KEMH (n = 199; 0.5%). For the remaining 548 neonates (1.5%), no reason for missing data was recorded. Paired samples from both the umbilical artery and umbilical vein were available in 29 874 neonates (79.1%), representing 85.3% of all neonates with UCBG analysis.
The application of the first four validation criteria to the KEMH population is presented in Figure 1. For the Westgate et al., modified Westgate et al. and KEMH criteria, the proportions of neonates excluded from analyses are identical until the application of AV differences for pH and Pco2. The KEMH criteria classified the greatest proportion of neonates with valid UCBG samples and the Kro et al. criteria classified the fewest. The Westgate et al. and modified Westgate et al. criteria have similar proportions of neonates with validated UCBG values. The discriminating point in the selection criteria where there was the greatest reduction in the number of valid samples for analyses was at the one-vessel/two-vessel stage.
The proportion of neonates with validated UCBG values varied in different clinical populations (Table 3). The Kro et al. criteria had the lowest proportion of validated UCBG values for each clinical outcome, whereas the minimal criteria had the highest proportion. In comparison to the original Westgate et al. criteria, the modified Westgate et al., KEMH, minimal and lactate criteria had significantly higher proportions of neonates with validated blood gas values for each outcome (P < 0.001). Conversely, the Kro et al. criteria defined a significantly lower proportion of neonates with UCBG values for each clinical outcome when compared with the Westgate et al. criteria (P < 0.001). Details of the proportion of neonates with validated UCBG values that experience adverse outcomes are summarised in Table 4. Similar to the nursery admission data, the Kro et al. criteria provided UCBG data on fewer neonates than the Westgate et al. criteria (P < 0.001) and the modified Westgate et al., KEMH, minimal and lactate criteria had significantly higher proportions of neonates with validated UCBG data (P < 0.001).
Table 3. Cord blood gas values for clinical outcomes stratified by selection criteria
Cohort for evaluation (from n = 37 763)
Term healthy neonates (from n = 23 536)
Term non-vigorous neonates (from n = 338)
Term neonates to ward (from n = 25 630)
Term nursery admissions (from n = 3203)
Term neonates to NICU (from n = 491)
NICU, neonatal intensive-care unit.
*P < 0.05 compared with baseline Westgate et al.11 criteria.
**UA and UV pH and Pco2 present with ΔUAV pH ≥0.01 and ΔUAV Pco2≥3.4 mmHg.
***UA and UV pH and Pco2 present with ΔUAV pH ≥0.01 and ΔUAV Pco2≥0.2 mmHg.
****UA pH, Pco2 and base excess/lactate present or UV pH, Pco2 and base excess present.
The proportion of neonates with normal UCBG values (arterial pH >7.10 and arterial base excess >−10 mmol/l) was consistent across the Westgate et al., modified Westgate et al. and KEMH validated populations (95.6%). Application of the minimal and lactate criteria resulted in lower proportions of neonates with normal UCBG values (87.3 and 92.7%, respectively), whereas the Kro et al. criteria resulted in a higher proportion (96.5%). The proportion of neonates with severe umbilical artery metabolic acidaemia (arterial pH value <7.00 and base excess values <−12 mmol/l) with valid UCBG using the Kro et al. criteria was 0.2% whereas it was 0.4% using the other five selection criteria (all P < 0.001).
The proportion of neonates with various clinical outcomes with valid UCBG data for the six selection criteria is presented in Table 5. The Westgate et al., modified Westgate et al. and KEMH selection criteria had similar proportions of healthy neonates at term with normal UCBG values. In contrast, the Kro et al. criteria had a higher proportion and the minimal and lactate criteria had lower proportions of neonates with normal UCBG values than the other selection criteria. The proportion of term healthy neonates diagnosed with abnormal UCBG values was identical using five of the six selection criteria with the Kro et al. criteria having a lower proportion (0.1%) than the other five (0.2%).
Table 5. Neonates with normal and abnormal blood gas values stratified by selection criteria and clinical outcome
Criteria (n; % total cohort)
Normal UCBG values* (%)
Borderline UCBG values** (%)
Abnormal UCBG values*** (%)
*Umbilical artery pH >7.10 and umbilical artery base excess >−10 mmol/l.
**Umbilical artery pH ≥7.00 and ≤7.10 and umbilical artery base excess ≥−12 and ≤−10 mmol/l.
***Umbilical artery pH <7.00 and umbilical artery base excess <−12 mmol/l.
A similar pattern of valid UCBG data was seen in term nonvigorous neonates (Table 5). Westgate et al., modified Westgate et al. and KEMH criteria had consistent proportions of neonates with normal UCBG values (76.0–76.1%) whereas the proportion was higher using the Kro et al. criteria (80.9%) and lower for the minimal and lactate criteria (68.2 and 71.9%, respectively). Among term nonvigorous neonates, the six selection criteria had different proportions of neonates diagnosed with abnormal UCBG values ranging from 3.2% (Kro et al. criteria) to 8.2% (lactate criteria).
Term neonates admitted to the nursery had similar patterns of valid normal UCBG values to the other neonatal outcomes previously described. The proportion of valid abnormal UCBG values was lower using the Kro et al. (0.8%) criteria than the other five criteria (1.3–1.5%).
In this study we have compared six selection criteria to determine ‘valid’, physiologically plausible, UCBG data in 37 763 consecutive deliveries at 23 weeks of gestation or later. Comparing the four validation criteria for paired UCBG samples, the KEMH criteria excluded the fewest samples from the study population and provided ‘valid’ data for the greatest proportion of neonates for all clinical outcomes assessed from term healthy neonates to term neonatal deaths. This increase in the proportion of deliveries with valid UCBG data did not increase the proportion of healthy term neonates diagnosed with abnormal blood gas values (false-negative UCBG data). It did, however, result in an increase in the proportion of nonvigorous neonates at delivery with umbilical artery metabolic acidaemia, likely to be true positive UCBG results.
Single-vessel UCBG data are obtained in approximately 14% of term deliveries, with single-vessel data over-represented in babies with adverse neonatal outcomes. In this study we have evaluated two selection criteria to determine valid single-vessel samples. Of these criteria, the ‘minimal’ criteria excluded the fewest samples without any increase in the proportion of healthy neonates diagnosed with abnormal UCBG values. Moreover, the ‘minimal’ criteria provided valid blood gas data for the largest proportion of babies with hypoxic ischaemic encephalopathy, neonatal seizures, low Apgar scores and neonatal deaths. These valid UCBG results are likely to assist clinicians who are treating neonates with adverse outcomes to identify the potential aetiology for their clinical status and to identify neonates for potential interventions.
The presence of umbilical artery acidaemia was an inclusion criterion for many of the randomised control trials evaluating the clinical utility of neonatal neuroprotective hypothermia,15–20 and is one of the essential criteria for the diagnosis of acute intrapartum hypoxic events.10,14 Changing from the Westgate et al. to the KEMH criteria for determining accurate paired UCBG samples would result in a 6% increase in the proportion of neonates with hypoxic ischaemic encephalopathy considered to have valid cord-blood gas values. If the selection criteria were expanded to the ‘minimal’ criteria, there would be a 23% increase in the proportion of neonates with valid paired cord blood gas values available to consider for neuroprotective therapy. Similarly, among those infants that die within the neonatal period, there would be a 31% increase in the number of neonates with validated blood gas data if the ‘minimal’ criteria for accuracy were adopted.
In clinical care, one of the most problematic situations is a nonvigorous neonate with a single umbilical vessel sample or invalid values from paired samples using the current accuracy criteria. To address this issue, we have evaluated the impact of using the minimal criteria—a pH, Pco2 and an indicator of metabolic acidaemia (base excess or lactate) available from one vessel. In a large population of 18 909 term healthy neonates, using the minimal criteria as a method of determining validity of results did not increase the proportion of neonates diagnosed with abnormal UCBG results (0.2%). Of greater importance, the use of the minimal criteria increased the proportion of term nonvigorous neonates with abnormal or borderline UCBG results with a corresponding decrease in the numbers diagnosed with normal UCBG results. Taken together, these data demonstrate that if paired cord blood samples are not available in nonvigorous neonates, the use of pH, Pco2 and base excess/lactate from a single vessel is appropriate because it does not increase the number of false positives but does provide valuable acid–base data to the neonatal team.
Historically, the proportion of samples declared non-valid varies from 5 to 37% of the population evaluated.3,7,11,13,21–28 Apart from the inability to obtain an arterial and/or venous sample, the most common postulated mistake is the inadvertent switching of umbilical arterial and venous samples or sampling the same umbilical vessel twice (typically the umbilical vein). Early studies of UCBG values relied on the skill of the midwifery and medical staff members’ sampling technique, and typically only reported values from a single umbilical vessel. It was not until Westgate et al.11 developed a model to exclude results with a high likelihood of incorrect sampling that consideration was given to sample accuracy. The models derived to identify accurate sampling are based on a series of minimum accepted differences between arterial and venous values for pH and Pco2 derived from the underlying principles of hydrogen and carbon dioxide diffusion across the placenta. The optimum minimum difference between arterial and venous pH is contentious with considerable divergence in opinion.24,26,27,29 The two previous studies to evaluate accuracy of paired UCBG values have concurred on a minimum difference of 0.02 pH units.11,13
The primary issue associated with the use of a centile-based umbilical AV difference to define accurate sampling is that a set percentage of the population is always going to be classified as having invalid UCBG values. The Westgate et al. and Kro et al. criteria use the fifth centile AV pH difference and the tenth centile AV difference for Pco2 as criteria to exclude paired single-vessel samples.11,13 Westgate et al. detailed the choice of the tenth rather than fifth centile for AV difference for Pco2 because the Pco2 electrode is less accurate than the pH electrode in standard blood gas machines.11 The KEMH criteria described in this study also used the fifth centile for AV pH difference to exclude paired single-vessel samples; however, the fifth centile AV difference in our large population of consecutive deliveries was 0.01 pH units. This small change in the exclusion criteria resulted in an additional 2.2% of samples remaining for subsequent analyses.
The standard deviation for the pH electrode in most blood gas machines is ±0.005 pH units. During routine calibration of the pH electrode, any values beyond two standard deviations (±0.01 pH units) for the calibration solutions results in a failure of quality control. Interestingly, the fifth centile AV pH difference derived from the KEMH population is the same as the accuracy of the pH electrode. This provides further evidence for the AV pH difference of ≥0.01 to be used to exclude paired single vessel samples.
In the KEMH criteria, the fifth centile AV difference for Pco2 is used rather than the tenth centile detailed in the Westgate et al. and Kro et al. criteria. This change, in conjunction with the change in pH AV difference in the KEMH criteria, resulted in an additional 4.2% of neonates having validated UCBG samples. The increase in valid samples did not result in an increase in false-positive UCBG results: there was no increase in the proportion of healthy neonates diagnosed with abnormal cord blood gas results. The accuracy of the Po2 and Pco2 electrodes in most blood gas machines is ±0.5 mmHg. The AV differences in Pco2 required to exclude paired single-vessel samples is 7.6 standard deviations of electrode accuracy in the Westgate et al. criteria and 10.5 standard deviations in the Kro et al. criteria. Both of these AV differences are probably beyond that which is required to reliably identify paired single-vessel samples.
The Westgate et al. criteria were developed in a small cohort of 2013 pregnancies ≥34 weeks of gestation where each fetus was monitored with a fetal scalp electrode. Further, for the initial period of the Westgate et al. study, research personnel rather than clinical staff collected UCBG samples. There are also limitations in the population used by Kro et al. to evaluate the accuracy of paired UCBG samples. Although the Kro et al. study population was larger (n = 36 432 UCBG samples) than the sample used by Westgate et al., it was obtained from three different study populations: two low-risk populations and one intensively monitored population. Each study subpopulation had different inclusion and exclusion criteria complicating the applicability of the results to a general obstetric population. In this study, we have addressed these limitations by collecting a large sample of consecutive deliveries (37 763) where all UCBG samples were collected and measured by clinical rather than research staff.
During this study, UCBG analysis occurred in concert with the traditional approach to the management of the third stage, including early clamping of the umbilical cord, if delayed cord clamping was adopted more broadly, further analyses of AV differences would be required. It is important to note, however, that the greatest value of UCBG analysis is in the management of babies that are nonvigorous at birth. In such a clinical situation, typical management would entail early cord clamping and transfer of the newborn to a resuscitation cot and into the care of neonatologists. Given the relative rarity of delayed cord clamping, it would currently be difficult to accumulate a database of the size in this manuscript for the development of applicable validation criteria.
Use of the KEMH criteria for identifying paired ‘accurate’ UCBG samples results in an increase in the proportion of deliveries where UCBG results are considered accurate. This change occurs without increasing the rate of false-positive diagnoses of acidaemia in vigorous neonates. It does, however, provide more ‘valid’ data to aid clinicians when neonates are born in poor condition. If only one blood sample can be collected from the umbilical cord at delivery, using the ‘minimal’ criteria described in this study results in a 31% increase in the number of neonates with valid blood gas data to aid clinicians in the management of neonates born in poor condition.
Disclosure of interest
The authors have no relevant financial (for example patent ownership, stock ownership, consultancies, speaker’s fees, shares), personal, political, intellectual (organising education) or religious interests requiring disclosure.
Contribution to authorship
The study was conceived and designed by CRW, CEP, DAD and JPN. and data were acquired by CRW, DAD, CEP and RK. CRW, CEP and DAD were responsible for the analysis and interpretation of data. The manuscript was drafted by CRW, CEP and DAD and revised by CEP, DAD, RK and JPN. DAD was the statistical expert and RK was responsible for administrative, technical and material support. The study was supervised by CEP, JPN and DAD.
Details of ethics approval
This study received ethics approval from the Women’s and Children’s Health Service Ethics Committee on 4 September 2007 and the 3 February 2009 (reference numbers EC07-12 and 1616/EW, respectively).
CRW was supported by an Australian Postgraduate Award from the Commonwealth of Australia and a PhD Top-Up Scholarship from The University of Western Australian, Perth, Western Australia.
We thank all the medical and midwifery staff members at King Edward Memorial Hospital for collecting all the required samples and the medical records staff for facilitating access to clinical records where required. James Humphreys and Paul Antoine provided expertise in computer programming and development of the study database. Maureen Hutchinson extracted the necessary data from the obstetric databases at King Edward Memorial Hospital.