Umbilical cord characteristics and their association with adverse pregnancy outcomes: A systematic review and meta-analysis

Objective Current data on the role of the umbilical cord in pregnancy complications are conflicting; estimates of the proportion of stillbirths due to cord problems range from 3.4 to 26.7%. A systematic review and meta-analysis were undertaken to determine which umbilical cord abnormalities are associated with stillbirth and related adverse pregnancy outcomes. Methods MEDLINE, EMBASE, CINAHL and Google Scholar were searched from 1960 to present day. Reference lists of included studies and grey literature were also searched. Cohort, cross-sectional, or case-control studies of singleton pregnancies after 20 weeks’ gestation that reported the frequency of umbilical cord characteristics or cord abnormalities and their relationship to stillbirth or other adverse outcomes were included. Quality of included studies was assessed using NIH quality assessment tools. Analyses were performed in STATA. Results This review included 145 studies. Nuchal cords were present in 22% of births (95% CI 19, 25); multiple loops of cord were present in 4% (95% CI 3, 5) and true knots of the cord in 1% (95% CI 0, 1) of births. There was no evidence for an association between stillbirth and any nuchal cord (OR 1.11, 95% CI 0.62, 1.98). Comparing multiple loops of nuchal cord to single loops or no loop gave an OR of 2.36 (95% CI 0.99, 5.62). We were not able to look at the effect of tight or loose nuchal loops. The likelihood of stillbirth was significantly higher with a true cord knot (OR 4.65, 95% CI 2.09, 10.37). Conclusions True umbilical cord knots are associated with increased risk of stillbirth; the incidence of stillbirth is higher with multiple nuchal loops compared to single nuchal cords. No studies reported the combined effects of multiple umbilical cord abnormalities. Our analyses suggest specific avenues for future research.


Introduction
Umbilical cord abnormalities (UCA) usually describe situations where fetal blood flow is reduced or interrupted due to altered structure or function of the umbilical cord. UCA are associated with adverse pregnancy outcomes including stillbirth, birth asphyxia and emergency Caesarean birth. However, estimates of the contribution of UCA to these outcomes vary; for example between 3.4% to 20% of stillbirths are reported to be caused by UCA [1]. Some of the variation may be due to the use of different classification systems for stillbirth, not all of which include UCA as a cause of death.
Of the possible reported UCA, nuchal cord, where the umbilical cord is wound at least once around the fetal neck [2], has been the subject of the most studies; its incidence increases throughout gestation, peaking at birth [3,4]. While there are reports of nuchal cord in individual cases of stillbirth [5], data from larger studies are conflicting, with some finding significant associations [5,6] and others reporting no effect of nuchal cord on stillbirth [7,8]. Other UCA, including true knots and cord prolapse are rarer, but are also linked to adverse outcomes; cohort studies have demonstrated associations between true knots and perinatal death and between cord prolapse and low Apgar scores [9,10]. In addition, an excessive or reduced number of coils of blood vessels within the cord has also been associated with various adverse outcomes [11]. UCAs can also present in combination, for example true knots may occur more in longer cords which are also more prone to entanglement [12,13], complicating the appreciation of the significance of individual abnormalities.
Variation in published results may be due to differences in study design, mode of detection (at birth or antenatal ultrasound), definitions of abnormalities, and lack of information about characteristics such as the number of loops of nuchal cord [14], tightness of cord loops or knots [15], or duration of UCA. To address these uncertainties and to better understand the association between UCA and adverse pregnancy outcomes we undertook a systematic review and meta-analysis of observational studies to describe the normal characteristics of human umbilical cord, the incidence of UCA in singleton pregnancies, and to determine the association between UCA and adverse pregnancy outcomes. We also aimed to understand potential sources of variation between studies.
incidence of UCA and their relation to adverse pregnancy outcomes or case control studies that compared pregnancies with and without UCA, or that looked at the incidence of UCA in adverse outcomes were also included. Inclusion criteria were studies of singleton pregnancies after 20 weeks of gestation, without congenital abnormalities, conducted in hospital settings (secondary or tertiary centres). Studies reporting UCA in multiple pregnancies were excluded as cord entanglement is a specific complication of monoamniotic twins. Studies of vasa praevia were not included as there is a recent systematic review [17]. All other umbilical cord abnormalities were considered for inclusion in this review.
Literature searches were conducted in MEDLINE, EMBASE, CINAHL and Google Scholar to identify relevant papers published since 1960. In addition, references from articles found, conference proceedings, and bibliographies from review articles and book chapters were examined for appropriate references. Searches were initially performed in May 2018 and updated on 1st December 2019. Example search strategies for the association between cord abnormalities and adverse pregnancy outcomes can be seen in Appendix A.

Outcomes of interest
The primary outcome for this review was stillbirth or intrauterine fetal death (IUFD), defined as death of a baby before birth and after 20 weeks' gestation (although the definitions employed in studies were anticipated to vary according to geographical location). Secondary outcomes studied were: neonatal intensive care unit (NICU) admission, preterm birth (<37 weeks' gestation), small-for-gestational-age (birthweight <10th centile or as defined by study), low birth weight at term (<2500g), low Apgar score (<7 at 1 minute and 5 minutes) and frequency of caesarean birth. These outcomes were selected because they reflect a proposed pathway through which UCA can lead to fetal death either acutely antenatally or intrapartum, or, if the UCA were present chronically, cause fetal vascular malperfusion leading to small for gestational age infants or sufficient intrapartum compromise [18], subsequent intervention in labour (Caesarean section), low Apgar score, and/or NICU admission (Fig 1). We anticipated that a positive association would be more compelling if it was associated with a number of these related outcomes.

Study selection and data extraction
Titles and abstracts were reviewed by two authors (from DH, JB, LC, AH) to identify relevant studies and full text papers were obtained. Data were extracted by two authors using a prepiloted data extraction form (from DH, JW, RB, MP, JB, LC, AH); disagreements were resolved by consultation with a third author. Studies not published in English were translated where possible. When full text was not available for a study its authors were contacted, abstracts were not included if all necessary information was not present.
from Cochran's chi-squared statistic Q, was calculated to describe the percentage of variability in effect estimates that is due to heterogeneity. Heterogeneity was classified as low (I 2 = 0-40%), moderate (I 2 = 41-60%), substantial (I 2 = 61-80%), or considerable (I 2 = 81-100%) [22]. Subgroup analyses were performed to investigate heterogeneity where appropriate and funnel plots were created to test for sample size effects.
Incidences of UCA were calculated using the command metaprop [23]. The relationship between presence of UCA and adverse outcome was investigated using the command metan [24]. Planned subgroup analyses were performed to examine the effects of different forms of UCA such as number of loops of cord and whether the cord could be unwound at birth. Although not originally an aim of this study, we also found papers that allowed us to calculate the diagnostic accuracy of ultrasound for detecting UCA. The STATA command metandi [25] was used to calculate the summary sensitivity and specificity from these studies and to produce an HSROC curve.

Study selection and study characteristics
After screening of 2,755 abstracts, 275 full text manuscripts were assessed and 145 studies met the inclusion criteria for this review (Fig 2). Two authors [26,27] provided further information about their studies when contacted. Key characteristics of included studies are presented in Table 1.

Risk of bias of included studies
Quality of included studies was mostly judged to be fair: 35 studies were judged to be good and 7 poor, with the remaining 103 studies judged as fair quality using the NIH quality assessment tools. Most studies had issues with at least one of the following criteria: providing sample size justifications; measuring different levels of exposures, for example the number of loops of nuchal cord; defining exposure or outcome measures, such as the definitions of UCA or gestational age at birth or blinding of exposure and outcome assessors.

Synthesis of results
Results are presented in four sections: normal characteristics of umbilical cord, incidence of UCA, diagnostic accuracy of ultrasound, and associations between UCA and adverse pregnancy outcomes.
Normal characteristics of umbilical cord. The average cord length at birth was found to be 56.0±11.1cm, using data from 39 studies of 94,849 pregnancies. Studies used a range of definitions, but if a study presented data for several different gestational age periods, then the one closest to term was used for this analysis. The average cord length at birth at 39 weeks' gestation was 55.6±12.4cm, using data from 11 studies of 13,263 pregnancies; this was chosen as it was the gestational age reported by the most studies. The mean umbilical coiling index at birth, defined as the complete number of vascular coils divided by the cord's length in centimetres [28], was 0.24±0.10 coils/cm using data from 21 studies of 8,315 pregnancies.
It was not possible to calculate the incidence of other entanglements due to variation in study definitions and outcomes.
Cord prolapse. Incidence of cord prolapse was calculated from 21 studies of 11,057,165 pregnancies; the overall incidence was 0.17%.
Abnormal coiling. Twenty-one studies reported the frequency of abnormal coiling, but the incidences of hypercoiling and hypocoiling were not calculated as they were generally defined using the 90 th and 10 th centiles respectively.
It is important to note, however, that the actual measurements used to define these centiles differed between studies depending on the populations.
Abnormal length. The incidence of abnormal cord length could not be recorded due to wide variation in study definitions. Definitions of a long cord ranged from >59.0cm to >95.0cm, and definitions of a short cord ranged from <35.0cm to <50cm.
Diagnostic accuracy of ultrasound. Nuchal cord. We identified 12 papers which reported the diagnostic accuracy of ultrasound scanning for predicting nuchal cord at birth, these are described in the characteristics of included studies table (Table 1). For this element, a positive index test result was any nuchal cord suspected antenatally using ultrasound and the reference standard was the presence of a nuchal cord at birth. Results were combined for ultrasound screening at any gestation; four studies [7,[30][31][32] performed screening immediately prior to induction or during labour and in all but two studies all measurements were performed after 36 weeks [33,34].
QUADAS-2 was used to quantify the risk of bias and applicability concerns for each included study. Most papers were at low or unclear risk of bias for all domains. Akkaya et al.  [30] was judged to be at high risk of bias for patient selection and index test domains, while Gonzalez-Quintero et al.
[34] was judged to be at high risk of bias for patient selection; these were both case control studies. Studies were all low or unclear risk of bias for applicability concerns. Only one study [32] was judged to be low risk for all domains. Six studies blinded reference standard results [3,4,32,[35][36][37] all but one of these [36] also blinded index test results. All other studies did not state whether blinding took place. Summary sensitivity for ultrasound at all gestations was 80.5 (95% CI 66.3, 89.6), summary specificity 86.6 (95% CI 80.0, 91.2). However, there was considerable variation in sensitivity of individual studies ranging from 29.0 to 96.8%, with specificities ranging from 57.0% to 96.6%. The positive likelihood ratio (LR+) was 6.01 and the negative likelihood ratio (LR-) was 0.17. Sensitivities and specificities from each study were used to produce an HSROC plot (Fig 3); the diagnostic odds ratio (DOR) for ultrasound scanning at all gestations was 26.6 (95% CI 9.46, 74.7). There did not appear to be a linear relationship between accuracy and gestational age although for studies where ultrasound scanning was performed in early labour the sensitivity values were higher, ranging from 90.2 to 96.8%.
True knots. The accuracy of ultrasound for the detection of true knots at birth could not be analysed due to a lack of available study data.
Associations between UCA and adverse pregnancy outcomes. Nuchal cord and stillbirth. When data for any nuchal cord at birth were pooled, no statistically significant association was detected between presence of any nuchal cord and stillbirth (OR 1.11; 95% CI 0.62, 1.98). Heterogeneity was moderate (I 2 44.4%, p = 0.055). As no association was detected for a single loop of nuchal cord versus controls (OR 0.87; 95% CI 0.56, 1.35), data were combined for an analysis comparing multiple loops to combined data for no loop and a single loop. This resulted in an OR of 2.36 (95% CI 0.99, 5.62; p = 0.053) for multiple loops of nuchal cord (Fig 4). Heterogeneity for this analysis was low (I 2 7.0%, p = 0.372). Comparing multiple loops to no loops, excluding single loops from the analysis, resulted in an OR of 1.91 (95% CI 0.90, 4.06). Heterogeneity was again low (I 2 0.0, p = 0.623). 123 stillbirths from 40,114 pregnancies were included in the pooled analysis. There was no evidence of small study effects (Harbord's test, p = 0.137).
Three studies presented data for the relationship between nuchal cord detected by ultrasound at any gestation and stillbirth, no statistically significant association was detected (OR 0.72; 95% CI 0.17, 3.05). Heterogeneity for this analysis was low (I 2 = 25.8%, p = 0.260). Data from this analysis were obtained from 1955 pregnancies, 24 of which were stillbirths.
Nuchal cord and other adverse outcomes. Results from analyses of the relationships between nuchal cord and all secondary adverse outcomes are shown in Table 2. Analyses could not be performed for the association between nuchal cord detected using ultrasound and Apgar scores <7 at 5 minutes, NICU admission, small-for-gestational age, or preterm birth, or for the association between nuchal cord at birth and preterm birth.
A single loop of nuchal cord at birth was only associated with a 1 minute Apgar score <7 whereas multiple loops of cord were associated with increased likelihood of caesarean section and Apgar scores <7 at both one and five minutes. Tight loops of nuchal cord but not loose loops were associated with low Apgar scores.
We found no evidence for an association between nuchal cords at birth and NICU admission. Overall heterogeneity was substantial at 66.8% (p<0.01); this was due to considerable heterogeneity in the data from multiple loops (I 2 88.5%, p = 0.00), whereas heterogeneity in the data for single loops was low (I 2 30.5%, p = 0.22). However, the likelihood of NICU admission with a tight nuchal cord at birth was twice as high as with no nuchal cord, although this was not statistically significant.
No included studies specified nuchal cord as an indication for birth; if a study presented emergency caesarean section or caesarean section for fetal distress separately, then these data

PLOS ONE
Umbilical cord characteristics and adverse pregnancy outcomes were used instead of the overall rate. However, caesarean section was significantly more likely in pregnancies with nuchal cord detected via ultrasound (OR 1.64; 95% CI 1.07, 2.51) Heterogeneity was low (I 2 33.5%, p = 0.211).
Sensitivity analyses for nuchal cord papers. No studies that were rated poor by quality assessment presented data for nuchal cord and its relationship to adverse outcome so planned sensitivity analyses were not performed.
True knots and stillbirth. The likelihood of stillbirth was significantly higher in pregnancies with a true knot in the umbilical cord at birth than in those without, with an OR of 3.96 (95% CI 1.85, 8.47; 7 studies of 930,314 births) (Fig 5) Heterogeneity was moderate (I 2 60%, p<0.05).
True knots and other adverse outcomes. Results from these analyses are presented in Table 3. Statistically significant associations with modest effect sizes were found between true cord knots at birth and all of our secondary outcomes except for caesarean section.We were not able to look at the association between true knots at birth and low Apgar scores at 1 minute. No evidence of small study effects was seen for our main outcome of stillbirths in studies of true knots; Egger's test gave a p value of 0.27 (Fig 6).
Abnormal coiling and intrauterine fetal death. We were unable to perform meta-analysis to analyse the relationship between abnormal coiling of the umbilical cord and stillbirth Abnormal coiling and other adverse outcomes. Outcome data are shown in Table 4. For all coiling analyses the hypo-or hypercoiled group was compared to the group with normal coiling only. The only exception is a study of hypocoiling by Strong,Finberg & Mattox [46] where the control group was all cords with an umbilical coiling index (UCI) above the 10 th centile (so would also have included hypercoiled cords). Analyses were also performed combining all thresholds for hypo-and hypercoiling as in most cases variation was minimal; for UCI at birth the range of thresholds classed as hypocoiling was from <0.6 to <0. 17  stated that they used <10 th and <90 th centile thresholds but did not specify the actual measurements to which these corresponded. Studies that measured UCI antenatally were not analysed with studies that measured coiling at birth; definitions for hypo-and hypercoiling from these studies also tended to differ, potentially due to the gestation at measurement. Apgar score <7 at 1 minute was measured with UCI at birth by three studies [48-50] and antenatal UCI by another [51]. One study defined a low Apgar score at 1 minute as below 4 [52] and another combined all poor Apgar scores [47]. These studies were not included in this analysis. Thresholds for diagnosis of hyper-or hypocoiled cords are displayed on the forest plots and did not appear to lead to any variation in the effect sizes between studies. Sensitivity analyses were also performed for the SGA/FGR analyses based on whether a definition for this outcome was provided by the study, but no effect was seen. For hypercoiling the OR was reduced once studies with unclear definitions were removed but there was still a statistically significant association with SGA.
Abnormal cord length and adverse outcomes. We did not perform any analyses of the relationship between abnormal cord length and adverse outcomes due to variation in study definitions, as described earlier. Cord prolapse and adverse outcomes. We were unable to perform meta-analysis to investigate the relationship between cord prolapse and any of our outcomes of interest due to a lack of available data.
A summary of normal cord characteristics, UCA incidences, diagnostic accuracy data, and associations between UCA and stillbirth is provided in Table 5.

Discussion
Our systematic review was able to combine a large amount of data to determine the normal characteristics of umbilical cord and report the frequency of abnormalities. Some cord abnormalities are common, for example nuchal cord was found in 22% of births (95% CI [19][20][21][22][23][24], whereas true knots and cord prolapse are less common (1% and 0.1% of births respectively). The definition of some abnormalities e.g. UCI were consistent between studies but others, such as the length of cord were heterogeneous, due to the thresholds applied to define abnormality which often overlapped with the normal ranges (e.g. pooled mean cord length 56.0±11cm, "long cord" defined as >59cm). Estimates of frequency also varied by gestation studied. We selected stillbirth as our primary outcome, and identified secondary outcomes to reflect diagnosis of fetal compromise that was not sufficiently severe to cause fetal death or represent intervention that prevented it. None of the abnormalities studied showed a significant association with all outcomes and the observed odds ratios were in the range of 1 to 5. This review found the diagnostic accuracy of antenatal or antepartum ultrasound to identify cord abnormalities was modest; the diagnosis of nuchal cord was most accurate when performed in early labour (all studies had sensitivity >90% and specificity >83%). This may be because its incidence is highest at term and there is less time for fetal movements to affect whether the cord is around the fetal neck or not [36,53]. There are insufficient data to determine whether other abnormalities of the umbilical cord can be reliably detected by antenatal ultrasound.

Strengths and limitations
This systematic review was strengthened by being conducted according to a pre-specified protocol by an international multidisciplinary review team to maximise the inclusion of relevant data. Up to 270,973 births were included in the meta-analyses giving robust estimates of effect size. However, this review is limited by variation in the definitions used to define both UCA and the associated outcomes, which restricted the number of analyses that can be reliably performed. We were also not able to identify any unpublished data suitable for inclusion, meaning that some of our effect sizes may be overstated due to publication bias.
Our proposed pathway for the differential effects of chronic vs. acute UCA is also limited in that we could mostly report evidence of UCA at birth without knowing how long it had been present, and could not address temporal variation (i.e. whether nuchal cord had been intermittently present). Due to the nature of our included studies we were also unable to distinguish between events that occurred antepartum vs. intrapartum or acute vs. chronic effects, nor could we look at the effects of combinations of UCA, which could affect the likelihood of adverse outcomes, for example shorter cords may lead to tighter nuchal cords and knots when they are present while longer cords may be more prone to entanglements (studies have shown the average cord length to be higher in cases with nuchal cord) [54]. Limb and body entanglements were also not recorded by the majority of studies.
Quality assessment showed that recording of UCA needs to be far more stringent, especially in nuchal cords; number of loops, tightness (which is unlikely to be a true dichotomous variable [55] and also may change during labour so tightness at birth may not reflect tightness antenatally) [56], and type of nuchal cord (A or B, indicating whether the cord is in a locked pattern or can easily be unwound) [57] should all be recorded along with whether other entanglements were present. Classification of stillbirth also requires improvement so that umbilical cord pathology is accurately recorded. Early classification systems such as the Wigglesworth classification did not include umbilical cord complications as a cause of perinatal death. Even when modern classification systems are applied, there is variation in recording of umbilical cord complications resulting in the estimated incidence of cord complications varying from 3.4% to 20%. A recent detailed analysis using the INCODE system suggested 19% of stillbirths are due to cord accident [58]. Variation in reporting of umbilical cord pathology would be reduced by a core outcome set for studies examining the association between UCA and adverse outcome.

Clinical implications
Our data demonstrate that UCA are associated with adverse perinatal outcomes. The broadest range of associations with stillbirth and associated adverse outcomes were seen for true knots, following by coiling abnormalities then nuchal cord, whereas the strongest effect sizes were for tight nuchal cords. Robust information about the diagnostic accuracy for UCA is only available for nuchal cords, in this case the pooled sensitivity and specificity of antenatal ultrasound was 80.5% and 86.6% respectively. Nuchal cords were only associated with adverse outcomes when either multiple or tight loops were present. On the basis of this information identifying an isolated nuchal cord antenatally is unlikely to prevent adverse outcome, but may increase intervention. However, combining identification of nuchal cord and abnormal umbilical artery flow increases the likelihood of intrapartum compromise [7,59]. Further testaccuracy studies are needed, but must be appropriately blinded to prevent intervention altering the outcome.
We hypothesised that umbilical cord abnormalities act via a common pathway of restricting blood flow to the fetus which may be acute or chronic (Fig 1). Thus, we expected to see associations with stillbirth and the secondary outcomes investigated in this meta-analysis. Given the comparatively modest effect size of the relationship between UCA and the outcomes studied here, we conclude that all stillbirths and adverse perinatal outcomes should be thoroughly investigated, even when UCA are present at birth to determine whether a) histopathological changes consistent with UCA are present, including lesions of fetal vascular malperfusion [42], and b) to exclude other possible causes, in order that a robust link may be made between the outcome and antecedent cord complications. In the context of stillbirth, the triple risk model proposes that fetal deaths can result from a combination of fetal stressors, maternal factors, and placental or fetal vulnerability [60]. Applying this model, UCA is a fetal stressor, where stillbirths occur with combinations of risk factors such as reduced placental perfusion, and maternal factors such as maternal obesity, or maternal sleep position. Further studies are also needed to understand the biological mechanisms underpinning UCA and adverse outcomes. For some, such as tight loops of cord or a true knot, this may be from direct occlusion, whereas in hyper-or hypo-coiled cords this may reflect haemodynamic consequences or developmental abnormalities. Larger datasets applying consistent thresholds for abnormalities are required to accurately determine the relationship of UCA to adverse perinatal outcomes.

Conclusion
This systematic review and meta-analysis has demonstrated links between UCA and several adverse pregnancy outcomes, although not all analyses were adequately powered and some comparisons were restricted by the methodologies of the original studies. Further studies are needed to allow robust clinical recommendations on the management of UCA to be made. These should make use of the information presented about normal cord characteristics to inform thresholds for abnormalities and examine multiple UCA and a range of adverse perinatal outcomes. Ideally, UCA should also be recorded antenatally in blinded studies so that prognostic accuracy can be calculated. Until such data are available, clinicians should be cautious about assigning causality of an adverse outcome based on an isolated observation of UCA.