Measurement of spiral artery jets: general principles and differences observed in small-for-gestational-age pregnancies

Authors


Abstract

Objective

To investigate whether the jets of blood from the mouths of the spiral arteries could be measured reliably, as well as their relationship with the uterine artery (UtA) and any differences in small-for-gestational-age (SGA) pregnancies.

Methods

Participants underwent serial ultrasound scans, from 11 weeks' gestation. Pulsatility index (PI) and resistance index (RI) of jets into the intervillous space (IVS) and UtA were recorded at every visit. Intra- and interobserver variability studies were performed. Customized birth weight centiles were calculated and SGA was defined as < 10th centile. Linear mixed model analysis was used to allow for the longitudinal nature of the data.

Results

Sixty-six women were recruited; 58 remained normotensive and delivered at term. Of these, six women delivered SGA newborns and 52 delivered appropriate-for-gestational-age newborns. All had pulsatile jets until 20 weeks' gestation. The PI and RI of the jets decreased with advancing gestation, following a trend similar to that of the UtAs. There was no correlation between the jets and UtA waveforms when gestational age was controlled for. For intraobserver variability the intraclass correlation coefficient was 0.9. The interobserver study showed no significant difference between the observers. Mixed model analysis demonstrated that PI and RI of jets were different in SGA pregnancies (P < 0.06). This difference was not seen for the UtAs (P = 0.8).

Conclusion

This technique enables examination of characteristics of the jets of blood flowing from spiral arteries into the IVS. It is both precise and reproducible, with biologically plausible results. Further work is required to assess differences in pregnancies with adverse outcomes. Copyright © 2012 ISUOG. Published by John Wiley & Sons, Ltd.

Introduction

Measurement of impedance of flow through the uterine artery (UtA) is an important but imperfect component of screening for pre-eclampsia and fetal growth restriction (FGR), two of the most serious and common problems in pregnancy. In both, histological data have implicated retention of the musculoelastic walls of the spiral arteries resulting from inadequate trophoblastic invasion1–3. It is widely purported that UtAs show an ‘increased resistance’ waveform because of these poorly converted, tighter, spiral arteries, although direct evidence of this mechanism is lacking. These changes may result from direct ‘back pressure’ effects or may be instigated simply by the same overall controlling mechanism4, 5. The possibility of measuring actual pathology at the uteroplacental interface rather than an ‘upstream’ surrogate marker could improve our understanding of the mechanism, and may increase the accuracy of screening tests for resulting adverse outcomes.

Examination of spiral arteries has been attempted in both normal pregnancies and those with adverse outcomes6–11 but results are conflicting and the correlations between normal and adverse pregnancy outcomes are weak. The commonly employed method relies on identification of spiral arteries by approximate location in the placental bed and by their recognizable bidirectional waveform8. This method has major limitations as no exact anatomical landmarks are used. Therefore, it is impossible to measure vessels at the same point. Recognition by accepted waveform alone might also inherently select ‘normal’ vessels, thereby missing pathology, and angle correction is inappropriate because of the tortuous nature of these vessels.

The decidua–placental interface where spiral arteries discharge into the intervillous space (IVS) is readily identifiable with grayscale ultrasound. We hypothesize that, using color Doppler, the single unidirectional signal appearing from the tangle of maternal vessels at the decidua–placental interface are jets of blood discharging from spiral arteries into the IVS. Identification of this point allows measurement at a fixed anatomical site: the mouth of the spiral artery. As blood flow is unidirectional, angle correction can be used. Furthermore, as flow is into the low resistance IVS, the waveform should largely represent the musculoelastic nature of the terminal end of the vessel. Therefore, this should allow direct examination of the pathology implicated in impaired placentation.

We sought to investigate: 1) whether the waveform of the jet of blood from the mouth of the spiral artery could be reliably and reproducibly measured using two-dimensional color Doppler; 2) whether this waveform altered in a gestation-dependent manner in concordance with histological data; 3) the nature of the relationship between this waveform and that of the UtAs; 4) whether jets are different in pregnancies resulting in small-for-gestational-age (SGA) newborns at term.

Methods

Women with singleton pregnancies attending the ultrasound department for routine scan were invited to participate. Women under the age of 16 years, those with a body mass index (BMI) > 35 kg/m2, those with significant chronic illness including diabetes, and those under treatment with medications associated with FGR such as beta-blockers12 were excluded. The study involved seven ultrasound scans during pregnancy, starting between 11 and 13 weeks' gestation. Sociodemographic and obstetric data were collected, including age, parity, family history and past medical and obstetric histories. Gestational age was calculated from crown–rump length measurement at the first scan, performed between 11 + 0 and 13 + 6 weeks' gestation13. The subsequent three scans were undertaken at fortnightly intervals after recruitment, with the last three taking place at 23, 28 and 33 weeks' gestation. Written consent was obtained and the study was conducted with NHS REC ethical approval (REC ref: 08/H0604/163).

Scans were undertaken by a single operator (S.C.) using an M6C curved array abdominal transducer on a Voluson E8 (GE Healthcare Ultrasound, Milwaukee, WI, USA). Machine settings were constant for all participants; in particular, the wall motion filter was set to ‘low 2’, the color Doppler gain at − 12.0 dB and the pulse repetition frequency at 0.9 kHz. After confirmation of viability the UtAs were identified using standard techniques. Before 16 weeks' gestation this was lateral to the cervix at the level of the internal os as the artery turns cranially14. After 16 weeks' gestation they were identified on a longitudinal axis, lateral to the uterus, where the uterine and external iliac arteries appear to cross15. Pulsed-wave Doppler ultrasonography, with a gate size of 2 mm, was then used to obtain at least three consecutive waveforms of good quality. The angle of insonation was kept at less than 30° and angle correction was applied if necessary. The pulsatility index (PI) and resistance index (RI) were calculated over at least three cardiac cycles. The UtA indices used were calculated as the mean of values for left and right vessels and also as the minimum value.

Blood flowing from the mouth of the spiral artery into the IVS was identified using the following technique. The interface between the decidua and the basal plate of the placenta was identified in the sagittal plane (or coronal plane if the placenta was lateral), ensuring that the probe was kept at exactly 90° to the uteroplacental interface. The area was then magnified and color Doppler signal was applied. The mouth of the spiral artery is readily identifiable as the flow becomes unidirectional (single color) at the bright decidua–placental interface (Figure 1a). Pulsed-wave Doppler ultrasonography, with a gate size of 2 mm, was then used to obtain at least three consecutive waveforms of good quality. The angle of insonation was kept at < 30° and angle correction was applied if necessary. The typical waveform is seen to be unidirectional (Figure 1b). Both the PI and RI were calculated over at least three cardiac cycles. This technique was used to measure the three most prominent spiral arteries in the central part of the placenta16. The mean jet PI and RI were calculated from the three individual vessels measured. The scanning time allowed for these measurements was limited to 20 min for safety reasons. Thermal index was kept at < 1.0 and mechanical index at < 0.7 throughout. If, by 20 min, the jets had not been measured the procedure was abandoned.

Figure 1.

(a) Placement of the Doppler gate at the mouth of the spiral artery where the jet of blood enters the intervillous space (IVS) (image has been ‘inverted’, therefore red indicates blood flowing into the placenta). (b) Recording of the typical ‘jet’ waveform, unidirectional with slightly turbulent flow.

Fourteen participants were selected as a subgroup to assess inter- and intraobserver variability. As the range and magnitude of the variance components for each characteristic were unknown, we were unable to define a sample size in advance. Therefore, we carried out ongoing statistical analysis until parameters could be estimated and an adequate sample size defined. As participation in these subgroups required more exposure to Doppler ultrasound, recruitment was discontinued once the sample sizes were deemed by the statistician to be sufficient.

Intraobserver variability was assessed by one observer (S.C.) measuring the waveform three times at the mouth of the same vessel in the same woman at the same visit. For assessment of interobserver variability, measurements were made by two observers (S.C. and L.I.) who were blinded to each other's results. They worked separately to identify the three most prominent vessels and measure these in the same woman at the same visit within 5 min of each other, without the participant moving around between measurements.

Pregnancy data were collected by reviewing the woman's hospital notes. This included antenatal complications, highest recorded blood pressure, gestational age at delivery, mode of delivery, intrapartum complications, sex and birth weight, Apgar scores, admission to the neonatal intensive care unit (NICU) and placental weight. The customized birth weight centile was calculated for each newborn using the Grow software package (version 7.5.1, West Midlands Perinatal Institute, Birmingham, UK). SGA was defined as < 10th centile on customized birth weight charts. Pregnancy-associated hypertensive disorders were defined according to the International Society for the Study of Hypertension in Pregnancy (ISSHP) guidelines17. A normal pregnancy outcome was defined as an uncomplicated livebirth after 37 completed weeks of gestation, customized birth weight ≥ 10th centile, no admission to the NICU and no evidence of maternal hypertensive disorders or gestational diabetes.

Statistical analysis

Statistical analyses were performed in collaboration with the Centre for Statistics in Medicine, University of Oxford, using SPSS (version 16.0, IBM Corporation, NY, USA), SAS (version 9, SAS Institute Inc., Cary, NC, USA) and Stata (version 10.0, Stata Corps, College Station, TX, USA). The methods of linear mixed models were used to analyze data. The dependent variables jet PI, jet RI, mean UtA-PI and UtA-RI, were each measured for each participant at seven gestation time points, between 12 and 35 weeks' gestation. The change in each variable over gestation was modeled using a second-order regression equation. A random coefficient model was fitted, which allows for differences in the participants' rate of change in PI or RI. Random coefficients describe the fitted lines for each participant, and random effects are the deviations of individual random coefficients from coefficients of the mean line. A factor, SGA, indicating whether the newborn is < 10th centile or not, was included in the model, as a main effect and as an interaction with gestation and time, as fixed effects. These fixed effects may explain some of the variation among participants in the intercepts and gradients. Here, interest centered on the effect of interaction of the factor with time on the rate of change in the dependent variable. To model the correlation between measurements over time within participants, the unstructured covariance matrix was used where the variance at each time and covariance between two times are freely estimated. There was no functional relationship imposed on the covariance between two time points. Where measurements of the jet PI and jet RI were not possible due to very low pulsatility, the data were imputed as 0.05. The relationship between the mean jet PI and both mean and minimum UtA-PI at each individual visit was explored using Spearman's Rank correlation. This was repeated to compare the RI for jets and UtA. The PI and RI for SGA and appropriate-for-gestational-age (AGA) newborns were compared at first and last visits after adjusting for gestational age.

The level of statistical significance chosen is always arbitrary and should depend on the type of study. A value of P < 0.05 is common, but for more rigorous tests P < 0.01 is often used. For exploratory studies where the effect size is unknown, and therefore a power calculation impossible, less stringent levels such as P < 0.1 may be chosen. This study is exploratory, involving only a small number of participants, and therefore the level of significance was set at P < 0.1.

Interobserver variability is often assessed using the Bland–Altman technique, which is a specific, simplified derivation of another general method. However, it can be used only when there are two repeats. If the design is more complex, as in our study, the original method must be used. Therefore, the intraobserver variability analysis was carried out using the mixed procedure of SAS. Patients were considered a random effect. Gestational age was considered a fixed effect which has the effect of adjusting for gestational age. The within-participant and between-participants components of variance were estimated. The intraclass correlation coefficient (ICC) was derived from these variance components. Repeatability is calculated as 2√2× within subject variation. It is not dimensionless and depends on the scale of the measurement. It is defined as the upper prediction limit for the absolute difference of two measurements by the same method. Here it refers to measurement of the same vessel, at the same time, by the same observer.

To analyze interobserver variability the data for each assessor were initially analyzed separately. These analyses showed that there might be a difference between assessors for variations within and between participants. The full model, therefore, was designed to take this into account. We used the method of linear mixed models, which allows for correlation among the three measurements for one participant. The assessor was a fixed effect term in the model; the difference between assessors' results over all participants was tested. The mean participant measurements are the mean of the results of measuring three different vessels per participant. As the question of interest is how the mean participant measurements of PI and RI differ between assessors, the limits of agreement were calculated as mean between-participant difference ± 2 × SD (mean between-participant difference).

Results

Sixty-six women were recruited between 11 + 0 weeks' and 13 + 6 weeks' gestation. One participant emigrated after the first two scans and was lost to follow-up.

Three women developed pregnancy-induced hypertension, two developed pre-eclampsia and two delivered before 37 completed weeks' gestation. Therefore, 58 women were normotensive throughout pregnancy and delivered at term. Six of these women gave birth to infants < 10th centile on customized birth weight centiles. Of the 58 women included, 55 underwent all seven scans, while the remaining three underwent only three, four or six scans. Delivery details were available for all 58 women. Therefore, data from 398 longitudinal ultrasound scans in 58 women were available for final analysis.

Baseline characteristics are shown in Table 1 and pregnancy outcomes in Table 2. In one participant jet measurements could not be obtained in the allocated 20 min at first visit. It was possible to measure three jets in all other women at the first four scans, and at the fifth and subsequent scans; from 23 weeks' gestation onwards, the overlying fetus occasionally obscured posteriorly sited placentas. This was most marked at the seventh scan (33 weeks' gestation) when jets could be measured in only 48 (92%) participants.

Table 1. Baseline demographics of participants
CharacteristicSGA (n = 6)AGA (n = 52)P
  • Data are given as mean (range) or n (%).

  • *

    t-test.

  • Fisher's exact test.

  • In AGA group.

  • AGA, appropriate-for-gestational age; IUD, intrauterine death; IVF, in-vitro fertilization; PET, pre-eclampsia; SGA, small-for-gestational age.

Age (years)31 (19–44)28 (16–37)0.44*
Body mass index at booking (kg/m2)23.2 (18–31)23.4 (17–35)0.95*
Smoker1 (17)9 (17)1.00
Parous4 (67)26 (50)1.00
 History of SGA3 (75)6 (23)0.12
 History of PET2 (50)1 (4)0.06
Known obstetric family history (FHx) (n = 51)   
 FHx of SGA1 (17)3 (6)0.5
 FHx of PET1 (17)9 (18)0.6
 FHx recurrent miscarriage01 (2)
 FHx late IUD (> 34 weeks)01 (2)
Self-reported ethnicity   
 White European648
 Indian01
 Pakistani01
 Mixed02
IVF pregnancy050.7
Placental site   
 Anterior322
 Posterior023
 Lateral24
 Fundal12
 Previa01
Table 2. Delivery details
CharacteristicSGA (n = 6)AGA (n = 52)P
  • Data are given as mean (range) or n (%).

  • *

    Student's t-test.

  • AGA, appropriate-for-gestational age; NA, not applicable; NICU, neonatal intensive care unit; SGA, small-for-gestational age.

Customized birth weight centile5 (0–8)47 (10–99)NA
Birth weight (g)2706 (2059–3010)3402 (2450–4730)< 0.001*
Placental weight (g)496 (390–647)637 (380–872)0.002*
Gestational age at delivery (weeks)39 + 2 (38 + 1 to 40 + 4)39 + 4 (35 + 4 to 42 + 2)0.6*
Onset of labor   
 Spontaneous4 (66)41 (79)
 Induced010 (19)
 Elective Cesarean2 (34)1 (2)
Mode of delivery  
 Spontaneous vaginal2 (34)39 (75)
 Instrumental2 (34)8 (15)
 Cesarean section2 (34)5 (10)
Admission to NICU00NA

All jets were visibly pulsatile and could be measured with the autotrace feature of the ultrasound machine until the fourth visit (c. 20 weeks' gestation). At this gestational age one participant was found to have jets with pulsatility so reduced that it could not be measured reliably; the waveform appeared to have a possible suggestion of pulsatility but the autotrace feature recorded it as continuous. The number of women with only non-pulsatile jets gradually increased with advancing gestational age until, at 34 weeks, 19 (37%) had no jets with measureable pulsatility. Consequently, the mean RI of jets was seen to decrease with increasing gestational age (Figure 2a). This follows the same pattern as both the mean and minimum UtA-RI (Figures 2b and 2c). The PI showed the same pattern for jets and UtAs.

Figure 2.

Relationship between mean resistance index (RI) of spiral artery jets (a), mean RI of both uterine arteries (b) and minimum RI of the two uterine arteries (c) and gestational age in appropriate-for-gestational age newborns. Mean and 95% band limits are shown.

Although the PI and RI for jets and UtAs show a similar relationship there was no significant correlation seen between their trends when the strong effect of gestational age was controlled for. There was also no trend observed in the correlation coefficient values for the PI or RI between jets and UtAs when they were analyzed separately at each visit (Table 3).

Table 3. Correlation coefficients at each visit comparing mean jet pulsatility (PI) and resistance (RI) indices to mean and minimum uterine artery (UtA) PI and RI
 Pulsatility indexResistance index
VisitJet vs mean UtAJet vs min UtAJet vs mean UtAJet vs min UtA
10.03− 0.020.02− 0.06
20.410.520.480.55
30.010.03− 0.03< 0.01
40.370.190.280.23
50.290.200.240.17
60.140.350.160.34
70.300.180.290.16

Initially, fourteen women were enrolled to assess intra- and interobserver variation. The data were analyzed at this stage. For intraobserver variability the ICC for PI and RI was 0.9. On assessing the interobserver variability it was seen that this sample size gave a good estimate of variance (Table 4). There was no significant difference between the overall mean established by Observers 1 and 2 over all participants. The decreasing within participant variance for Observer 2 over time showed a possible learning effect. Therefore, with time, the limits of agreement between the two observers may have decreased further. However, using the ‘as low as reasonably achievable’ principle, it was felt that this was sufficient to confirm that there was appropriate reproducibility with no systematic bias.

Table 4. Interobserver reproducibility
IndexMean (SD) for Observer 1Mean (SD) for Observer 2Mean difference (SD) Observer 1− Observer 2Limits of agreement
Pulsatility index0.38 (0.08)0.37 (0.08)0.01 (0.06)− 0.12 to 0.13
Resistance index0.31 (0.05)0.30 (0.05)0.01 (0.03)− 0.06 to 0.07

There were no significant differences in baseline demographics between the SGA group and the controls. The SGA group had significantly smaller placentas at delivery (P = 0.002) despite there being no difference in gestational age at delivery (P = 0.6). No newborns were admitted to the neonatal intensive care unit. All women had UtA Doppler PIs < 95th centile at 23 weeks' gestation.

Figure 3 shows the change in mean RI of jets with increasing gestational age for AGA newborns, with the individual point represented for SGA newborns. The trend for both PI and RI appeared shallower in SGA newborns but did not meet statistical significance. Figure 4 shows the same graph for UtA RI. Results of the linear mixed model are presented in Table 5. The coefficients of gestation2 are significant (P < 0.0001) for both PI and RI, which indicates that the change in PI and RI over time is not linear and that a second-order model is a better fit. This second-order model demonstrated that, at the P < 0.1 level, both PI and RI for SGA newborns were significantly different from those of AGA newborns (P = 0.06). There was no significant difference in either PI or RI of the UtAs between SGA and AGA newborns (P = 0.8). There was also no significant difference between SGA and AGA groups for PI or RI at visits 1 or 7, after adjustment for gestational age (Table 6).

Figure 3.

Change in jet resistance index (RI) with increasing gestational age. equation image, mean for appropriate-for-gestational age newborns, with 95% prediction intervals (equation image). equation image, mean for small-for-gestational age (SGA) newborns. equation image, individual results for SGA newborns.

Figure 4.

Change in uterine artery resistance index (RI) with increasing gestational age. equation image, mean for appropriate-for- gestational age newborns, with 95% prediction intervals (equation image). equation image, mean for small-for-gestational age (SGA) newborns. equation image, individual results for SGA newborns.

Table 5. Linear mixed-model coefficients for Doppler indices over gestation for small-for-gestational age (SGA) newborns compared with appropriate-for-gestational age newborns
 Pulsatility indexResistance index
Fixed effectsCoefficient (SE)95% CIPCoefficient (SE)95% CIP
  1. SE, standard error.

Constant baseline0.97 (0.05)0.86 to 1.080.66 (0.04)0.59 to 0.74
Gestation− 0.05 (0.005)− 0.05 to − 0.04< 0.0001− 0.026 (0.003)− 0.03 to − 0.02< 0.0001
Gestation20.0006 (0.0001)0.0004 to 0.0008< 0.00010.00030 (0.0001)0.0002 to 0.0004< 0.0001
SGA− 0.12 (0.08)− 0.28 to 0.034− 0.092 (0.053)− 0.20 to 0.01
SGAr* gestation0.007 (0.003)− 0.0002 to 0.0130.060.0050 (0.0026)− 0.0001 to 0.010.06
Table 6. Comparison of appropriate-for-gestational age (AGA) and small-for-gestational age (SGA) newborns at first and last visits with respect to jet pulsatility and resistance indices
IndexAGASGAP
  1. Data are given as mean (SD).

Pulsatility index   
 Visit 10.50 (0.13)0.47 (0.15)0.7
 Visit 70.17 (0.11)0.22 (0.14)0.1
Resistance index   
 Visit 10.37 (0.07)0.36 (0.09)0.6
 Visit 70.15 (0.09)0.19 (0.11)0.1

Discussion

In this study we describe the methodology for measurement of the jet of blood from the mouth of the spiral artery into the IVS using Doppler ultrasound. We have shown that the method is feasible and reproducible, and that the jet PI and RI decrease with advancing gestational age, which is likely to reflect changes in the properties of the spiral artery wall. The study also suggests that the jet waveform may be different in pregnancies resulting in SGA newborns, despite little difference being observed in the UtAs.

To the best of our knowledge this is the first time this technique has been formally described and validated. A method has been reported for measuring blood flow in the IVS in the cynomolgus monkey18; however, the exact sampling site was not described and it does not appear to have been reported in humans.

The decrease in PI and RI of the jets (Figure 2a) shows that there is a gestation-dependent change that occurs in the vascular tree at the uteroplacental interface. Histological studies have demonstrated trophoblast-invasion-associated loss of the musculoelasticity of the spiral artery wall1, 3, 19. If it is assumed that the normal IVS has low resistance to the blood flowing into it20, the change in waveform of the jet is likely to reflect the change in properties of the spiral artery wall. As the vessel widens, and the elasticity and subsequent wall rebound is lost, the pulsatile nature of the flow would be expected to diminish, ultimately leading to continuous forward flow, which is the pattern observed in this study. As the last ultrasound scan took place at 34 weeks' gestation, we cannot comment on the number of jets that remained pulsatile at term. However, the overall pattern closely resembles that predicted by histology and models of placental hemodynamics5. If the underlying pathology of this growth restriction at term is due to residual musculo elasticity in the terminal end of the vessel, it is biologically plausible that the jet PI and RI increase as a result of the retained musculoelasticity and increased vascular resistance. As the difference in these indices becomes increasingly prominent in the third trimester, it is possible that this could be the underlying mechanism that results in late-onset growth restriction. Histological data6–8 would not support the alternative possibility, that the cause of the observed difference is a change in the nature of the IVS leading to increased resistance to forward flow.

Any correlation between early pregnancy waveforms and SGA outcome appears greater in jets than in UtAs. This would be surprising if changes in UtAs are caused directly by changes in the downstream spiral arteries, as the total of all vessels in the placental bed should be more sensitive than three individual vessels. If confirmed, this difference may have many potential causes. Failure of trophoblastic invasion leads to the myometrial segment of the spiral arteries retaining its musculoelastic wall in pregnancies complicated by normotensive growth restriction1–3 and pre-eclampsia. It is not clear whether these two entities represent different clinical manifestations of the same pathology, possibly due to variation in the number of affected vessels or the degree of vessel change, or represent maternal response to the pathology1, 21. It is possible that we have managed to measure three affected vessels and the UtA result reflects that there are more normally adapted vessels than abnormal vessels and is therefore overall normal. However, it is more likely that blood supply from the uterine to the spiral arteries is not a simple ‘series’ circuit. Evidence for myometrial arteriovenous anastomoses was presented in a paper by Schaaps et al.22 which included the finding of low resistance in UtAs after delivery, suggesting that the vascular network remains functional after placental expulsion. If these anastomoses are able to adequately ‘buffer’ some of the back pressure from the decidua–placental interface, the pathology would need to be severe before changes were seen in the UtA waveform. This would explain the more reliable prediction for the severe end of the spectrum23.

We acknowledge a number of limitations to the study. We could not measure all jets from the basal plate in every participant. This was attempted in three women before the study was undertaken, and it took 50–70 min per participant. Twenty-nine jets were measured in one woman (30 and 32 jets, respectively, in the other two women). The results demonstrated a difference of < 0.06 between maximum and minimum PI and RI values. Therefore, measuring three jets is likely to be representative of all jets. This maximized the number of women who could be scanned in the amount of time available, while minimizing color Doppler exposure. Furthermore, consistency was demonstrated between two operators independently measuring what they regarded as the ‘most prominent’ jets. While the numbers used to assess both intra- and interobserver variability were small, they were sufficient to demonstrate that the method is both precise and reproducible. The gestational age and position of placenta alter the ease with which measurements can be made. It can be particularly difficult to measure jets in lateral placentae at early gestational ages. Another limitation is that we excluded some women where maternal pathology might have been important in the pathogenesis of FGR, e.g. obese women; however, we felt that our threshold, at BMI of 35 kg/m2, was reasonable when examining a new ultrasound technique. Another limitation of the study was the small number of SGA pregnancies. We acknowledge that there is a 6% chance of the null hypothesis being true, but as this was a small pilot study, this level of significance indicates that further investigation is warranted.

One strength of the study is that we used customized birth weight to define the AGA newborns. There is increasing evidence that this allows better differentiation of clinically relevant small newborns at risk of developing adverse outcomes24. The use of customized centiles has also been shown to improve significantly the correlation of UtA-RI and birth weight25.

Manual measurement of jets from the mouth of the spiral arteries is fairly time-consuming and thus is likely to remain a research tool. However, it has enabled in vivo examination of the site of vascular pathology believed to underpin impaired placentation. Our finding that characteristics of blood flow at the uteroplacental interface may be different in pregnancies which result in SGA newborns suggests that this site may hold promise for future research into fetal growth restriction.

Acknowledgements

We thank the staff of the Ultrasound Department at the Women's Centre, John Radcliffe Hospital, especially Mrs K. Blissett, for all their help in recruiting the study participants. The Oxford Partnership Comprehensive Biomedical Research Centre supports S.C., G.S., J.B. and A.P. with funding from the Department of Health NIHR Biomedical Research Centres funding scheme.

Ancillary