The clinical use of umbilical venous blood flow was first suggested in the early 1980s1, 2, when reduced flow was demonstrated in growth-restricted fetuses. Later studies3–5 stated that, though feasible, umbilical venous blood flow calculation had little clinical potential because of measurement inaccuracies and low reproducibility. Much clinical research then shifted focus to the emerging umbilical artery Doppler, which was to become the benchmark for clinical management of growth-restricted fetuses. More recently, the advent of high-technology ultrasound and pulsed Doppler has brought umbilical venous blood flow back onto the research agenda of several groups. Pioneering studies found a correlation between placental mass and umbilical venous blood flow6 and a reduction in blood flow was demonstrated to occur weeks before significant changes were evident in the umbilical artery Doppler waveform7. Hence, umbilical venous blood flow could be considered a more direct and physiological measurement of vascular placental function than are umbilical artery Doppler indices, representing indirectly the quantity of oxygen and nutrients reaching the fetus. Nevertheless, there are concerns regarding its accuracy and reproducibility, and no clear methodological recommendations have been issued. Clinical research on this issue has been advancing slowly and there are doubts as to whether inconsistent findings could be attributable to methodological differences. This paper aims to review the methodological aspects of umbilical venous flow measurement.
Accuracy refers to the degree of veracity of a measurement. In-vivo models have been used for assessing the accuracy of umbilical venous blood flow measured by Doppler. Schmidt et al.8 reported in fetal lambs a slight overestimation of umbilical vein flow volume compared with radionuclide-labeled microspheres and electromagnetic techniques, but good correlation overall (r = 0.91). Galan et al.6 demonstrated, also in an ovine model, that umbilical venous blood flow volume determined by triplex-mode ultrasonography differed by less than 1% from the true flow measurement obtained by the steady-state diffusion technique. Konje et al.9 measured umbilical venous blood flow by ultrasonic transit time flowmetry in human fetuses during Cesarean section and compared their results with previously reported values10, 11 obtained by Doppler, concluding that values were ‘comparable’ when the umbilical vessel diameter is less than 4 mm. Thus, Doppler measurement of umbilical venous blood flow has been found to be accurate when compared with several gold standards for in-vivo flow calculation.
Flow (Q) is defined as the volume of fluid that passes a given cross-sectional area (CSA) per time unit (t). It is directly proportional to the spatial mean velocity of the flow and to the area of the vessel: Q(t) = Vmean(t) × CSA(t). Even small errors in volume flow components, the mean velocity and the vessel area, result in disproportionately large errors in the calculation of volume flow12. Therefore, technique standardization is of paramount importance when measuring any blood flow.
Venous cross-sectional area
The umbilical vein is the longest venous vessel in the human fetus and it has been demonstrated in vivo that, at its free-floating portion, its diameter decreases progressively from the fetus to the placenta13. Both intra-abdominal1, 4, 10, 14–24 and free-floating21, 25–29 portions of the umbilical vein have been used to estimate vessel area. When a free-floating portion of the umbilical vein is used, the area is calculated directly26, 28, 29 or from the vessel diameter21, 25, 27 (CSA = π × (diameter/2)2). Not surprisingly, normal ultrasound values differ beyond the expected biological variability according to the technique used for estimation: at 30 weeks the mean venous cross-sectional area is 22.6 mm2 at the intra-abdominal portion24, while at the free-floating portion it is 32.2 mm2 using diameter calculation27 and 40 mm2 using direct area calculation28. These methodological differences may account for inconsistent clinical findings: whilst for some groups28 the reduction in umbilical venous blood flow in growth-restricted fetuses was attributed mainly to decreased venous area, for others21 this was attributed principally to a reduction in the mean blood velocity, with no change in the area of the vessel.
Transverse and longitudinal sections of the same portion of the vein, with angles as close as possible to 90° and 0° with respect to the direction of flow, are required for cross-sectional area and Doppler mean velocity calculations, respectively. It could be argued that these requirements are more easily fulfilled in free-floating than in intra-abdominal portions of the umbilical cord. Validation studies in animal models by means of modern ultrasound and Doppler technology and under experimental conditions, i.e. anesthetic and surgery, which allow reasonable extrapolation to clinical use, have demonstrated accurate venous blood flow measurements by estimating the vessel's cross-sectional area from perpendicular views of longitudinal sections of free-floating portions of the cord6, 21. Therefore, this method constitutes a strong candidate for standard methodological recommendations. The vessel's diameter should be measured where the cross-sectional perimeter of the vein has a circular shape. Since the radius (diameter/2) is squared for cross-sectional area calculation, a small error in the diameter measurement results in a disproportionately large error in the cross-sectional area calculation. Therefore, high-resolution digital ultrasound equipment and adequate image zooming (occupying at least 30% of the screen), with inner-to-inner caliper placement to the nearest one tenth of a millimeter, are required for accurate measurements. In addition, there is clear experimental evidence that the accuracy of diameter measurements improves by averaging between three and five separate measurements30.
When the Doppler beam insonates the vessel at a defined point in time, the blood velocities (i.e. the shift of frequencies) are displayed in a format similar to that of a histogram, with the occurrence of frequencies encoded by brightness. When this is performed continuously, the addition of the histograms creates the Doppler spectrum. This process, whereby echo-strength values are converted into digital values, is highly software-dependent and is not under the direct control of the machine operator. For each point in time, only a discrete random set of frequencies is calculated and, therefore, only an approximation of the actual velocity distribution is obtained31. Mean velocity is represented by the intensity-weighted Doppler shift mean frequency for each unit of time, i.e. the ‘intensity-weighted mean velocity’ (IWMV). If the Doppler beam is sufficiently wide to produce a uniform insonation of the entire cross-sectional area of the vessel, the shift frequencies will be approximately proportional to the average spatial velocity in the vessel. In other words, it will represent the whole spectrum of the different velocities within the blood flow. However, in many situations, the Doppler beam does not cover the entire vessel. The Doppler sample volume is limited laterally by the Doppler beam dimensions (0.5–2 mm) and axially by the electronic gating of the received signal, which in most equipment can be set by the operator within a range of 1–15 mm. For correct mean velocity estimation from the IWMV, the sample volume should extend laterally and longitudinally to include the complete cross section of the vessel. The mean velocity estimated from the IWMV is only accurate if the transmission and reception profiles are uniform throughout the vessel's cross section32. The size of the Doppler gate affects the calculation of mean velocity from the IWMV by a process known as ‘spectral broadening’. When the spectral broadening is narrow, the resulting mean is closer to the maximum velocity33, overestimating the mean velocity. In addition, the high-pass filters can alter the number of signals that are evaluated. A high high-pass filter may exclude signals from slow blood, thereby also overestimating the mean velocity34. On the other hand, lower high-pass filters result in unstable measurements of the mean velocity because of the noise of vessel wall movements and echoes from the neighboring blood vessels35. Generally speaking, the effects of the high-pass filters on the IWMV are unpredictable, since they are related to the type of transducer, the flow velocity, the insonation angle and other instrument settings36.
Alternatively, a straightforward estimation of mean velocity can be made if the relationship between the mean and the maximum velocities is known, represented by the spatial velocity distribution coefficient. This value describes how the velocities are distributed across the cross-sectional diameter of the vessel. In a perfect parabolic flow profile, i.e. a laminar flow in a circular vessel, the maximum velocity is in the geometric center of the vessel and this coefficient is 0.5. Therefore, under the assumption of parabolic flow, the mean velocity can be estimated as being half the maximum velocity (0.5 × Vmax). This method is less sensitive to the overestimation phenomenon that results from filtering off low velocities and incomplete covering of the entire cross-section of the vessel, as long as the central axis of the vessel is included within the Doppler sample volume. Figure 1 depicts the reference ranges for flow, normalized for fetal weight, reported from 1990 onwards20, 22–28, 37, when it can be assumed that Doppler technology was comparable with current standards. It clearly shows that there is a trend to find higher values of flow when mean velocities are estimated from the IWMV compared with from the mean and maximal velocity20, 23, 25, 26, 37. Accordingly, Gerada et al.29, summarizing the findings of 20 published studies on umbilical venous flow, also reported that studies in which mean velocities were estimated from the IWMV show higher blood flow values at 30 weeks. However, estimating the mean velocity from Vmax cannot be considered error-free either. First, it assumes implicitly a steady blood flow, but under abnormal conditions, as in severe growth restriction or hydrops, heart-synchronous pulsations may be found in the umbilical venous flow. In fact, even a proportion of normal fetuses show a pulsatile umbilical venous flow pattern in the third trimester5. In these cases, only by estimating IWMV could umbilical venous blood flow be quantified accurately. Second, both the helicoidal arrangement of the vein and blood viscosity contribute to a departure from a perfect parabolic flow through the umbilical cord, which theoretically will lead to an underestimation of the mean velocity. Using spatial-velocity two-dimensional (2D) analysis, Pennati et al.38 calculated an underestimation of 18% of the mean velocity at a free-floating portion of the cord, and found this effect to be even more pronounced (32% underestimation) if the Doppler measurement was performed closer to the placenta. Thus, the velocity profile appears to become increasingly parabolic the closer it is to the fetus rather than the placenta, and it is possible that it would eventually become parabolic at the intra-abdominal site. In-vivo measurements of the spatial-velocity profile shape at this site have not been described, so we cannot make direct assumptions about this portion of the cord. However, Acharya et al.24 recently demonstrated that mean values of normalized flow showed little difference when calculated by 0.5 × Vmax and by IWMV at this site, suggesting that the spatial velocity profile in the intra-abdominal part of the umbilical vein is parabolic. As shown in Figure 1, recent studies investigating a free loop25–28 did not report flow values systematically different from those found by studies investigating flow in an intra-abdominal portion of the cord20, 22–24, 37. Although further investigation is required to examine the velocity profile through the cord and so to establish the true relationship between maximum and mean velocities in different parts of the umbilical cord, measurement errors yielded by the maximum velocity method seem to be more predictable and systematic than those associated with the IWMV method. In addition, since the former technique is less software-dependent, it is more likely to be more reproducible by different groups.
A new method for umbilical venous blood flow calculation was described recently, to measure Vmax by means of color Doppler cine-loop analysis29. As this method is less dependent on the insonation angle, it allows us to make fewer assumptions of the spatial-time velocity profile.
Whichever method is used for calculation of mean velocity, adherence to the basic principles of Doppler is of the utmost importance. A complete cross section of the vessel should be insonated uniformly. This is limited by the width of the Doppler beam, the size of the vessel and the angle of insonation. The effect of visual error, introduced inevitably by the operator when angle assignment is used, is minimized when the vessel is closer to the vertical on-screen. It has been demonstrated mathematically39 that at insonation angles of less than 30°, small angle errors (< 5°) have little impact on velocity estimation; beyond 30°, the impact increases exponentially. The Doppler sample gate can only be increased in length by increasing the electronic gating of the received signals32, and cannot be increased in width. This has implications when the vessel is large (such as the umbilical vein at > 32 weeks' gestation) and the longitudinal section of the vessel is parallel to the Doppler beam40. In this situation, to calculate IWMV, the Doppler gate should include the inner wall of one side of the vessel and the center of the vessel. To calculate mean velocity from Vmax, the Doppler gate should be located in the center of the vessel equidistant from both vessel walls. In all situations, it is recommended to maintain the Doppler gate between 2 and 2.5 mm.
Reproducibility refers to the closeness of the agreement between the results of measurements of the same parameter when carried out under changed conditions, such as time or operator. It is affected by biological variability and measurement accuracy of the parameter in question. In a pulsed Doppler phantom study41 aimed at analyzing the contributions of examiner and machine-related factors to the variability of Doppler flow measurements, it was shown that only 16% of the overall variability was observer- or equipment-related. The authors concluded that if the vessel could be located and examined properly, Doppler flow measurements would be reasonably repeatable. The question raised is whether the umbilical vein fulfills these requirements. Even though reproducibility should have been investigated before addressing validity issues, few of the studies on umbilical venous flow using modern technology have addressed this issue. In 10 fetuses, Barbera et al.27 reported interobserver coefficients of variation for the diameter, mean velocity from Vmax calculation, and blood flow of 2.9%, 7.9% and 12.7%, respectively. The intraobserver coefficients of variation for each parameter were 3.3%, 9.7% and 10.9%, respectively. In seven fetuses, Lees et al.26 reported an interobserver systematic error of − 0.8% with a coefficient of variation of 6.5% for umbilical venous area measurement, and for IWMV these values were 10.7% and 33%, respectively. In addition to the limitations resulting from the small sample size of both studies, the coefficient of variation is a measure of the degree of spread of the differences around the mean and is considered to be incomplete as a measure of reproducibility42. This method measures only absolute agreement, while the extent to which agreement is present against a background of the true variability in the observations is not assessed numerically. The significance of the concordance between the measurements is not evaluated against the diversity of the measurements routinely encountered in clinical practice. Using the intraclass correlation coefficient (ICC) to assess agreement overcomes all these limitations and is preferred as an index of concordance for such measurement variability. The following benchmarks are used for ICC characterization43: slight agreement (ICC, 0–0.2), fair agreement (0.21–0.4), moderate agreement (0.41–0.6), substantial agreement (0.61–0.8) and almost perfect agreement (0.81–1.0). No previous reports of reproducibility using this method in the umbilical vein have been published. In 63 consecutive singleton pregnancies between 24 and 42 weeks, two operators in our group performed blinded measurements of the umbilical vein (unpubl. data). Venous diameter and mean velocity were calculated according to standard recommendations27: inner-to-inner diameter from perpendicular views of a longitudinal section in a free-floating portion of the cord and, by rotating the image 180°, mean velocity estimation as 0.5 × Vmax. Satisfactory Doppler parameters were obtained successfully from all fetuses. The intraobserver ICC (95% CI) for diameter, mean velocity and blood flow were 0.7 (0.55–0.81), 0.59 (0.4–0.74) and 0.55 (0.35–0.7), respectively, and the interobserver ICCs were 0.65 (0.48–0.78), 0.46 (0.23–0.64) and 0.60 (0.4–0.74), respectively. These results highlight the fact that by adhering to methodological recommendations, umbilical venous blood flow calculation has moderate to good intra- and interobserver reproducibility, comparable to values reported for the umbilical artery indices of resistance44 and pulsatility45.
Adherence to sound methodological recommendations results in reasonably accurate measurements of umbilical venous blood flow. Umbilical venous diameter measurement in a longitudinal section of the vessel is reproducible and the resulting blood flow estimations have been validated in animal models by means of modern technology under experimental conditions that allow reasonable extrapolation to clinical use. Estimation of the mean velocity from the maximum velocity, rather than using the IWMV, is less software-dependent and more clearly defined, yielding estimates with more predictable and systematic errors. Further research is needed to define the blood velocity profile along the umbilical vein, which would result in more accurate estimates of blood flow at intra-abdominal and free-loop sites.