Associate Professor L.M.E. McCowan, Department of Obstetrics and Gynaecology, School of Medicine, University of Auckland, National Women’s Hospital, Private Bag 92 189, Auckland 3, New Zealand
Objective To assess the value of uterine artery Doppler ultrasound screening, when performed in a clinical setting, to predict complications of impaired uteroplacental blood flow in high-risk women.
Design A prospective audit.
Subjects A total of 116 pregnancies in 114 women at high risk of pre-eclampsia and/or small-for-gestational-age (SGA) babies attending a maternal–fetal medicine clinic at National Women’s Hospital, a tertiary referral hospital, Auckland, New Zealand.
Methods Uterine artery Doppler screening was performed as part of clinical practice between 22 and 24 weeks’ gestation. A resistance index (RI) was calculated from each uterine artery and the presence or absence of a notch was determined. An RI of > 0.58 was defined as abnormal and an RI of ≥ 0.7 was defined as very abnormal. The main outcome measures were: pre-eclampsia, SGA baby (birth weight < 10th centile), placental abruption, intrauterine death, ‘all’ and ‘severe’ outcomes.
Results Thirty-two (27.5%) women developed pre-eclampsia, 31 (26.7%) had SGA babies, 23 (20%) were delivered at < 34 weeks because of pregnancy complications, and there were three (2.6%) placental abruptions and three (2.6%) perinatal deaths. The sensitivity of any RI of > 0.58 for pre-eclampsia, SGA, ‘all’ outcomes and ‘severe’ outcome was 91%, 84%, 83% and 90%, respectively. The specificity of any RI of > 0.58 for these outcomes was 42%, 39%, 47% and 38%, respectively. The positive predictive value of any RI of > 0.58 for the same outcomes was 37%, 33%, 58% and 24%, respectively. Among women with both RI values of ≥ 0.7, 58%, 67%, 85% and 58% developed pre-eclampsia, SGA, ‘all’ and ‘severe’ outcomes, respectively. In women with bilateral notches, 47%, 53%, 76% and 65% developed the respective outcomes. Women with both RI values of ≥ 0.7 and women with bilateral notches had relative risks of 11.1 (95% CI 2.6–46.4) and 12.7 (95% CI 4.0–40.4) for developing severe outcome, respectively. Only 5% of women with both RI values of < 0.58 developed a severe outcome.
Conclusion In high-risk women, uterine artery Doppler waveform analysis performed best in the prediction of severe adverse outcome and was better than clinical risk assessment in the prediction of pre-eclampsia and SGA babies. Further studies are necessary to determine how information from uterine artery Doppler studies should modify current practice in high-risk women.
The role of uterine artery color Doppler waveform analysis, as a screening test in women at high risk of developing complications of impaired uteroplacental blood flow, has not been well evaluated. Abnormal uterine artery Doppler waveforms reflect increased impedance in the uterine circulation 1. The increased impedance is thought to occur secondarily to failure of trophoblastic invasion of the muscular spiral arteries2,3. The few screening studies of women at high risk of pre-eclampsia and small-for-gestational-age (SGA) babies are limited by small numbers, inconsistent methods, variable gestation at screening and differing definitions of a high-risk population 4–10.
Screening the unselected, low-risk population for pre-eclampsia using uterine artery Doppler waveform measurements has limited value, with positive predictive values between 4 and 20% 11–13. The low positive predictive values largely reflect the low prevalence of complications in unselected populations.
The optimal method for defining an abnormal uterine artery flow velocity waveform is uncertain. In unselected women, the presence of a diastolic notch is reported to perform better in predicting pre-eclampsia than an elevated resistance index (RI) or systolic/diastolic ratio, with positive predictive values up to 31%14,15. The significance of notching at 22–24 weeks in high-risk women is still uncertain. Notch assessment is subjective and interobserver measures of agreement have been determined in only a few studies12,16.
We hypothesized that, in high-risk women, Doppler screening of the uterine circulation would double the positive predictive value for estimating either pre-eclampsia or an SGA baby, compared to clinical risk assessment or prevalence. Also, we hypothesized that the presence of a diastolic notch would be a better predictor of pre-eclampsia or an SGA baby than the RI. A secondary aim was to assess the agreement between observers for notch determination in high-risk women.
Materials and Methods
This prospective cohort study was conducted in a tertiary referral center (National Women’s Hospital) over a 3-year period (1994–96). Women considered at high risk of complications resulting from poor uteroplacental function were studied. Women attending a maternal–fetal medicine clinic and who had the following conditions were included:
(1) Essential hypertension;
(2) Secondary hypertension;
(3) Pre-existing renal disease;
(4) Systemic lupus erythematosus;
(5) Antiphospholipid syndrome;
(6) Previous recurrent pre-eclampsia;
(7) Previous early-onset pre-eclampsia requiring delivery at ≤ 32 weeks;
(8) Previous placental abruption.
Multiple pregnancies and pregnancies with recognized fetal abnormalities were excluded.
Uterine artery color Doppler ultrasound was performed as part of routine clinical care in the maternal–fetal medicine clinic for women with pregnancies at high risk of pre-eclampsia and/or an SGA baby. Doppler studies were performed between 22 and 24 weeks’ gestation. Gestational age was determined by menstrual history and confirmed in nearly all cases by an early ultrasound scan. All Doppler examinations were performed in the Ultrasound Department at National Women’s Hospital by experienced ultrasonographers. Recordings were made using a Toshiba 270 machine (Toshiba Medical Systems, Tokyo, Japan), or Diasonics Masters Series (Diasonics, California, USA) real-time, color Doppler ultrasound system. A 3.75-MHz convex array or 3.5-MHz phased array sector transducer was used with the Toshiba machine, and a 5-MHz phased array sector transducer with the Diasonics machine. The high-pass filter was set at 100 Hz or lower. All ultrasound examinations were performed in the semirecumbent position and the uterine artery was identified, as previously reported 13. Three to five consecutive waveforms from each artery were obtained and the images frozen, and the RI was calculated.
Ultrasound films were reported by an obstetrician or radiologist. The results were available to the clinicians responsible for each patient. Timing of delivery and initiation of aspirin therapy were not influenced by the uterine artery Doppler results.
The RI (for either side) was defined as abnormal if it was > 0.584,17 or very abnormal if it was ≥ 0.7, which represented the 95th centile cut-off in a low-risk population 13. Hard copies of Doppler waveforms were independently reviewed by two observers (MC and LMcC) for the presence of notches. The observers were blinded to the clinical details. The waveforms were graded as 1–4 (1 = normal waveform; 2 = probably normal waveform; 3 = mild notch; 4 = severe notch). The grades were then grouped according to absence (grade 1 or 2) or presence (grade 3 or 4) of a notch and Cohen’s kappa statistic was calculated as a measure of interobserver agreement. In cases where there was major disagreement, for example grade 1 vs. grade 3 or 4, the films were reviewed by a third observer (RN) and a final classification was made.
The definition of ‘any’ notch included pregnancies with either unilateral or bilateral notching. Similarly, ‘any’ RI included either unilateral or bilateral RI values.
In cases in which it was not possible to obtain Doppler waveforms from both sides, results were reported from the single side. Clinical data and pregnancy outcome information on all women were retrieved from the hospital records.
The major end-points were:
(1) Pre-eclampsia defined as gestational hypertension and proteinuria (≥ 0.3 g/24 h or at least ‘2+’ proteinuria on dipstick, in the absence of urinary tract infection). Gestational hypertension was defined as a blood pressure of at least 140/90 mmHg with an increase of at least 15 mmHg in diastolic blood pressure taken on two occasions more than 4 h apart after the 20th week of pregnancy. Superimposed pre-eclampsia was defined as a systolic blood pressure of > 140 mmHg with an increase in systolic blood pressure of ≥ 30 mmHg and/or a diastolic blood pressure of > 90 mmHg with a rise in diastolic blood pressure of ≥ 15 mmHg, combined with either new proteinuria (≥ 0.3 g/24 h or ≥ ‘2+’ on dipstick) or a doubling of the 24-h urinary protein excretion.
(2) SGA baby, defined as having a birth weight less than the 10th centile 18.
(3) Intrauterine death after 20 completed weeks of gestation.
(4) Placental abruption, defined as retroplacental clot at delivery.
(5) All outcomes, defined as the development of pre-eclampsia and/or an SGA baby and/or intrauterine death and/or abruption at any gestation.
(6) Severe outcome, defined as pre-eclampsia and/or an SGA baby necessitating delivery before 34 weeks and/or placental abruption and/or intrauterine death at any gestation.
Statistical analysis was performed using the SAS system for Windows (release 6.11) statistical package. The χ2 test or Fisher’s exact two-tailed test was used to compare the outcome of pregnancies with a normal screening test and those with abnormal Doppler studies. The sensitivity, specificity, positive and negative predictive values as well as relative risk were calculated for pre-eclampsia, SGA baby, placental abruption, ‘all’ and ‘severe’ outcomes. Relative risk was calculated between groups with abnormal Doppler studies and a group of 38 pregnancies (33%) with both RI values of ≤ 0.58 who were defined as normal. Only one pregnancy in this normal group had a unilateral notch, and this pregnancy was uncomplicated. Relative risks for the presence of any notching or bilateral notches were calculated by comparing this group to those with no notching.
For the purposes of performing a power calculation we estimated that the prevalence of pre-eclampsia and/or an SGA baby would be 26% 19. The test would be considered clinically useful if, in the presence of an abnormal test, the risk of these complications was doubled to 50%. A false-negative rate of 5% was considered clinically acceptable. A sample size of 100 was required to show this increase to 50% (95% confidence interval (CI) 35–65%).
A total of 116 pregnancies in 114 women were studied over 3 years. Fifty-one (44%) had essential hypertension, 18 (15.5%) secondary hypertension, 17 (14.6%) glomerulonephritis, 10 (8.6%) reflux nephropathy, one (0.8%) renal transplant, 13 (11.2%) systemic lupus erythematosus, five (4.3%) antiphospholipid syndrome, 24 (20.6%) previous recurrent pre-eclampsia, 25 (21.6%) previous severe pre-eclampsia and delivery at ≤ 32 weeks, and 10 (8.5%) previous placental abruption. Seventy pregnancies (60%) had more than one entry criterion.
Forty-one per cent of the women were European, 26% were Maori, 21% were Polynesian and 11% were of other ethnic origin. The median age at delivery was 31 (range 19–43) years. Thirty-one (26.7%) were nulliparous and 18% smoked during pregnancy.
Seventy-eight pregnancies (67.2%) had either an RI of > 0.58 and 12 (10.3%) had bilateral RI values of ≥ 0.7. A total of 196 right and left hard copies of Doppler waveforms were available for analysis. Thirty-nine (40%) pregnancies had a notch on any side, 22 (23%) had unilateral and 17 (17%) had bilateral notches. In 30 (15%) waveforms there was major disagreement between initial observers for notch determination. Two-thirds of this subgroup (20/30) were finally graded as ‘no notch’ and one-third (10/30) as ‘notch present’. Cohen’s kappa statistic for interobserver agreement for notch identification was 0.81 (95% CI 0.48–1).
The proportion of pregnancies with waveform notches according to the uterine artery RI value on any or both sides is shown in Figure 1. As the RI value increased, the proportion of pregnancies with notches also increased (P < 0.001).
Ninety-five per cent of the pregnancies with early-onset pre-eclampsia and delivery at < 34 weeks had any RI of > 0.58 (data not shown). However, the test did not discriminate these pregnancies from the total group, as the positive predictive value was 25%, whereas the prevalence of early-onset pre-eclampsia was 18%. Both RI values of ≥ 0.7 detected early-onset pre-eclampsia with sensitivity, specificity, positive and negative predictive values of 41%, 96%, 70% and 88%, respectively (RR 5.7; 95% CI 2.8–11.5). Bilateral notching had sensitivity, specificity, positive and negative predictive values of 62%, 89%, 47% and 94%, respectively (RR 9.3; CI 95% 2.8–31.1), for predicting early onset pre-eclampsia.
Gestational hypertension developed in 25 (21.5%) pregnancies. The test performed poorly as a predictor of gestational hypertension, and no test variation achieved statistical significance.
Of the 31 SGA babies, 11 were associated with preeclampsia and five with gestational hypertension. The ability of uterine artery Doppler waveforms to predict SGA babies is shown in Table 3.
Table 3. The prediction of small-for-gestational-age babies by uterine artery Doppler screening
The test characteristics for predicting all outcomes are shown in Table 4. Eighty-three per cent of pregnancies with an abnormal outcome had any RI of > 0.58, but the positive predictive value (58%) was not significantly greater than the prevalence of ‘all’ outcomes (47%, P = 0.40). More severely abnormal waveforms improved the positive predictive value, but with reduced sensitivity.
Table 4. The prediction of ‘all’ outcomes by uterine artery Doppler screening
The prevalence of a ‘severe’ outcome was 18.1%. Table 5 outlines the test characteristics for predicting severe outcome. Nearly two-thirds of women with both RI values of ≥ 0.7 (58%) or both sides notched (65%) developed a severe outcome. The negative predictive value of all test variables was high (86–95%).
Table 5. The prediction of ‘severe’ outcome by uterine artery Doppler screening
There were two intrauterine deaths and one neonatal death. The clinical and Doppler details of these cases are shown in Table 6. Placental abruption occurred in three (2.5%) pregnancies. Two had abnormal RI values with notches on both sides. One resulted in an intrauterine death at 23 weeks and the other woman with pre-eclampsia and fetal growth restriction delivered at 27 weeks. The third woman had normal RI values and the abruption was an incidental finding at Cesarean section performed at 37 weeks.
Table 6. Details of perinatal deaths
Gestation at uterine Doppler scanning
Gestation at delivery
RI = resistance index; IUD = intrauterine death; NND = neonatal death; NA = not available.
In six women the Doppler waveform was obtainable on only one side. Four of the six had a normal RI on the obtainable side; all delivered at term, one with an SGA baby. The remaining two had abnormal RI values on the obtainable side (0.71, 0.88) and associated notching. Both developed pre-eclampsia and were delivered at 27 weeks, one with an SGA baby (birth weight 520 g).
Forty-three (37%) pregnancies were treated with aspirin (75–150 mg/day) prior to 20 weeks. There was no significant difference in the RI values or presence of notching between those women treated with prophylactic aspirin compared to those who did not receive aspirin. In pregnancies with abnormal Doppler studies, there was no significant difference in the rate of pre-eclampsia or SGA babies between those who were treated with aspirin compared with those who were not (data not shown).
This is one of the first clinical studies of uterine artery color Doppler ultrasound used as a screening test in a population with a confirmed high rate of pre-eclampsia, SGA babies and resultant delivery before 34 weeks. In the few previous studies of high-risk women the ultrasound scans were performed by a small number of researchers, rather than routinely by ultrasonographers working in a clinical department. In this study a high proportion of women had an abnormal Doppler study. Uterine Doppler waveform analysis performed modestly in the prediction of pre-eclampsia. However, the positive (37–60%) and negative (73–92%) predictive values for pre-eclampsia based on the various Doppler indices performed better than clinical risk assessment based on the prevalence of pre-eclampsia (27.5%) in this population. As expected, the positive predictive values were higher than those obtained in low-risk populations. This in part reflects the higher prevalence of pre-eclampsia and is consistent with other studies of high-risk women9,10. Both these former studies used a single researcher to perform the Doppler studies and were limited by smaller numbers. We found comparable results using multiple operators in a clinical setting.
Of the babies, 27% were SGA and the test performance for predicting SGA babies was similar to that for pre-eclampsia, with positive predictive values of 33–70%. Only the very abnormal Doppler waveforms significantly improved clinical risk prediction.
The women in this study were at high risk of complications resulting from poor uteroplacental blood flow. They were particularly concerned about the likelihood of a severe complication, especially if it would result in fetal death or delivery of a very premature baby. The higher RI cut-off value (≥ 0.7) or the presence of notching enabled the Doppler test to identify these women. In particular, 58% of women with bilateral RI values of ≥ 0.7 and 65% with bilateral notching developed a severe outcome. This was significantly better than the pretest likelihood of a severe outcome (18.1%). Further, only 5% of women with a normal uterine Doppler study (both RI values of < 0.58) developed a severe complication. This information may be useful when counselling women about the probability of severe adverse outcome.
This study is the first to report on the significance of notching in a large series of high-risk women. A secondary aim of this study was to assess the agreement between observers for the presence of a notch in high-risk women. A Cohen’s kappa statistic of 0.81 indicates very good agreement between observers. When the RI was ≤ 0.58, notching was extremely uncommon. As the RI increased, the proportion of waveforms with notches increased. Ninety-one per cent of women with both RI values of ≥ 0.7 had bilaterally notched waveforms ( Figure 1). Consistent with our hypothesis, in this population the presence of notching improved the test performance compared to an RI cut-off value of > 0.58 alone. The positive predictive value of bilateral notching for a severe outcome (65% (95% CI 42–88%)) was not significantly different from that for bilateral RI values of ≥ 0.7% (58%); however, the sensitivity was better for bilateral notching (69% vs. 33%).
One-third of women in this study received low-dose aspirin from early pregnancy. This study was not designed to determine whether aspirin treatment could improve pregnancy outcome in high-risk women. However, no difference in the rates of pre-eclampsia or SGA babies was seen between women treated and those not treated with aspirin. This is consistent with the findings of a large randomized controlled trial 20 in which low-dose aspirin treatment did not improve pregnancy outcome in high-risk women. In a small study in which women with abnormal uterine Doppler findings were randomized to treatment with low-dose aspirin at 24 weeks, a reduction in severe pre-eclampsia was found in aspirin-treated women 21. Further studies are necessary before aspirin can be recommended as treatment in pregnancies with abnormal uterine Doppler findings. In the current study, no women commenced aspirin in response to or after their uterine Doppler examination.
The results of this study suggest that uterine artery Doppler may be a useful screening method for the prediction of severe adverse outcome in high-risk women. The test performed better than clinical risk assessment in the prediction of both pre-eclampsia and SGA babies. Although, at present, there is no effective intervention to alter the outcome in women with an abnormal Doppler study 21, the level of antenatal surveillance could be modified by the Doppler result. Further prospective and blinded studies will be necessary to determine whether this information should be integrated into the antenatal care of high-risk women.
The contributions of the following are acknowledged: Jeannie McDonald, Jenny Mitchell, Alistair Stewart and Vivienne Topping.