SEARCH

SEARCH BY CITATION

Abstract

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. METHODS
  5. RESULTS
  6. DISCUSSION
  7. References

Objective To evaluate the clinical usefulness of Doppler analysis of the uterine artery velocity waveform in the prediction of pre-eclampsia and its associated complications of intrauterine growth retardation and perinatal death.

Design Quantitative systematic review of observational diagnostic studies using online searching of the MEDLINE database coupled with scanning of the bibliographies of primary and review articles including known unpublished studies.

Material Twenty-seven studies involving 12,994 subjects stratified into population subgroups at low and high risk of developing pre-eclampsia and its complications.

Outcome measures The outcome measures studied were: 1. the development of pre-eclampsia; 2. intrauterine growth retardation; and 3. perinatal death. The main meta-analyses were the flow velocity waveform ratio ± diastolic notch derived by transabdominal Doppler ultrasound as the measurement parameter. The analyses were conducted using likelihood ratio as a measure of diagnostic accuracy. A likelihood ratio of 1 indicates that the test has no predictive value for the outcome. Prediction for the outcome event is considered conclusive with likelihood ratios of > 10 or < 0.1 for a positive and negative test result, respectively. Moderate prediction can be achieved with likelihood ratios of 5–10 and 0.1–0.2 whereas likelihood ratios values of 1–5 and 0.2–1 would generate only minimal prediction.

Results In the low risk population a positive test result, predicted pre-eclampsia with a pooled likelihood ratio of 6.4 (95% CI 5.7–7.1), while a negative test result had a pooled likelihood ratio of 0.7 (95% CI 0.6–0.8). For intrauterine growth retardation the pooled likelihood ratio was 3.6 (95% CI 3.2–4.0) for a positive test result and 0.8 (95% CI 0.8–0.9) for a negative test result. Using perinatal death as outcome measure, the pooled likelihood ratio was 1.8 (95% CI 1.2–2.9) for a positive test result and 0.9 (95% CI 0.8–1.1) for a negative test result. In the high risk population a positive test result predicted pre-eclampsia with a pooled likelihood ratio of 2.8 (95% CI 2.3–3.4), while a negative test had a likelihood ratio of 0.8 (95% CI 0.7–0.9). For intrauterine growth retardation the pooled likelihood ratio was 2.7 (95% CI 2.1–3.4) for a positive test result and 0.7 (95% CI 0.6–0.9) for a negative result. For perinatal death the pooled likelihood ratio was 4.0 (95% CI 2.4–6.6) for a positive test result and 0.6 (95% CI 0.4–0.9) for a negative result.

Conclusion Uterine artery Doppler flow velocity has limited diagnostic accuracy in predicting preeclampsia, intrauterine growth retardation and perinatal death.


INTRODUCTION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. METHODS
  5. RESULTS
  6. DISCUSSION
  7. References

Despite recent advances in antenatal care, pre-eclampsia has remained a major cause of maternal and perinatal morbidity and mortality. Data from the Confidential Enquiries into Maternal Deaths in the United Kingdom 1991–1993 confirm that hypertensive disorders of pregnancy is still the second most common cause of maternal mortality, accounting for 20/128 (15.5%) direct deaths1. Hypertension in pregnancy is also responsible for 18% of fetal (> 19 weeks of gestation) and infant mortality and 46% of infants born small for gestational age2. Early screening for pre- eclampsia may allow vigilant antenatal surveillance and appropriate timing of fetal delivery in order to avoid serious sequelae. Unfortunately, various haemodynamic and biochemical measures have been found to have limited accuracy as screening measures for this condition3,4.

Pre-eclampsia is characterised by an imbalance between prostacycline and thromboxane production5, as well as failure of the second wave trophoblastic invasion of the endometrio- myometrial vasculature. The result is abnormal uteroplacental blood flow, and this has lead to the idea of using Doppler assessment of uterine artery velocity waveforms as a method of screening for this antenatal complication6. An abnormal test result is represented by either an abnormal flow velocity ratio (systolic to diastolic velocity (S:D) ratio, diastolic to systolic velocity ratio (D:S) or resistance index) or the presence of an early diastolic notch. However, the wide variation in scanning techniques, Doppler measurement parameters and study protocols have resulted in conflicting results3. To address these concerns and to overcome the lack of precision in the individual studies, we decided to conduct a quantitative systematic review of the currently available literature in order to determine whether uterine artery Doppler ultrasound waveform analysis has any clinical value in the prediction of pre-eclampsia, intrauterine growth retardation and perinatal death. We also investigated whether any inconsistencies in the results between the various studies (heterogeneity) can be explained by differences in study quality, Doppler scanning techniques and gestational age at scanning.

METHODS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. METHODS
  5. RESULTS
  6. DISCUSSION
  7. References

The systematic review was based on a prospective protocol designed to address the above questions. Initially, two of us (A.G. and P.O.) independently conducted computerised MEDLINE searches (January 1966–January 1997) with the following search strategies: the first search was conducted with the term pregnancy in the title, abstract or medical subject heading (MeSH), combined with uterine or uteroplacental and Doppler as textwords and limited to human studies; the second search utilised the term pregnancy in Thesaurus, exploding it to include all subheadings, then combining it with the textwords uterine and Doppler and again limited to human studies. We used two independent electronic search strategies in order to maximise the number of published articles identified for the overview. The citation list in each search, containing only the titles, MeSH headings and abstracts (where available), was independently reviewed by two authors (A.G. and P.O.) who were blinded to information on the authors'names, their institutional affiliation, the publication language, the year of publication and the name of the journal. For inclusion in the overview, objective criteria were chosen for the study population, the diagnostic intervention and the outcome measure(s). Each citation was therefore deemed to be relevant or not relevant depending on whether it satisfied all the preset criteria. For both the searches, the complete manuscripts of all the citations considered relevant by the reviewer were then retrieved. We also supplemented the citation list from the electronic search with pertinent citations obtained from scanning of reference lists of all primary and review articles and reviews of recent journal issues not covered by the search. Any known unpublished studies were also included into the overview.

The eligibility of all the English language articles was independently assessed by two of us (P.C. and N.A.) by review of the entire manuscript under masked conditions (i.e. the names and affiliations of the authors, the date of publication, the journal's name and the sources of financial support were concealed in the manuscript). Any disagreement was resolved by consensus or arbitration by a third reviewer (A.G.). The review of articles published in non-English languages was performed by reviewers with medical training who comprehended the language.

In situations where there was suspicion of duplication in the reporting of data, the first authors of these articles were contacted to clarify the uncertainty. In the event that there was partial duplication of reporting of data, the article with the larger sample size was included in the overview.

The following criteria were used to select articles from the electronic database for full scrutiny of their entire manuscript:

  • 1
    Population: Pregnant women, both at low or high risk of developing hypertensive disorders and related complications.
  • 2
    Diagnostic intervention: Antenatal uterine artery Doppler flow velocimetry, conducted during second or third trimester of pregnancy including continuous wave and pulsed wave Doppler scanning, and colour flow mapping.
  • 3
    Outcome measures: Pre-eclampsia, intrauterine growth retardation and perinatal death.

Methodological quality criteria

Methodological quality of all the English language articles was assessed by two of us independently (P.C. and N.A.). The following criteria for methodological quality assessment were chosen based on the belief that they contribute to the internal validity of a diagnostic test study7,8:

  • 1
    Population: For the method of sampling of the study population, consecutive recruitment of eligible women was considered to be ideal. Convenience sampling, (i.e. arbitrary recruitment or nonconsecutive recruitment) was deemed to be inadequate. In the absence of any explicit information in the manuscript on the method of recruitment, the article was categorised as having unclearly reported population enrolment.
  • 2
    Diagnostic intervention: The description of the uterine artery Doppler velocity waveform analysis was considered to be ideal if the scanning method (i.e. whether continuous wave, pulsed wave, colour mapping, transabdominal or transvaginal scanning was used) was reported together with the gestational age at the time of Doppler scanning, the measurement parameter used and the cutoff level for an abnormal result. In the absence of any of the above information in the manuscript, then the diagnostic intervention was considered to be unclearly reported.
  • 3
    Outcome measures: Blinding of the outcome measure from the results of the Doppler velocimetry was considered ideal if it was clearly reported that the attending clinicians managing the subjects were kept unaware of the results of the Doppler findings. If the results of the Doppler recordings were divulged to the attending clinicians, then blinding was categorised to be inadequate. In the absence of any such reporting, blinding was considered to be unclearly reported. Information on the number of women recruited into the study and those whose outcome data were known was also sought from the manuscripts. Women whose outcomes were deleted from data analysis on the basis of fetal anomalies, multiple pregnancies, spontaneous abortion and scanning performed out with the specified gestational limits were considered to be legitimate exclusions. Withdrawal of women from the study, missing data and lack of outcome data due to delivery out with the hospital in which the study was conducted were categorised as lost to follow up. Follow up was considered to be ideal if > 90% of the women originally enrolled into the study without legitimate exclusions were included in the data analysis. If data from only 81%–90% of the recruited women were available for analysis, then it was classified as second best. Follow up was defined as inadequate when data from ≥ 80% of enrolled women were available for analysis.

Data extraction

Two authors (P.C. and N.A.) independently extracted information from each article in order to construct 2 × 2 tables of the diagnostic test result and pregnancy outcomes. Any disagreement was resolved by conference.

In order to reduce heterogeneity in the results, the target population of pregnant women was subdivided into those with low or high risk pregnancies. The assignment of the selected studies into these subgroups was performed by the authors of the original studies and it was confirmed independently by two of us (P.C. and N.A.). Studies which recruited an unselected obstetric population with normal pregnancies at the time of Doppler assessment were categorised as having low risk subjects. Pregnant women with advanced age (> 35 years), previous poor pregnancy outcome, vascular complications during previous pregnancies, previous small for gestational age babies, insulin-dependent or gestational diabetes mellitus, essential hypertension, previous severe pre-eclampsia, systemic lupus erythematosus, antiphospholipid syndrome, renal disease, unexplained elevated maternal serum α-fetoprotein were classified into the high risk population.

The diagnostic test result was also further stratified according to type of ultrasound scanning employed (transabdominal or transvaginal) and whether the measurement parameter was based on flow velocity waveform ratio ±diastolic notch or the presence or absence of a diastolic notch alone. The test result was considered to be abnormal if the diastolic notch was present in the velocity waveform of one or both uterine arteries.

Outcome data on pre-eclampsia, intrauterine growth retardation and perinatal death was recorded where available. We set a priori that the criteria for preeclampsia should include the presence of both hypertension and proteinuria. The definition of hypertension utilised by the selected studies included blood pressure ≥ 140/90 mmHg on two occasions which were more than 4 hours apart, a single reading of diastolic blood pressure ≥ 110 mmHg or a rise in blood pressure ≥ 30/15 mmHg on two occasions which were more than 4 hours apart. Significant proteinuria was considered to be present when more than 150–500 mg of protein was excreted per 24 hours or ≥ 1+ on labstix testing of the urine specimen. Intrauterine growth retardation was defined as infant birthweight < 10th centile for gestational age. Perinatal death included all intrauterine deaths beyond fetal viability and neonatal deaths during the first week of postnatal life.

Data analysis

To assess the reproducibility of study selection and methodological quality assessment, we evaluated the agreement between the two reviewers using percentage agreement and weighted kappa statistics9. Substantial reviewer agreement was considered to be present when the kappa levels were > 0.6510.

The main analysis was restricted to all studies in which 2 × 2 tables of the diagnostic test results and the relevant pregnancy outcomes could be extracted. The summary receiver-operator characteristic curve is still considered the preferred method of pooling dichotomous test result from primary studies8. Because there was considerable differences in the cutoff level for abnormality used in the subgroup of studies that utilised flow velocity waveform ratio ± diastolic notch as the measurement parameter, we initially tested for a high positive correlation (i.e. correlation coefficient > 0.6) between their true positive rates and the false negative rates using the Spearman correlation test to determine if a summary receiver-operator characteristic curve could be generated11. Since none of the relevant subgroup of studies generated a sufficiently high positive correlation, we then conducted our analyses using likelihood ratio as the measure of diagnostic attribute of a test.

Heterogeneity of results between different studies was formally assessed using the Breslow-Day test12 which compared the ratio of the odds of having the outcome of interest when the test result was positive to the odds of having the same outcome with a negative test result for each study. If significant heterogeneity was present, we explored for the potential sources of it by further stratifying the relevant studies according to variation in methodological quality (ideal versus inadequate or unclearly reported studies), method of Doppler analysis (continuous wave versus pulsed wave Doppler scanning) and gestational age of scanning (scans performed at 16–24 weeks of gestation versus > 24 weeks of gestation) for sensitivity analyses13.

Pooling of likelihood ratios was performed for predefined subgroups of study populations subject to the use of either uterine artery flow velocity waveform ratios ± diastolic notch or diastolic notch alone as the manner of determining test abnormality. For the purpose of meta-analysis, we weighted the log (likelihood ratio) from each study in inverse proportion to its variance in order to combine the likelihood ratios from each study. All the methods for data analysis used in this overview have been reported previously by us elsewhere14. The interpretation of likelihood ratios for positive and negative test results has been reported by Jaeschke et al.15. A likelihood ratio of 1 indicated that the test has no predictive value for the outcome of interest. In order for conclusive prediction of the outcome event of interest to be achieved, a likelihood ratio of > 10 or < 0.1 would be required for a positive and negative test result, respectively. Moderate prediction can be achieved with likelihood ratio values of 5–10 and 0.1–0.2 whereas likelihood ratios of 1–5 and 0.2–1 would generate only minimal prediction.

To demonstrate the practical application of the likelihood ratios generated, we calculated post-test probabilities for the various pregnancy outcomes by using Bayes'theorem. An estimate of the pre-test probability was obtained by calculating the prevalence of the outcome event in the population studied. The algorithm of equations used for calculating post-test probability and the 95% confidence interval around the point estimate have also previously been reported by us14. In the same manner, clinicians can also use the likelihood ratios generated from here to calculate the post-test probabilities of the relevant pregnancy outcomes based on the prevalence rates of their own practice population. The clinical usefulness of the test can therefore be assessed for different practice populations with the result of enhancing the external validity of the overview.

RESULTS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. METHODS
  5. RESULTS
  6. DISCUSSION
  7. References

The initial two independent MEDLINE searches by A.G. and P.O. yielded 501 and 479 citations, respectively. After blinded scrutiny of the titles and abstracts of these articles by the two independent reviewers, there were 46 articles which both reviewers thought were relevant and four articles which at least one of the reviewers considered to be relevant. All these four articles were identified by both searches. The complete manuscripts of all these articles considered relevant by any of the reviewers were then retrieved for further assessment. Of the 50 articles, 45 were published in English, two in German and one each in French, Portuguese and Japanese. Twenty-two further articles (20 published in English, one in French and one known unpublished study) were identified through examining the bibliography of known primary and review articles or obtained from personal files maintained through the constant searching of recent issues of journals addressing obstetric issues.

After independent review of the English manuscripts under masked conditions, 30 articles6,16–44 were considered to be eligible to be included in the overview. The agreement concerning eligibility was 92% (kappa 0.85). The five instances of disagreement were due to an oversight of one of the reviewers with regards to appropriateness of the study population (one article) and relevance of the reported outcome data (four articles). The disagreement was easily resolved by conference. The reasons for the remaining 42 articles to be excluded were inappropriate study design (11 articles), inappropriate study population (nine articles), lack of clarity of the cutoff level for abnormal test result (one article) and outcome measure (one unpublished article), and lack of relevant outcome data reported (twenty articles). The bibliography of the articles which were considered to be ineligible for inclusion into the systematic review can be obtained from the authors.

Following the assessment of eligibility, there were four articles6,18,36,37 in which partial duplication of data reporting was suspected. Two of these6,18 were excluded from the overview when the first authors of these articles confirmed that the data had been reported elsewhere. Thus, 28 articles16–17,19–44 were included into the overview; pre-eclampsia was reported in 22 studies (79%)16–17,19–20,22–25,27,28–33,38,40–42,44, intrauterine growth retardation was reported in 24 studies (86%)16–17,21–23,25,27–44 and eight studies (28%)17,23,25–26,28,31,36–37 reported on perinatal death (Table 1). Of the eight articles that reported on perinatal death, the gestational age for fetal viability was defined as 24 weeks of gestation in one study36, 18–22 weeks of gestation in another study17 and undefined in the remaining studies23,25–26,28,31,37. Sixteen of these articles16–17,20–21,25,27,33–37,39,41–44 provided data on low risk obstetric populations whereas twelve articles19,22–24,26,28–32,38,40 reported on antenatal women who were considered to be at high risk of developing antenatal complications

Table 1.  Description of studies used in systematic overview of uterine artery Doppler ultrasound. CW = continuous wave Doppler; PW = pulsed wave Doppler; CM = color mapping; RI = resistance index; S/D = systolic/diastolic ratio; D/S diastolic/systolic ratio; PET = pre-eclampsia; IUGR = intrauterine growth retardation; PD = perinatal death.
 PopulationUterine artery Doppler ultrasoundOutcome
StudyEnrolment methodDescription of scanning methodGestational age at scan (week)Measurement parameter(s) used in overviewCutoff level for waveform ratioBlinding of of resultsOutcome measures in overviewFollow-up (%)
  1. *Transvaginal ultrasound scan used, hence study was excluded from meta-analyses (see text for description). Same study but different measurement parameters reported in these articles.

Low risk population        
Campbell et al. (1986)25ConsecutivePW16–18RI>2 SDYesPET + IUGR + PD81–90
Schulman et al. (1987)34Unreportedcw20–40S/D>2SDYesIUGR> 90
Hanretty et al. (1989)39Consecutivec w26–30S/D> 95th centileYesIUGR81–90
Newnham et al. (1990)35Unreportedc w18S/D> 95th centileYesIUGR> 90
Steel et al. (1990)37Unreportedc w16–2224RI> 0.58YesPET + IUGR + PD> 90
Bewley et al. (1991)16Consecutivec w16–24RI> 95th centileYesPET + IUGR> 90
Bower et al. (1993)17ConsecutiveCW+PW+CM18–22,24RI ± notches> 95th centileYesPET + IUGR + PD81–90
Bower et al. (1993)20ConsecutiveCW+PW+CM18–22,24Notches aloneYesPET81–90
Valensise et al. (1993)33UnreportedPW+CM22,24RI>2 SDUnreportedPET + IUGR> 90
North et al. (1994)27UnreportedPW+CM19–24,22–24RI> 90th centileYesPET + IUGR> 90
Todros et al. (1995)21UnreportedCW+PW19–24SID> 2.7Y e sIUGR> 90
Harrington et al. (1996)36UnreportedCW+PW+CM18–21,24RI ± notches> 95th centileUnreportedPET + IUGR + PD> 80
Harrington et al. (1997)41UnreportedPW + CM*12–16Notches aloneUnreportedPET + IUGR> 90
Frusca et al. (1997)42UnreportedCW+PW+CM20,24RI> 0.58YesPET + IUGR> 90
Liberati et al. (1997)43ArbitraryPW+CM22–24RI + notches> 90th centileUnreportedIUGR> 90
Irion et al. (1998)44UnreportedPW+CM18–19,2–27S/D; notches> 90th centileYesPET + IUGR> 90
High risk popdation        
Trudinger et al. (1985)32UnreportedcwUnreportedD/S< 5th centileUnreportedPET + IUGR> 90
Jacobson et al. (1990)40ArbitraryPW20,24RI>2 SDYesPET + IUGR> 90
Aristidou et al. (1990)26Unreportedc w18–22,24–26RI ± notches> 97.5th GentileYesPD> 90
Pattinson et al. (1991)19Unreportedc w16–28. > 28RI> 95th centileYesPET> 90
Benifla et al. (1992)38UnreportedPW20–30D/S< 2 SDUnreportedPET + IUGR> 90
Guzman et al. (1992)31Unreportedc wUnreportedS/D ± notches> 2.6NoPET + IUGR + PD> 90
Caruso et al. (1993)23ConsecutivePW+CM18–24RI> 90th centileUnreportedPET + IUGR + PD8 1.90
Haddad et al. (1993)22UnreportedcwUnreportedD/S ± notches< 10th centileUnreportedPET + IUGR> 90
Ferrier et al. (1994)29ArbitraryPW+CM19–24RI> 90th centileYesPET + IUGR> 90
Chan et al. (1995)24Unreportedc w20,28, 36RI> 95th centileYesPET>90
Haddad et al. (1995)30Consecutivecw24,30D/S ± notches< 10th centileUnreportedPET + IUGR> 90
Konchak et al. (1995)28UnreportedPW+CMUnreportedRI; notches> 95th centileUnreportedPET + IUGR + PD> 90

Study quality

The reviewer agreement regarding the various components of study quality varied between 68% to 100%; kappa values were 0.48 for enrolment of study population; 0.85 for nature of study population, 0.94 for description of scanning method, 1.0 for description of measurement parameter used, 0.93 for description of cutoff level for abnormal flow waveform ratios, the presence or absence of a diastolic notch for test result and blinding of test results and 0.68 for follow up.

On reviewing the entire manuscripts of the 28 studies selected for overview, there were two articles17,20 that reported on the same study population but different measurement parameters (i.e. flow velocity waveform ratio ± diastolic notch17 and the presence or absence of a diastolic notch alone20) were used to report the data. To avoid duplication, we have therefore excluded the latter article20 in the assessment of methodological qualities (Table 2).

Table 2.  Methodological quality of uterine artery Doppler ultrasound studies included in overview. Values are given as dn (%).
Quality criteriaLow risk* populationHigh risk populationTOTAL
  1. *Two articles17–20 reported on the same population with different Doppler measurement parameters and hence the assessment of methodological quality was not duplicated on them.

Population   
Enrolment   
  Ideal (consecutive)4/15 (26.7)2/12 (16.7)6/27 (22.2)
  Inadequate(arbitrary)1/15 (6.7)2/12 (16.7)3/27 (11.1)
  Unclearly reported10/15 (66.6)8/12 (66.6)18/27 (66.7)
Diagnostic test   
Description of scanning method, gestation at scan, measurementparameter and cutoff level   
  Ideal15/15 (100)8/12 (66.7)23/27 (85.2)
  Unclearly reported0115 (0)4/12 (33–3)4/27 (14.8)
Outcome   
Blinding of results   
  Ideal11/15 (73.3)5/12 (41.7)16/27 (59.2)
  Inadequate0/15 (0)1/12 (8.3)1/27 (3.7)
  Unclearly reported4/15 (26.7)6/12 (50.0)10/27 (37.1)
Follow up   
  Ideal (> 90%)11/15 (73.3)11/12 (91.7)22/27 (81.5)
  Second best (81–90)3/15 (20.0)1/12 (8.3)4/27 (14.8)
  Inadequate (≤80%)1/15 (6.7)0/12 (0)1/27 (3.7)

In the low risk population the method of population enrolment was ideal (consecutive) in four out of 15 studies16–17,25,39 (27%). All 15 studies16–17,21,25,27,33–37,39,41–44 (100%) provided an adequate description of the uterine artery Doppler velocity waveform analysis. Eleven studies16–17,21,25,27,34–35,37,39,42,44 (73%) had adequate blinding of test results and follow up was > 90% in 11 studies16,21,27,33–35,37,41,42–44 (73%) Among the 12 studies19,22–24,26,28–32,38,40 in the high risk population, there were two studies23,30 (17%) with consecutive enrolment. Only eight studies19,23–24,26,29–30,38,40 (67%) provided an adequate description of the test. Adequate blinding of test results was present in five studies19,24,26,29,40 (42%) and follow up was > 90% in 11 studies19,22–23,26,28–32,38,40 (92%).

Among the 15 studies16–17,21,25,27,33–37,39,41–44 in the low risk population, the uterine artery Doppler velocity waveform analysis was performed using transabdominal ultrasound in all except for one study41. This study41 utilised transvaginal scanning for Doppler analysis and hence was excluded from meta-analyses. Doppler velocimetry in all the 12 studies19,22–24,26,28–32,38,40 in the high risk population were performed with the transabdominal approach.

Uterine artery Doppler ultrasound in low risk population

There were 15 articles16–17,20–21,25,27,33–37,39,42–44 used for meta-analyses in the low risk population. The likelihood ratios generated for the different measurement parameters and outcomes produced minimal to moderate shift in pre-test to post-test probabilities (Tables 3 and 4).

Table 3.  Likelihood ratios for predicting pre-eclampsia, intrauterine growth retardation and perinatal death in primary studies of uterine artery Doppler ultrasound on low risk population. LR = likelihood ratio; IUGR = intrauterine growth retardation.
 Positive testNegative test
Cutoff level for test and outcome measuren/n (%)LR (95% CI)*n/n (%)LR (95% CI)*
  1. *In the presence of zero in any cell of the 2×2 table, 0.5 was added to each cell to calculate the likelihood ratios.

  2. Outcome data calculated from prevalence, sensitivity and specificity reported in original article.

Flow waveform ratio ± diastolic notch    
Pre-eclampsia    
  Campbell et al. (1986)259/50 (18.0)1.8 (1.1–2.8)5/76 (6.6)0.6 (0.3–1.2)
  Steel et al. (1990)3727/118 (22.9)8.6 (7.3–9.1)7/896 (0.8)0.2 (0.1–0.4)
  Bewley et al. (1991)1610/51 (19.6)5.1 (2.7–9.4)32/866 (3.7)0.8 (0.7–0.9)
  Bower et al. (1993)1739/329 (11.8)5.2 (4.3–6.3)13/1729 (0.8)0.3 (0.24.5)
  Valensise et al. (1993)338/26 (30.8)13.0 (7.9–21.5)1/246 (0.4)0.1 (0.02–0.8)
  North et al. (1994)274/53 (7.5)2.4 (1.0–5.6)11/393 (2.8)0.8 (0.6–1.1)
  Harrington et al. (1996)3634/110 (30.9)11.8 (9.0–15.4)10/1094 (0.9)0.2 (0.144)
  Frusca et al. (1997)427/36 (19.4)9.0 (5.1–15.8)4/383 (1.0)0.4 (0.2–0.8)
  Irion et al. (1998)4412/118 (10.2)2.7 (1.6–4.6)34/1041 (3.3)0.8 (0.7–1.0)
IUGR    
  Campbell et al. (1986)2512/50 (24.0)1.9 (1.2–2.9)6/76 (7.9)0.5 (0.3–1.0)
  Schulman et al. (1987)346/16 (37.5)3.3 (1.5–7.2)5/55 (9.1)0.5 (0.3–1.0)
  Hanretty et al. (1989)391/19 (5.2)1.0 (0.1–6.7)15/272 (5.5)1.0 (0.9–1.1)
  Newnham et al. (1990)353/24 (12.5)1.3 (0.443)46/477 (9.6)1.0 (0.9–1.1)
Steels a/. (1990)3732/118 (27.1)3.5 (2.5–5.0)65/896 (7.2)0.7 (0.6–0.8)
  Bewley et al. (1991)1618/52 (34.6)3.6 (2.1–6.1)100/861 (11.6)0.9 (0.8–1.0)
  Bower et al. (1993)1784/329 (25.5)2.8 (2.3–3.4)141/1729 (8.2)0.7 (0.64.8)
  Valensise et al. (1993)3314/26 (534)13.9 (7.4–26.2)7/246 (2.8)0.4 (0,246)
  North et al. (1994)2715/56 (26.8)5.2 (3.3–8.3)15/401 (3.7)0.6 (0.4–0.8)
  Todros et al.(1995)215/59 (8.5)1.9 (0.84.6)37/857 (4.3)0.9 (0.8–1.0)
  Hanington et al. (1996)3642/110 (38.2)5.1 (3.6–7.1)89/1094 (8.1)0.7 (0.6–0.8)
  Frusca etal. (1997)4216/36 (44.4)9.4 (5.4–16.3)17/383 (4.4)0.5 (044.8)
  Liberati et al. (1997)4322/63 (34.9)4.8 (3.1–7.4)28/436 (6.4)0.6 (0.5–0.8)
  Irion et al. (1998)4433/118 (28.0)3.2 (2.2–4.5)94/947 (9.9)0.8 (0.7–0.9)
Perinatal death    
  Campbell et al. (1986)251/50 (2.0)1.9 (0.8–4.4)0/76 (0)0.4 (0.04–4.6)
  Steel et al. (1990)375/118 (4.2)3.0 (1.4–6.2)101896 (1.1)0.8 (0.5–1.1)
  Bower et al. (1993)175/329 (1.5)1.2 (0.5–2.6)22/1729 (1.3)1.O(0.8–1.2)
  Harrington et al. (1996)3611110 (1.0)0.9 (0.1–6.0)11/1094 (1.0)1.0 (0.4–1.2)
  Diastolic notch alone Pre-eclampsia    
  Bower et al.(1993)2037/301 (12.3)6.3 (5.3–7.5)8/1757 (0.4)0.2 (0.14.4)
  Hanington et al. (1996)3634/110 (30.9)11.8 (9.0–15.4)10/1094 (0.9)0.2 (0.144)
  Irion et al. (1998)4412/160 (7.5)2.0 (1.2–3.3)34/999 (3.4)0.8 (0.7–1.O)
IUGR    
  Harrington et al. (1996)3642/110 (38.2)5.1 (3.6–7.1)89/1094 (0.1)0.7 (0.648)
  Irion et al. (1998)4438/160 (23.8)2.5 (1.8–3.5)891999 (8.9)0.8 (0.7–0.9)
Table 4.  Pooled estimates of pre-test probabilities, likelihood ratios and post-test probabilities for uterine artery Doppler ultrasound in predicting pre-eclampsia, intrauterine growth retardation and perinatal death. IUGR = intrauterine growth retardation
Population, cutoff level for test and outcome measurePre-test probability % (95% CI)Likelihood ratio (95% CI)Post-test probability % (95% CI)
  1. Heterogeneity: *P ≤ 0.0001, **P ≤ 0.05; results homogenous for other meta-analyses; Breslow-Day test used.

LOW RISK POPULATION   
Flow waveform ratio ± diastolic notch   
Pre-eclampsia (n= 9 studies)   
  Positive test result3.5 (3.1–3.9)6.4 (5.7–7.1)*18.8 (16.4–21.5)
  Negative test result3.5 (3.1–3.9)0.7 (0.6–0.8)*2.5 (2.1–2.9)
IUGR (n= 14 studies)   
  Positive test result9.8 (9.2–10.4)3.6 (3.2–4.0)*28.0 (25.4–30.7)
  Negative test result9.8 (9.2–10.4)0.8 (0.8–0.9)*8.4 (7.8–9.0)
Perinatal death (n= 4 studies)   
  Positive test result1.3 (0.9–1.6)1.8 (1.2–2.9)2.3 (1.4–3.8)
  Negative test result1.3 (0.9–1.6)0.9 (0.8–1.1)1.2 (0.9–1.6)
Diastolic notch alone   
Pre-eclampsia (n= 3 studies)   
  Positive test result3.0 (2.5–3.6)6.8 (5.9–7.9)17.7 (14.7–21.2)
  Negative test result3.0 (2.5–3.6)0.7 (0.6–0.8)*2.2 (1.7–2.7)
IUGR (n= 2 studies)   
  Positive test result10.9 (9.7–12.2)3.5 (2.8–4.4)**29.9 (24.7–35.7)
  Negative test result10.9 (9.7–12.2)0.8 (0.7–0.8)**8.5 (7.4–9.8)
HIGH RISK POPULATION   
Flow waveform ratio ± diastolic notch   
Pre-eclampsia (n= 11 studies)   
  Positive test result9.8 (7.9–11.8)2.8 (2.3–3.4)23.5 (18.6–29.2)
  Negative test result9.8 (7.9–11.8)0.8 (0.7–0.9)7.8 (6.1–10.0)
IUGR (n= 9 studies)   
  Positive test result17.8 (14.4–21.1)2.7 (2.1–3.4)36.7 (29.5–44.6)
  Negative test result17.8 (14.4–21.1)0.7 (0.6–0.9)13.8 (10.7–17.6
Perinatal death (n= 4 studies)   
  Positive test result8.9 (5.4–12.3)4.0 (2.4–6.6)27.8 (16.4–42.9)
  Negative test result8.9 (5.4–12.3)0.6 (0.4–0.9)5.5 (3.1–9.7)
Diastolic notch alone   
Pre-eclampsia (n= 1 study)   
  Positive test result5.8 (1.3–10.3)20.2 (7.3–56.3)55.6 (25.1–82.3)
  Negative test result5.8 (1.3–10.3)0.2 (0.03–10.)1.1 (0.1–7.2)
IUGR (n= 1 study)   
  Positive test result17.5 (10.1–24.8)2.4 (0.6–8.6)33.3 (11.1–66.6)
  Negative test result17.5 (10.1–24.8)0.9 (0.7–1.1)16.0 (9.9–24.7)
Perinatal death (n= 1 study)   
  Positive test result1.9 (0.7–4.5)2.7 (0.2–32.2)5.0 (0.3–47.5)
  Negative test result1.9 (0.7–4.5)0.8 (04.1.8)1.6 (0.3–7.4)

The use of an abnormal velocity waveform ratio ± the presence of a diastolic notch to predict pre-eclampsia resulted in a heterogeneous estimate of likelihood ratios: 6.4 (95% CI 5.7–7.1) for a positive test result and 0.7 (95% CI 0.6–0.8) for a negative test result (Breslow-Day test, P= 0.0001). Exploration of the sources of heterogeneity by performing sensitivity analyses on studies subgroups based on the various features of methodological quality and types of Doppler scanning techniques used could not provide an adequate explanation. In general, studies with adequate methods generated more conservative estimates of diagnostic prediction but the confidence intervals in the subgroups overlapped (Table 5). Pulsed wave Doppler scanning also produced more conservative estimates of likelihood ratio for a positive test result but a more liberal estimate of likelihood ratio was obtained with a negative test result (Table 5). Furthermore, the exclusion of studies with nonadequate methods and those that only employed continuous wave scanning had minimal effect on the overall estimate. Since the ultrasound scans for all the studies reporting on this outcome measure was conducted ≤ 24 weeks of gestation, it was not possible to explore for sources of heterogeneity based on the gestational age of ultrasound scanning.

Table 5.  Sensitivity analysis of meta-analyses of uterine artery Doppler ultrasound using flow velocity waveform ratio f diastolic notch as measurement parameter in low risk population.
 Likelihood ratio (95% CI)
Outcome criteria (no. of studies)Positive testNegative test
  1. *Study quality was considered to be ideal when consecutive enrollment of study population was employed, a description of the gestational age at the time of Doppler scanning, the measurement parameter used and the cutoff level for an abnormal result was provided in the manuscript, Doppler results was adequately blinded to the attending physicians and > 90% of all the women enrolled into the study were available for data analysis. All other methods of study design for the population enrollment, diagnostic intervention used and ascertaining the outcome measure(s) were considered inadequate. In the absence of any explicit information on the above methodological quality, the study was classified as unclearly reported.

Pre-eclampsia  
Study quality*  
  Ideal (n= 1)5.1 (2.7–9.4)0.8 (0.7–1.0)
  Inadequate/unclearly reported (n= 8)6.4 (5.7–7.2)0.5 (040.6)
All studies (n= 9)6.4 (5.7–7.1)0.7 (0.6–0.8)
Doppler scanning method  
Continuous wave Doppler (n= 2)7.9 (6.2–10.0)0.7 (064.9)
Pulsed wave Doppler ± color flow mapping (n= 7)6.3 (5.5–7.2)0.5 (0.4–0.6)
All studies (n= 9)6.4 (5.7–7.1)0.7 (0.6–0.8)
Gestational age of scanning  
  Scan < = 24 weeks of gestation (n= 9)6.4 (5.7–7.1)0.7 (0.6–0.8)
  Scan > 24 weeks of gestation (n= 0)
  All studies (n= 9)6.4 (5.7–7.1)0.7 (0.6–0.8)
Intrauterine growth retardation  
Study quality*  
  Ideal (n= 1)3.6 (2.1–6.1)0.9 (0.8–1.0)
  Inadequate/unclearly reported (n= 13)3.6 (3,240)0.8 (0.8–0.9)
  All studies (n= 14)3.6 (3.240)0.8 (0.8–0.9)
Doppler scanning method  
Continuous wave Doppler (n= 5)3.2 (2.542)0.9 (0.9–1.0)
Pulsed wave Doppler ± color flow mapping (n= 9)3.7 (3.242)0.8 (0.7–0.8)
  All studies (n= 14)3.6 (3.240)0.8 (0.8–0.9)
Gestational age of scanning  
Scan ≤ 24 weeks of gestation (n= 12)3.6 (3.2–4.0)0.8 (0,849)
Scan ≥ 24 weeks of gestation (n= 2)2.8 (1.3–5.7)1.0 (0.9–1.1)
All studies (n= 14)3.6 (3.240)0.8 (0.8–0.9)

The likelihood ratios generated for the prediction of intrauterine growth retardation were also heterogeneous: 3.6 (95% CI 3.2–4.0) for a positive test result and 0.8 (95% CI 0.8–0.9) for a negative test result. Again, the sources of heterogeneity remained unexplained after performing sensitivity analyses on the various study subgroups. The likelihood ratios remained virtually unchanged irrespective of the methodological quality of the studies, method of Doppler scanning employed for screening for this outcome and gestational age of scanning (Table 5).

For the prediction of perinatal death, the pooled likelihood ratio was 1.8 (95% CI 1.2–2.9) and 0.9 (95% CI 0.8–1.1) for positive and negative test results, respectively. Consequently, the pre-test probability given a positive test result was only minimally increased from 1.3% (95% CI 0.9–1.6%) to 2.3% (95% CI 1.4–3.8%). With a negative test result, it only decreased to 1.2% (95% CI 0.9–1.6%). The pooled estimates of likelihood ratio for the prediction of perinatal death were homogeneous.

Data from the studies with test results based on the presence or absence of a diastolic notch alone only reported for pre-eclampsia and intrauterine growth retardation. The pre-test probability of pre-eclampsia was 3.0% (95% CI 2.5–3.6%). A positive result generated moderate shift in the pre-test probability; the pooled likelihood ratio for a positive test result was 6.8 (95% CI 5.9–7.9), raising the probability to 17.7% (95% CI 14.7–21.2%). A negative test result had a minimal effect; the pooled likelihood ratio was 0.7 (95% CI 0.6–0.8), reducing the probability to 2.2% (95% CI 1.7–2.7%). For intrauterine growth retardation, the pooled likelihood ratio for a positive test result was 3.5 (95% CI 2.8–4.4) and 0.8 (95% CI 0.7–0.8) for a negative test result. There was only a minimal increase in the pre-test probability from 10.9% (95% CI9.7–12.2%) to 29.9% (95% CI 24.7–35.7%) with a positive test result. It decreased only to 8.5% (95% CI 7.4–9.8%) with a negative test result. However, the pooled estimates of likelihood ratio in all these analyses were heterogeneous. Since the methodological quality of all the studies with test results based on the presence or absence of a diastolic notch alone were nonideal and employed pulsed wave Doppler scanning before 24 weeks of gestation, the exploration for the sources of heterogeneity was not possible with the meta-analyses based on both these two outcome measures.

Uterine artery Doppler ultrasound in high risk population

Twelve of the eligible studies19,22–24,26,28–32,38,40 were categorised as having recruited high risk subjects (Table 6). The pooled likelihood ratios from all the analyses in the high risk subgroup were homogeneous (Table 4).

Table 6.  Likelihood ratios for predicting pre-eclampsia, intrauterine growth retardation and perinatal death in primary studies of uterine artery Doppler ultrasound on high risk population. LR = likelihood ratio; IUGR = intrauterine growth retardation.
 Positive testNegative test
Cutoff level for test and outcome measuren/n(%)LR (95% CI)*n/n(%)LR (95% CI)*
  1. *In the presence of zero in any cell of the 2 × 2 table, 0.5 was added to each cell to calculate the likelihood ratios.

  2. outcome data calculated from prevalence, sensitivity and specificity reported in the original article.

Flow waveform ratio ± diastolic notch    
Pre-eclampsia    
  Trudinger et al. (1985)329/28 (32.1)3.1 (1.9–5.2)3/63 (4.8)0.3 (0.1–0.9)
  Jacobson et al. (1990)406/36 (16.7)1.8 (1.1–3.1)3/55 (5.4)0.5 (0.2–1.3)
  Pattinson et al. (1991)196/20 (30.0)2.8 (1.7–4.8)1/33 (3.1)0.2 (0.03–1.3)
  Benifla et al. (1992)384/6 (66.7)3.5 (1.7–7.1)0118 (0)0.1 (0.01–1.9)
  Guzman et al. (1992)312/4 (50.0)2.4 (0.4–14.0)6/23 (26.1)0.8 (0.5–1.3)
  Caruso et al. (1993)234/8 (50.0)4.0 (1.5–10.7)1/17 (5.9)0.2 (0.041 .5)
  Haddad et al. (1993)22215 (40.0)8.6 (2.8–26.2)0132 (0)0.2 (0.01–2.3)
  Ferrier et al. (1994)292/14 (14.3)2.0 (0.7–5.8)2/37 (5.4)0.7 (0.2–1.8)
  Chan et al. (1995)244/18 (222)3.9 (1.4–10.8)191316 (6.0) 0.9 (0.7–1.0)
  Haddad et al. (1995)305/26 (19.2)1.9 (1.3–2.8)0122 (0)0.2 (0.01–2.4)
  Konchak et a1. (1995)285/11 (45.4)13.5 (5.7–31.6)1/92 (1.1)0.2 (0.03–1.1)
IUGR    
  Trudinger et al. (1985)3215/28 (53.6)3.0 (1.7–5.5)10163 (15.9)0.5 (0.348)
  Jacobson et al. (1990)4012/36 (33.3)2.2 (1.4–3.4)5/55 (9.1)0.4 (0.2–0.9)
  Benifla et al. (1992)385/10 (50.0)4.0 (1.8–8.7)0118 (0)0.1 (0.01–1.5)
  Guzman et al. (1992)314/4 (100)28.3 (1.7463.0)223 (8.7)0.4 (0.1–1.0)
  Caruso et al. (1993)232/7 (28.6)2.9 (1.0–8.9)1/18 (5.6)0.4 (0.1–2.2)
  Haddad et al.(1993)22215 (40.0)8.6 (2.8–26.2)0132 (0)0.2 (0.01–2.3)
  Femer et al. (1994)295/14 (35.7)4.2 (2.1–8.3)1/37 (2.7)0.2 (0.03–1.2)
  Haddad et al. (1995)304/26 (15.4)1.6 (0.9–2.6)1/22 (4.5)0.4 (0.1–2.4)
  Konchak et al. (1995)283/11 (27.3)1.8 (0.540)15/92 (16.3)0.9 (0.7–1.1)
Perinatal death    
  Aristidou et al. (1990)267/18 (38.9)3.8 (1.8–8.2)7/80 (8.8)0.6 (0.3–1.0)
  Guzman et al. (1992)313/4 (75.0)14.6 (3.4–71.9)0/23 (0)0.1 (0.01–1.8)
  Caruso et al. (1993)232l8 (25.0)3.1 (1.3–7.0)0/17 (0)0.2 (0.02–2.9)
  Konchak et al. (1995)280/11 (0)2.2 (0.2–26.2)1/92 (1.1)0.8 (0.4–1.9)
Diastolic notch alone    
Pre-eclampsia    
  Konchak et al. (1995)28519 (55.6)20.2 (7.3–56.3)1/94 (1.1)0.2 (0.03–1.0)
IUGR    
  Konchak et al. (1995)28319 (33.3)2.4 (0.6–8.6)15/94 (16.0)0.9 (0.7–1.1)
Perinatal death    
  Konchak et al. (1995)280/9 (0)2.7 (0.2–32.2)1/94 (1.1)0.8 (0.4–1.8)

When the presence of an abnormal velocity waveform ratio ± diastolic notch was used to predict preeclampsia, the likelihood ratios generated were 2.8 (95% CI 2.3–3.4) and 0.8 (95% CI 0.7–0.9) for a positive and negative test result, respectively. The pre-test probability for this outcome was 9.8% (95% CI 7.9–11.8%). These likelihood ratios resulted in minimal increase in the pre-test probability to 23.5% (95% CI 18.6–29.2%) for a positive test result and decrease to 7.8% (95% CI 6.1–10.0%) for a negative test result. For intrauterine growth retardation, the pooled likelihood ratios were similarly disappointing; 2.7 (95% CI 2.1–3.4) for a positive test result and 0.7 (95% CI 0.6–0.9) for a negative test result. The pre-test probability increased from 17.8% (95% CI 14.4–21.1%) to 36.7% (95% CI 29.5–44.6%) with a positive test result. It decreased to 13.8% (95% CI 10.7–17.6%) in the presence of a negative test result. The pooled likelihood ratios for the prediction of perinatal death were 4.0 (95% CI 2.4–6.6) and 0.6 (95% CI 0.4–0.9) for a positive and negative test result, respectively. A positive test result increased the pre-test probability from 8.9% (95% CI 5.4–12.3%) to 27.8% (95% CI 16.4–42.9%), whereas it is decreased to 5.5% (95% CI 3.1–9.7%) with a negative test result.

There was only one study28 that used the presence or absence of a diastolic notch as the measurement parameter and hence pooling of results was not possible. The pre-test probability for pre-eclampsia was 5.8% (95% CI 1.3–10.3%). The likelihood ratio for a positive result was 20.2 (95% CI 7.3–56.3), resulting in a large change in pre-test probability to 55.6% (95% CI 25.1–82.3%). A negative test result generated a moderate shift in pretest probability; likelihood ratio was 0.2 (95% CI 0.03–1.0), resulting in a post-test probability of 1.1% (95% CI 0.1–7.2%). The limited sample size in this analysis resulted in the upper limit of the estimate of likelihood ratio for a negative test result to include unity, producing a considerable amount of uncertainty around it. There was limited prediction for intrauterine growth retardation; the likelihood ratios were 2.4 (95% CI 0.6–8.6) and 0.9 (95% CI 0.7–1.1) for a positive and negative test, respectively. The pre-test probability therefore increased from 17.5% (95% CI 10.1–24.8%) to 33.3% (95% CI 11.1–66.6%). It decreased to 16.0% (95% CI 9.9–24.7%) with a negative test result. For perinatal death, the likelihood ratio was 2.7 (95% CI 0.2–32.2) for a positive test result and 0.8 (95% CI 0.4–1.8) for a negative test result. The pre-test probability was raised from 1.9% (95% CI 0.7–4.5%) to 5.0% (95% CI 0.3–47.5%) with a positive test result and reduced to 1.6% (95% CI 0.3–7.4%) in the presence of a negative test result.

DISCUSSION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. METHODS
  5. RESULTS
  6. DISCUSSION
  7. References

In this review stringent criteria for a rigorous systematic review were observed. We had a prospective protocol, focused a priori on a research question, clearly stated our search strategies, excluded data which were subject to duplicate publication, assessed the study methodological quality in an unbiased and reproducible fashion, quantitatively summarised the evidence of diagnostic attribute of the test, explored for heterogeneity in the results and also for the possible sources of variation in the data from study to study45–46. Our results suggests that an abnormal flow waveform ratio ± diastolic notch as the measurement parameter used for uterine artery Doppler flow velocity has limited predictive value for pre-eclampsia, intrauterine growth retardation and perinatal death. The presence of a diastolic notch alone as the criterion for a positive test result in low risk population also has limited predictive value.

It is generally agreed that pre-eclampsia should only be diagnosed in the presence of gestational hypertension and proteinuria47. However, the criteria for defining hypertension and proteinuria is still lacking agreement. The significant variation in the definition of hypertension and proteinuria present in the selected studies is again another potential contributor of heterogeneity in the results. Furthermore, other hypertensive conditions in pregnancy such as chronic hypertension, chronic renal disease and chronic hypertension with superimposed pre-eclampsia can sometimes mimic pre-eclampsia47. Until there is an objective reference standard to verify the disease, we are left with no option but to rely on this condition being diagnosed correctly by clinical means. This problem was further compounded by the finding that 11 out of 27 selected studies (41%) did not adequately blind the Doppler test results from the clinicians who subsequently verified for the presence or absence of disease. Not surprising, statistically significant heterogeneity was therefore encountered in the low risk population given that there was such significant variation between the various studies with regards to the criteria used to diagnose this outcome measure.

All except one of the studies selected for the overview defined proteinuria as > 300 mg per 24 hours and/or ≥ 2+ on labstix urine analysis. The remaining study16 (with exception to the above definition) considered significant proteinuria as > 150 mg per 24 hours or > 1+ on labstix testing of the urine sample. Such a level of urinary protein excretion or semi-quantitative protein concentration may be considered to be inappropriately low for defining significant proteinuria in pre-eclampsia47. This study reported on the prediction of pre-eclampsia and intrauterine growth retardation with uterine artery Doppler velocimetry in the low risk population16. The omission of this study16 from the meta-analyses produced pooled likelihood ratios which were virtually identical to that presented in Table 4 (pooled likelihood ratios were 6.4 (95% CI 5.7–7.2) for positive test result and 0.7 (95% CI 0.6–0.7) for negative test result for preeclampsia; 3.6 (95% CI 3.2–4.0) for positive test result and 0.8 (95% CI 0.8–0.9) for negative test result for intrauterine growth retardation). Therefore, the inclusion of this study did not alter the result of the overview.

We also evaluated the influence of gestational age on the uterine artery velocity waveform pattern by performing sensitivity analyses stratified according to gestational age of scanning. The physiological trophoblastic invasion of the uterine spiral arteries occurs at 14–16 weeks of gestation48. The failure of this physiological process is known to subsequently lead to the development of pre-eclampsia. Hence, uterine artery Doppler velocimetry performed before 16 weeks of gestation is unlikely to be a useful screening test for this condition. In addition, it is irrational to employ this test to screen for pre-eclampsia in the third trimester of pregnancy (> 24 weeks of gestation) as the disease usually would have been established by then48. For these reasons, the studies were dichotomised to whether the ultrasound scans were performed between 16–24 or > 24 weeks in the sensitivity analyses. All nine studies in the low risk population with pre-eclampsia as the outcome employing screening ≤ 24 weeks of gestation (Table 5). With regards to the 14 studies in the same population with intrauterine growth retardation as the outcome, the likelihood ratios for studies with Doppler ultrasound scanning at 16–24 and > 24 weeks of gestation were similar with overlapping confidence intervals to the overall pooled likelihood ratios (Table 5). We have therefore presented the result with the overall pooled likelihood ratios in order to achieve a more precise estimate.

The studies categorised into the high risk subgroup included subjects with different risk factors for developing pre-eclampsia. Although a rather heterogeneous group of conditions have been incorporated into the meta-analyses of this subgroup, the pooled estimates of likelihood ratios from all these analyses remained homogenous (Table 4).

Other measures of diagnostic accuracy such as sensitivity and specificity could also have been employed to evaluate the performance of the test. These measures lack stability when subjected to meta-analysis (i.e. significant heterogeneity is more likely to be encountered)14,49. Moreover, inference based on sensitivity and specificity tend to overestimate the clinical value of the test50. In this overview we have chosen to use likelihood ratio as our preferred measure of diagnostic accuracy because it is independent of the prevalence of disease and the post-test probability of disease can always be calculated for any known pre-test probability51. Hence, the likelihood ratios generated by this overview can be applied to any patient population provided the prevalence of disease for that population is known. Because the post-test probability refers to the probability of disease given a certain test result, it is immediately more clinically useful to clinicians compared with sensitivity and specificity14,52.

Bayes'theorem implies that the predictive accuracy of any test is conditional on the overall prevalence of disease in the population tested. The prevalence (i.e. pre-test probability) of pre-eclampsia can be considered low in relative terms for both the low risk (3.0%–3.5%) and high risk (5.8%–9.8%) populations. Therefore, uterine artery Doppler flow velocimetry will need to have a high likelihood ratio in order to have a reasonable amount of predictive ability for pre- eclampsia in the presence of a positive test result, whereas a low likelihood ratio would be required to confidently exclude the disease with a negative test result. Unfortunately, the likelihood ratios generated by uterine artery Doppler waveform analysis are generally conservative, therefore the clinical usefulness of this test in the prediction of pre-eclampsia is limited.

This overview suggest that the use of uterine artery flow waveform ratio ± diastolic notch as the measurement parameter for the test result has limited diagnostic prediction for pre-eclampsia, intrauterine growth retardation and perinatal death. Future research should instead focus on Doppler ultrasonic detection of uterine artery diastolic notches alone to predict for pre-eclampsia, especially in pregnant women considered to be at high risk for this condition. Decisions on whether to initiate medical or nonmedical interventions to prevent pre-eclampsia should therefore not be based on these test results alone.

Acknowledgements

The authors would like to thank Drs T. Shinomura, R. Devlieger, M. Jacobs, M. Quaresma and A. Kolle for their contribution in extracting data from the non-English articles.

References

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. METHODS
  5. RESULTS
  6. DISCUSSION
  7. References