Background: Early identification of pregnant women at risk for preeclampsia is a priority to implement preventive measures. Some biochemical and ultrasonographic parameters have shown promising predictive performance, but so far there is no clinically validated screening procedure.
Content: Using a series of keywords, we reviewed electronic databases (Medline, Embase, all records to May 2009) reporting the performance of biological and ultrasonographic markers to predict preeclampsia, both single markers and combinations of markers. We analyzed the data according to gestational age and risk levels of the studied populations. We evaluated the methodological quality of included publications using QUADAS (quality assessment of diagnostic accuracy studies). We identified 37 relevant studies that assessed 71 different combinations of biochemical and ultrasonographic markers. Most studies were performed during the second trimester on small-scale high-risk populations with few cases of preeclampsia. Combinations of markers generally led to an increase in sensitivity and/or specificity compared with single markers. In low-risk populations, combinations including placental protein 13 (PP13), pregnancy-associated plasma protein A (PAPP-A), a disintegrin and metalloprotease-12 (ADAM12), activin A, or inhibin A measured in first or early second trimester and uterine artery Doppler in second trimester appear promising (sensitivity 60%–80%, specificity >80%). In high-risk populations, the combination of PP13 and pulsatility index in first trimester showed 90% sensitivity and 90% specificity in a single study limited to severe preeclampsia.
Summary: Combinations of biochemical and ultrasonographic markers improved the performance of early prediction of preeclampsia. From a perspective of integrative medicine, large population-based studies evaluating algorithms combining multiple markers are needed, if screening approaches are to be eventually implemented.
Preeclampsia (PE),1 a leading cause of adverse pregnancy outcomes worldwide, remains a major cause of maternal and perinatal mortality (1)(2)(3). It is defined as de novo hypertension (>140/90 mmHg) appearing after 20 weeks of gestation accompanied by proteinuria (>0.3 g/24 h) (4)(5). An increasing body of evidence indicates that women affected by PE will be at higher risk for cardiovascular disease later in life (6)(7)(8)(9). WHO has recognized the importance of PE by launching a program specifically dedicated to the study and treatment of this syndrome (10).
PE is characterized by a complex pathophysiology and heterogeneous clinical and laboratory findings (11)(12)(13). Numerous pathophysiological mechanisms, alone or in combination, have been suggested to be responsible for the diverse subsets of PE. They include impaired vascular remodeling of the maternal–fetal interface, excessive immune response to paternal antigens, systemic inflammatory response, and dysfunctional placental or endothelial response, all of these processes being modulated by genetic and environmental parameters (1)(14)(15)(16). Such heterogeneity of potential processes leading to, or resulting from, PE has contributed to the lack of diagnostic means for identification of women susceptible to developing PE, resulting in delayed recognition and severe complications (17)(18) and impeding evaluation of new preventive interventions. The latter problem holds particular consequences for high-risk women, in whom treatments have shown potential protective effects (19)(20)(21)(22). A recent metaanalysis suggests that, in high-risk women, low-dose aspirin started before 16 weeks’ gestation could prevent up to 50% of PE, severe PE, and intrauterine growth restriction (IUGR) (23). This reinforces the need for early identification of at-risk women with the objective of implementing targeted interventions for improving both perinatal and maternal outcomes.
In 2004, after performing a systematic review of screening tests for PE, WHO reaffirmed that “there is no clinically useful screening test to predict the development of preeclampsia in either low-risk or high-risk populations. Further prospective, longitudinal studies are needed” (24). Since this assertion, many groups have identified or studied potential biochemical and/or biophysical markers based on physiological mechanisms, some of which showing encouraging results. Systematic reviews and/or metaanalyses assessing the clinical utility of, most often, single markers, have recently been published (25)(26)(27)(28), but so far no marker has demonstrated an appropriately high accuracy level to justify its clinical application.
Because of the heterogeneous nature of PE, a combination of 2 or more independent biomarkers, each reflecting a different pathophysiological process, should potentially increase the likelihood to derive suitable predictive algorithms. So far, however, no systematic review has evaluated whether biochemical markers improved the performance of ultrasonographic markers, and vice versa. In this study, we performed a systematic review of the available evidence pertaining to the predictive characteristics of combinations of biochemical and ultrasonographic markers in the first and second trimester. The development of efficient prediction algorithms would be a major step toward the early identification of women destined to suffer from PE, to classify them according to risk subsets and eventually to offer them appropriate surveillance and prophylactic interventions.
Materials and Methods
literature search and study selection
We made electronic searches of Medline and Embase bibliographic databases from inception to May 2009. The search strategy consisted of MeSH (medical subject heading) terms, Emtree terms, and keywords related to the disease (hypertensive disorders of pregnancy, preeclampsia, gestosis, eclampsia, pregnancy-induced hypertension, etc.) combined with biochemical and biophysical (ultrasonographic, Doppler, etc.) markers. We developed the search strategy through an iterative and collaborative process in close collaboration with information specialists and clinical experts in laboratory medicine and obstetrics, particularly for the development of the biochemistry-marker search strategy. Special care was taken to include proper synonyms, abbreviations, and spelling differences of biochemical markers, as well as to look for and to select more general MeSH terms when necessary (for example, the marker IL-10 is linked to the MeSH “interleukin-10,” which is included within the MeSH “Interleukins,” which, in turn, is part of “Cytokines”; we thus selected “cytokines” as an encompassing MeSH term). Overall, this strategy produced 68 MESH terms that were combined with 303 keywords, which represented more than 118 different molecules to produce the search strategy in Medline (for Embase, we used 311 keywords). Supplemental Table 1 (which accompanies the online version of this article at www.clinchem.org/content/vol56/issue3 ) presents the research strategy used for Medline searching. The search strategy used for Embase is available on request from the authors. Bibliographies of selected publications were manually screened to include potential articles not captured by electronic searches. We initially included all original studies reporting any ultrasound methodology for assessment of either maternal or fetal health and any biochemical marker (from blood, urine, or any biological fluid) in any healthcare setting, in any population level of PE risk, performed before the diagnosis of the disease and reported in either English or French. Studies allowing development of 2 × 2 tables for combined test accuracy were included.
data extraction and quality assessment
Titles and abstracts of citations were first screened by 2 independent reviewers (N. Bernard, S. Grenier, or M. Charland) for their relevance. Citations without abstracts were rejected. Screened articles for which there were disagreements between the assessments of 2 independent reviewers were resolved by the third reviewer. After agreement, the full-text article was assessed by 2 reviewers for inclusion and data extraction. For multiple publications of the same data set, only the most relevant study was included.
Methodological quality of the publications was evaluated by two independent reviewers (N. Bernard and M. Charland) using the QUADAS (quality assessment of diagnostic accuracy studies) tool, a questionnaire that uses 14 criteria for evaluating the avoidance of different types of bias (29)(30)(31). Disagreements were resolved by consensus. Because the QUADAS tool requires a clear definition of a reference standard, the definition of the International Society for the Study of Hypertension in Pregnancy was selected: PE is diagnosed in women with persistent high systolic (>140 mmHg) or diastolic (>90 mmHg) blood pressure and proteinuria (>0.3 g/24 h or dipstick test result ≥1+) arising after the 20th week of pregnancy (4).
data synthesis: analysis
We were able to build 2 × 2 tables for combinations of markers from 24 of the included studies in relation to PE. Data related to the predictive performance (sensitivity, specificity, predictive values, and likelihood ratios) of studied combinations were either directly extracted or calculated from retrieved raw data. Using the classical equations for calculation of these parameters, the prevalence of the disease is taken into account as it influences the rate of false positives and false negatives; that is, the calculated positive predictive value (PPV) and negative predictive value (NPV) will vary according to the prevalence of the disease under study. Unfortunately, the large diversity of combinations, the variability related to study characteristics (cutoffs, risk-level of studied populations, etc.) and the small number of included studies for a given combination of biochemical and ultrasonographic markers did not permit meta-analyses to be performed.
The initial electronic search strategy led to screen titles and abstracts of 394 citations. Fig. 1⇓ shows the screening and selection process that was followed for the identification and inclusion of studies combining biochemical and ultrasonographic markers to predict PE. We retrieved 92 potentially eligible primary studies for detailed evaluation and inclusion in the systematic review, and an additional 15 potentially eligible publications from the bibliographies of included studies. Detailed evaluation led to the exclusion of 70 publications that did not meet the selection criteria. Overall, 37 studies were considered relevant and were included in the systematic review; their characteristics are shown in online Supplemental Table 2.
Included studies were conducted in 12 different countries; among them, 19 were from the U.K., whereas 3 each were from Italy and Turkey. Most selected studies were performed using a prospective design including nested case-control studies that were carried out as part of larger prospective studies, consistent with our inclusion criterion that the tests had to be performed before the occurrence of PE. More than 60% (23 of 37) of included studies were conducted on high-risk populations, a characteristic that was either directly stated by the authors or deduced from the study data. For example, a study including women after an abnormal test result (either biochemical or ultrasonographic) was considered to be performed on a high-risk population. Almost half of included studies (17 of 37) were based on fewer than 25 preeclamptic women as cases, whereas only 5 studies investigated more than 100 PE women (see online Supplemental Table 2).
Of the 37 included studies, biochemical and ultrasonographic markers were simultaneously evaluated (same gestational age) in 23 studies, whereas they were performed sequentially in the remaining 14, the biochemical marker always being assessed before the ultrasonographic one. For studies with simultaneously evaluated markers, 8 combinations of tests were conducted during first trimester, 1 in the early second trimester (before 20 weeks), and 14 after 20 weeks of gestation. For studies in which markers were sequentially evaluated, the biochemical markers were sampled during the first trimester of pregnancy in 5 studies and during the second trimester in 9; ultrasonographic markers were assessed during the second trimester in all but 1 study. The most frequently studied biochemical markers were human chorionic gonadotropin (hCG, 8 studies), inhibin A, and soluble fms-like tyrosine kinase 1 (sFlt-1) in 6 studies and α-fetoprotein (AFP), activin A, pregnancy-associated plasma protein A (PAPP-A), and placental growth factor (PlGF) in 5 studies. The uterine artery Doppler pulsatility index (PI) was studied in 20 studies (in 14 as the sole ultrasonographic marker and in 6 as a component of a suite of ultrasonographic markers). Few studies (n = 9) simultaneously combined more than 1 biochemical marker with ultrasonographic measurements to evaluate the predictive capability of a combination of tests for predicting PE.
The methodological quality of selected studies as assessed by their compliance with the 14 items included within QUADAS is shown in Fig. 2⇓ . More than 75% of studies met the following criteria: clear description of patient selection criteria, avoidance of partial and differential verification bias, use of an independent reference test, adequate description of index test, blind assessment of both index and reference tests, and explanation of withdrawals. However, fewer than 50% of studies met the following criteria: appropriate patient spectrum, availability of clinical data, and reporting of uninterpretable results.
Data extraction from the 37 studies allowed us to build 2 × 2 tables and calculate the performance characteristics of combinations drawn from 24 studies (Table 1⇓ ⇓ ⇓ ⇓ ⇓ ). Performance characteristics of the test combinations were expressed as sensitivity, specificity, PPV, and NPV. When this was not possible (13 studies), we used the major conclusions given by the study authors to qualify the predictive ability of the combinations (Table 2⇓ ).
Plotting test sensitivity as a function of false-positive rate (1 − specificity) (Figs. 3⇓ and 4⇓ ) allowed comparison of the performance characteristics of markers taken alone or in combination. The data plotted were stratified according to the risk level of the population (high vs low risk) and the time of marker sampling or measurement (first, early second, or late second trimester). Overall, combinations of markers generally led to an increase in sensitivity and/or specificity compared with single markers. In low-risk populations, several combinations, including PP13, PAPP-A, a disintegrin and metalloproteinase 12 (ADAM12), activin A, or inhibin A measured early in first or early second trimester combined with uterine artery Doppler (pulsatility index or resistance index and/or presence of a notch) showed reasonable accuracy (sensitivity 60%–80% for specificity >80%). In 1 high-risk population, placental protein 13 (PP13) and PI measured in the first trimester showed 90% sensitivity and 90% specificity when limited to the diagnosis of severe preeclampsia.
We evaluated the performance characteristics of 71 combinations of ultrasonographic and biological markers in identifying women susceptible to development of PE. Although 65 combinations of markers included at least 1 marker that was assessed during the second trimester, only 6 combinations included both biological and ultrasonographic markers measured during the first trimester (32)(33)(34)(35)(36)(37). No given combination was assessed more than once. The large variety of marker combinations observed across studies reflects the complexity and heterogeneity of the pathophysiology of PE—or the lack, at this point, of powerful individual candidate markers. Such heterogeneity and the limited number of published studies for each combination precluded any further comparative analyses such as meta-analyses.
Among the numerous marker combinations studied during the second trimester, several appeared to have promising predictive characteristics (sensitivity 60%–90%, specificity >90%). Overall, the addition of uterine artery Doppler data to biochemical marker data improved the predictive performance of biochemical markers alone to a greater extent than did the addition of biochemical markers to uterine artery Doppler data alone. In 1 study (the only first-trimester study of high-risk populations), however, the addition of PP-13 to uterine artery PI increased sensitivity from 40% to 90% (at a specificity of 90%) (34). This study was limited to only 10 cases of early-onset PE. However, the results are in agreement with data from Romero et al. (38), who observed that PP-13 appears to be an interesting marker for risk assessment of preterm PE but a weak marker for severe PE at term, and ineffective for identifying mild PE at term. Also, data from Spencer et al. (39) suggested that first trimester serum PP-13 as a single marker in a low-risk population showed poor predictive performance. We were unable to find published data on first trimester PP-13 in combination with Doppler parameters in a low-risk population.
We observed that first trimester or early second trimester PAPP-A, inhibin A, and activin A in combination with uterine artery Doppler measurements may provide good predictive performance for PE in both low- and high-risk populations. Data from recent publications have indicated that these biochemical markers, taken individually, can predict up to 42% of early PE, but up to 89% when combined with maternal characteristics (race, body mass index, parity) and uterine artery PI (at a false-positive rate of 10%) in a low-risk population (32). PAPP-A and PlGF were recently found to be significantly decreased during first trimester [0.53 multiple of the median (MoM) and 0.61 MoM respectively] in women destined to experience early-onset PE, whereas they were only mildly diminished in women who developed late PE. When these variables were combined in logistic regression algorithms for early PE, a predictive rate of 94% at a false-positive rate of 5% was observed (35). Therefore, a growing body of evidence suggests that combination of first-trimester biochemical and ultrasonographic markers may be useful in prediction of early PE; the addition of maternal characteristics to the prediction algorithms might further improve their performance (33)(35)(36)(37).
This systematic review has several strengths: (a) an exhaustive coverage of existing literature from sound and highly recognized databases, (b) an assessment of articles not limited to those written in English, and (c) a quality assessment of studies based on recognized criteria (30)(31). On the other hand, we also recognize several limitations: (a) the definition of PE has changed over time and between studies using different national guidelines, introducing heterogeneity in the classification of the syndrome; (b) variability in the inclusion criteria, further increasing the heterogeneity between studies; (c) study-to-study variation in the criteria defining level of risk of a given population: some studies on “low-risk” populations showed incidence rates of PE that were higher than “high-risk” populations included in other studies, (d) a multiplicity of potential tests, test combinations, and timings of sampling during pregnancy, which hampered detailed analysis; and (e) a lack of precise information in several studies, flawed study design and/or conduct, although it is unclear how such flaws may have affected the results of a systematic review or a metaanalysis of diagnostic accuracy (40)(41). For each given study on a combination of markers, it would have been beneficial to have ready access to the screening performance characteristics of the markers, either individually or in combination. The absence of such data led us to remove numerous studies from the core analysis. Extraction of data from ROC curves and provision of sensitivity data based on a given level of specificity (e.g., 80% or 95%) limit the ability to fully evaluate test accuracy. Among the greatest weaknesses of the included studies, affecting the interpretation of accuracy data, were the lack of a systematic approach in the reporting of performance assessment of markers, alone or in combination, and the absence of a standardized definition of the syndrome.
Our systematic review reveals a real need to develop experimental approaches that would allow comparison of candidate biomarker combinations. Such information would likely improve statistical power and our capability for finding a screening procedure that can better identify at-risk women. Larger prospective studies including more patients with clearly specified outcomes would likely lead to more meaningful conclusions. Of importance, studies performed during the first trimester could guide the assessment of new prophylactic interventions (such as acetylsalicylic acid) in at-risk women (23). In such studies, specific care should be taken to (1) describe the population (including risk level, means of recruitment, follow-up) and methodological aspects (including a clear and recognized definition of PE, adequate identification of outcomes in accordance with internationally recognized guidelines, and thorough description of the tests); (2) differentiate between mild and severe disease as well as early and late onset, as these factors confer differences in terms of maternal and fetal morbidity and mortality; and (3) report all data related to single markers and report all data with similar cutoffs to facilitate between-study comparisons. The above criteria should be fulfilled, as stated by WHO in their systematic review on screening tests for PE (24) and by the Standards for Reporting Diagnostic Accuracy (STARD) initiative (42), on reporting studies of test diagnostic accuracy.
Furthermore, it has been shown recently that the inclusion of maternal parameters, namely body mass index, ethnicity, and previous maternal medical and obstetrical history, in prediction algorithms further improves predictive accuracy of combinations of biochemical and ultrasonographic markers (35)(36)(37). The addition of such easily accessible variables that increase sensitivity, and thus decrease false-negative results, is of great value. Numerous papers have been published on potential biomarkers for identifying women predisposed to development of PE before the onset of clinical symptoms. As our understanding of PE improves, new tests that will contribute to better predictive performance characteristics of a PE-risk model need to be developed. In addition, this systematic review underscores the need for well-conducted investigations that integrate various candidate parameters (clinical, biochemical, ultrasonographic) into algorithms capable of identifying women at risk of PE early in pregnancy so that early preventive measures can be instituted.
From the perspective of integrative medicine, there is a clear need for prospective large-scale studies with rigorous study design criteria to determine the clinical usefulness of combinations of biomarkers in different geographic and healthcare environments. In conducting such studies, it will be essential to incorporate a universally agreed-on definition of hypertensive disorders of pregnancy. The ultimate outcome of research efforts like these will be the development of an efficient screening procedure that is evidence based and that uses a multivariate algorithm of selected maternal characteristics and measurements of biochemical and ultrasonographic markers to identify women at risk for PE who would benefit from early targeted preventive interventions.
Author Contributions: All authors confirmed they have contributed to the intellectual content of this paper and have met the following 3 requirements: (a) significant contributions to the conception and design, acquisition of data, or analysis and interpretation of data; (b) drafting or revising the article for intellectual content; and (c) final approval of the published article.
Authors’ Disclosures of Potential Conflicts of Interest: Upon manuscript submission, all authors completed the Disclosures of Potential Conflict of Interest form. Potential conflicts of interest:
Employment or Leadership: None declared.
Consultant or Advisory Role: None declared.
Stock Ownership: None declared.
Honoraria: None declared.
Research Funding: The project “Healthy Pregnancy” is supported by a grant from the Canadian Institutes for Health Research [Institute of Human Development, Child and Youth Health (IHDCYH)].
Expert Testimony: None declared.
Role of Sponsor: The funding organizations played no role in the design of study, choice of enrolled patients, review and interpretation of data, or preparation or approval of manuscript.
1 95th%, 95th percentile; UtA, uterine artery; βhCG, β hCG; NS, not significant; D/S, diastolic/systolic; RI, resistance index; NA, not applicable; MAP, mean arterial pressure; MMP-9, matrix metalloproteinase 9.
2 Maternal factors included ethnicity, body mass index, and history of PE.
3 Maternal factors included ethnicity, parity, and history of PE or chronic hypertension.
4 Maternal factors included ethnicity, parity, body mass index, and history of PE.
5 Maternal factors included ethnicity, parity, body mass index, weight, and history of PE.
1 sVEGFR-1, soluble vascular endothelial growth factor receptor 1; PAI, plasminogen activator inhibitor; ADMA, asymmetric dimethylarginine; CRP, C-reactive protein; TNF-R1, tumor necrosis factor receptor 1; SGA, small for gestational age.
Parts of this manuscript have been presented as a poster at the 18th IFCC-EFCC European Congress of Clinical Chemistry and Laboratory Medicine, Innsbruck, Austria, June 7–11, 2009.
↵1 Nonstandard abbreviations: PE, preeclampsia; IUGR, intrauterine growth restriction; MeSH, medical subject heading; QUADAS, quality assessment of diagnostic accuracy studies; PPV, positive predictive value; NPV, negative predictive value; hCG, human chorionic gonadotropin; AFP, α-fetoprotein; PAPP-A, pregnancy-associated plasma protein A; PlGF, placental growth factor; PI, pulsatility index; ADAM12, a disintegrin and metalloproteinase 12; PP-13, placental protein 13; MoM, multiple of the median; STARD, Standards for Reporting Diagnostic Accuracy; sFlt1, soluble fms-like tyrosine kinase 1.
- © 2010 The American Association for Clinical Chemistry