Keywords
multivariable prediction models, proteomics, metabolomics, cellfree fetal DNA, podocyturia, angiogenic factors, uterine artery Doppler velocimetry
Editors’ comment: As noted by the authors in their Introduction, prediction tests had hardly been studied when Chesley published his singleauthored edition. However, this concept was never far from his mind, and he was amongst the pioneers in the field suggesting urate clearance tests to predict the disease. As noted in the current chapter, this approach proved imprecise. One problem with the prediction literature is its reliance primarily on clinical criteria alone to establish a diagnosis of preeclampsia. As discussed further in Chapter 6 and elsewhere, diagnosis by purely clinical criteria can be erroneous in a considerable number of patients, especially in multiparous women, when histopathological confirmation is sought (e.g,. renal biopsy). Suggestions are made elsewhere in the text that specific biomarkers be sought and utilized in conjunction with classical clinical signs and symptoms; these are likely to aid in the accurate diagnosis of preeclampsia and should enhance the precision of predictive tests .
Introduction
The prediction of preeclampsia is a worthwhile goal because identification of patients at risk could result in earlier diagnosis of the disease, monitoring the mother and fetus at risk, and testing/implementing preventive strategies. In addition, longitudinal studies of patients at risk are a means to gain insight into the pathogenesis of the disease and the development of mechanisticbased strategies for prevention and treatment.
In the singleauthored first edition of this textbook, Chesley foresaw the importance of predictive tests; yet, the available tests were few. Chesley highlighted three possible methods: indices of exchangeable sodium, the flicker fusion test (the point at which a rapidly flickering light is perceived as steady), and a variety of pressor responses ranging from angiotensin infusion to the rollover test. Much work has been conducted since the publication of the first edition of the textbook, and this chapter will review tests which have been proposed for the prediction of the syndrome of preeclampsia.
Assessing the Quality of Tests to Predict Disease
“Prediction” refers to the administration of a test to asymptomatic individuals with the goal of assessing the likelihood of developing a particular disease. The evaluation of predictive tests in preeclampsia is important and necessary for several reasons. Preeclampsia is associated with increased maternal and perinatal morbidity and mortality. Interventions to prevent preeclampsia (e.g., lowdose aspirin and calcium) can have small/moderate benefits that caregivers may wish to offer women at higher risk of developing the disease. Intensive maternal and fetal surveillance could be offered, and earlier interventions may decrease the severity of maternal morbidity.
The properties of a predictive test are traditionally assessed using sensitivity, specificity, and predictive values. Sensitivity is the probability of a test result being positive in a patient who will develop the disease, and specificity is the likelihood of a test result being negative in a patient who will not develop the disease. The performance of a test can also be measured in terms of positive and negative predictive values. Sensitivity and specificity cannot be used to estimate the probability of disease for an individual patient.
Predictive values (positive and negative) allow estimation of the probability of subsequently developing the disorder or not developing the disorder; yet they depend on the prevalence of the disorder in the population. Likelihood ratios are alternative indices for summarizing the performance of a test, and are independent of the prevalence of the disease. Likelihood ratios depend on the inherent value of the test to distinguish between patients who will and will not develop the disease, and are calculated by combining sensitivity and specificity.
The likelihood ratio of a positive test is the ratio of the probability of a positive predictive test result in women who subsequently develop preeclampsia to the probability of a positive predictive test result in women who do not develop the disease (sensitivity/[1−specificity]). The likelihood ratio of a negative test is the ratio of the probability of a negative predictive test result in women who subsequently develop preeclampsia to the probability of the negative predictive test result in women who do not develop the disease ([1−sensitivity]/specificity). The greater the positive likelihood ratio, the larger is the increase in the posttest probability of developing disease, while the smaller the negative likelihood ratio, the larger is the decrease in the probability of developing disease.
Tests with a high positive and a low negative likelihood ratio are considered to be clinically useful. Tests with positive likelihood ratios of 10 or greater and negative likelihood ratios of 0.1 or lower are most likely to be useful in clinical practice. Moderate prediction can be obtained with tests with a likelihood ratio value of 5–10 and 0.1–0.2, respectively, whereas those below 5 and above 0.2, respectively, yield only minimal prediction.
The Bayes theorem permits the use of likelihood ratios in conjunction with pretest probability of preeclampsia to estimate the posttest probability that an individual will develop this disorder once the result of a test is known. The use of odds rather than risk makes the calculation slightly complex; however, a nomogram can be used to avoid having to make conversions between odds and probabilities. The Fagan nomogram is a useful and convenient graphical tool that permits estimation of the posttest probability of disease based on the pretest or anterior probability of disease and the likelihood ratio.
Likelihood ratios indicate by how much a given test result will increase or decrease the probability of developing preeclampsia. To use the Fagan nomogram (as depicted in Fig. 11.1 ), a line is drawn from the estimated pretest probability (left vertical line) through the likelihood ratio of the test result (center vertical line) and the intersection of the line with the right vertical line provides the posttest probability. For example, if a test has a likelihood ratio of a positive result of 15 and a negative likelihood ratio of 0.1, and if a pregnant woman has a pretest probability of 5% and the test is positive, the posttest probability of disease would be 44% (red line). On the other hand, if the test is negative, the posttest probability would be approximately 0.6% (blue line).
When assessing a test during pregnancy, its accuracy is a function of the prevalence of the disease in the population under study. Since incidence rates of preeclampsia for all women in developing (1.3 to 6.7%) and developed countries (0.4 to 2.8%) are relatively low, any predictive test for this disorder would require a very high positive likelihood ratio (>15) to increase the probability that preeclampsia will occur, and a very low negative likelihood ratio (<0.1) to confidently exclude the probability that the woman will develop the disorder. Criteria for the ideal predictive test for preeclampsia are shown in Table 11.1 .

Table 11.2 depicts tests proposed for the prediction of preeclampsia. Such methods have been chosen on the basis of proposed mechanisms of disease implicated in the pathophysiology of preeclampsia.
1. Placental perfusion/vascular resistance dysfunctionrelated tests 
Rollover test 
Isometric handgrip exercise test 
Cold pressor test 
Flickerfusion threshold test 
Pressor response to aerobic exercise 
Intravenous infusion of angiotensin II 
Mean arterial blood pressure 
Assessment of arterial stiffness 
Platelet angiotensin II binding 
Platelet calcium response to arginine vasopressin 
Renin 
24hour ambulatory blood pressure monitoring 
Transcranial Doppler velocimetry 
Ophthalmic artery Doppler velocimetry 
Ultrasonographic placental volume, location, and vascularization 
3D power Doppler of the uteroplacental circulation space 
Uterine artery Doppler velocimetry 
2. Fetal and placental unit endocrinology dysfunctionrelated tests 
Human chorionic gonadotropin 
Alpha fetoprotein 
Estriol 
2Methoxyestradiol 
Inhibin A 
Pregnancyassociated plasma protein A (PAPPA) 
Activin A 
Placental protein 13 (PP13) 
Corticotrophinreleasing hormone 
A disintegrin and metalloprotease 12 (ADAM12) 
Kisspeptin 
Pregnancyspecific β _{1 }glycoprotein (SP1) 
3. Renal dysfunctionrelated tests 
Serum uric acid 
Microalbuminuria 
Urinary calcium excretion 
Urinary kallikrein 
Microtransferrinuria 
N Acetylβglucosaminidase 
Cystatin C 
Podocyturia 
4. Endothelial/oxidant stress dysfunctionrelated tests 
Platelet count and volume 
Fibronectin 
Endothelin 
Neurokinin B 
Prostacyclin 
Thromboxane 
Creactive protein 
Cytokines 
Matrix metalloproteinase9 (MMP9) 
Homocysteine 
Isoprostanes 
Serum lipids 
Ceruloplasmin 
Insulin resistance 
Adiponectin 
Resistin 
Inositol phosphoglycan Ptype 
Antiphospholipid antibodies 
Plasminogen activator inhibitor 
Leptin 
Dimethylarginine 
Pentraxin 3 
Paraoxonase 1 
Tryptophan 
Endothelial cell adhesion molecules (P and Eselectin, vascular cell adhesion molecule1 (VCAM1), and intercellular adhesion molecule1 (ICAM1)) 
Angiogenic factors (placental growth factor; vascular endothelial growth factor; fmslike tyrosine kinase receptor1 [sFlt1]; endoglin; angiopoietin) 
5. Others 
Hematocrit 
Carboxyhemoglobin 
Total proteins 
Antithrombin III 
Magnesium 
Calcium 
Ferritin 
Transferrin 
Haptoglobin 
Atrial natriuretic peptide 
β _{2 }Microglobulin 
High temperature requirement A3 (HtrA3) protease 
25Hydroxyvitamin D 
Thyroid functionrelated tests 
Liver functionrelated tests 
Histidinerich glycoprotein 
Lymphocyte micronucleus (maternal chromosomal damage) 
Hydroxysteroid (17β) dehydrogenase 1 
Testosterone 
Cellfree fetal DNA 
Insulinlike growth factors/insulinlike growth factor binding protein1 
Proteomic, metabolomic and transcriptomic markers 
Combination of tests and maternal characteristics 
We will summarize the accuracy of predictive tests for preeclampsia based on the best available evidence (published and updated systematic reviews and metaanalyses on the topic). In accordance with recent recommendations, the terms “early preeclampsia” or “earlyonset preeclampsia” will refer to preeclampsia that resulted in delivery before 34 weeks of gestation, whereas the terms “late preeclampsia” or “lateonset preeclampsia” will refer to preeclampsia requiring delivery at or after 34 weeks of gestation.