External validity of first-trimester algorithms in the prediction of pre-eclampsia disease severity

Transcription

1 Ultrasound Obstet Gynecol 2014; 44: Published online 7 August 2014 in Wiley Online Library (wileyonlinelibrary.com). DOI: /uog External validity of first-trimester algorithms in the prediction of pre-eclampsia disease severity N. OLIVEIRA*, L. E. DOYLE*, R. O. ATLAS*, C. B. JENKINS, M. G. BLITZER* and A. A. BASCHAT* *Department of Obstetrics, Gynecology and Reproductive Sciences, University of Maryland School of Medicine, Baltimore, MD, USA; Department of Obstetrics & Gynecology, Mercy Medical Center, Baltimore, MD, USA; Department of Obstetrics & Gynecology, MedStar Harbor Hospital, Baltimore, MD, USA; Departments of Pediatrics, University of Maryland School of Medicine, Baltimore, MD, USA KEYWORDS: pre-eclampsia first-trimester pre-eclampsia screening; HELLP syndrome; maternal outcomes; neonatal outcomes; severe ABSTRACT Objective To compare disease features in women with pre-eclampsia between those who are correctly identified (true ) and those who are missed (false ) when applying first-trimester prediction algorithms for pre-eclampsia to a prospectively enrolled population. Method Six first-trimester early (requiring delivery < 34 weeks gestation) pre-eclampsia algorithms were applied to a prospective cohort of singleton pregnancies enrolled at first-trimester screening. Maternal outcomes, neonatal outcomes and severity parameters for pre-eclampsia were compared between true- and false- predictions. Results Twenty of 2446 (0.8%) women developed early pre-eclampsia, with 65% of these developing severe features and 20% HELLP syndrome. At enrollment, true- cases were more likely to be African American and chronically hypertensive, while false- cases were more likely to be Caucasian. At delivery, true- cases were more likely to have pre-eclampsia superimposed on hypertension, severely elevated blood pressure and creatinine level > 1.1 mg/dl. - cases were more likely to have HELLP syndrome (all P < 0.05). Conclusion In an urban population with a high prevalence of chronic hypertension, patients who are correctly identified by first-trimester screening models are more likely to develop pre-eclampsia superimposed on chronic hypertension with severely elevated blood pressure and evidence of renal failure. In contrast, patients who are missed by these algorithms are more likely to have HELLP syndrome. Further research is needed to confirm these findings and the algorithm adjustments that may be necessary to better predict pre-eclampsia phenotypes. Copyright 2014 ISUOG. Published by John Wiley & Sons Ltd. INTRODUCTION Pre-eclampsia affects 2 3% of pregnancies and is a leading cause of maternal and perinatal mortality and morbidity 1. The maternal risks are potentiated by the presence of end-organ complications such as renal disease and hemolysis, elevated liver enzymes and low platelet count 2 7. Prediction algorithms that incorporate multiple contributing risk factors can predict the development of pre-eclampsia as early as the first trimester of pregnancy, offering the potential benefits of prophylactic intervention and enhanced surveillance In recognition that the phenotype of pre-eclampsia differs across gestational stages, these algorithms are modeled to predict either early-onset disease (requiring delivery before 34 weeks gestation) or late-onset disease occurring after this cut-off. Before the uniform application of these prediction rules, determination of their external validity is critical to assure comparable performance in the general population. In a previous study 21, we evaluated the external validity of six published first-trimester early pre-eclampsia algorithms by applying their prediction rules to a cohort of pregnant women who had been prospectively enrolled in the first trimester. This analysis evaluated algorithms for their ability to predict disease characterized by gestational age at delivery. While we showed that the algorithms underperformed compared with their performance described in the original publications, we did not evaluate algorithm performance in predicting the Correspondence to: Dr A. A. Baschat, Department of Gynecology and Obstetrics, Johns Hopkins University School of Medicine, 600 North Wolfe Street, Phipps 287, Baltimore, MD 21287, USA ( aabaschat@hotmail.com) Accepted: 30 May 2014 Copyright 2014 ISUOG. Published by John Wiley & Sons Ltd. ORIGINAL PAPER

2 First-trimester prediction of severity of pre-eclampsia 287 severity of disease. It was the aim of this study to compare the disease features of women with pre-eclampsia who were correctly identified with those that were missed by the previously evaluated six algorithms. METHODS This was a secondary analysis of a prospective, observational study that had the primary aim of developing a first-trimester predictive model for pre-eclampsia. For the primary study women presenting with a singleton pregnancy at 9 14 weeks gestation and who were willing to participate, signed written informed consent forms. Subsequently, maternal characteristics, fetoplacental ultrasound parameters and maternal biomarkers were ascertained as previously described 16,21,22. Pregnancy outcome was ascertained by study personnel and verified by source documentation. We identified six publications that described formulae for the prediction of early pre-eclampsia that were derived from studies with patients and methods that matched ours 12,13,15,17,18,20,21. Applying these prediction rules to our population we identified true- patients, who were correctly identified by the prediction algorithms at the cut-off defined by the Youden index. Using the same cut-off we identified false- cases, i.e. women who were considered to be at low risk but who nevertheless developed pre-eclampsia. In this study we analyzed the maternal and fetal outcomes in the true- and false- screened patients who did not receive first-trimester aspirin. We excluded patients who received first-trimester aspirin, since the effect of aspirin could potentially modify presentation of the disease 23. Pre-eclampsia was defined as new-onset or worsening proteinuria and systolic blood pressure 140 mmhg and/or diastolic blood pressure 90 mmhg on two separate occasions 6 h apart after 20 weeks gestation. Pre-eclampsia superimposed on chronic hypertension was defined as worsening blood pressure and increasing proteinuria after 20 weeks gestation. The severity of pre-eclampsia was based on the highest maternal blood pressure and worst maternal blood test at delivery. Systolic and diastolic blood pressure 160 or 110 mmhg, respectively, was considered severe 24. Thrombocytopenia was defined as a platelet count of < /μL. Transaminase elevation twice the normal range was defined as abnormal ( 72 IU/L for alanine aminotransferase (ALT) and/or 104 IU/L for aspartate aminotransferase (AST). A serum creatinine level > 1.1 mg/dl was considered as evidence of acute renal injury. The diagnosis of HELLP syndrome was established when at least one of three criteria was present: platelet count < /μL, elevated plasma level of transaminases (AST and ALT > 70 IU/L) and lactic dehydrogenase enzyme > 600 IU/mL. Placental abruption was defined as premature separation of the normally implanted placenta before delivery, clinically manifested by antepartum hemorrhage and a retroplacental blood clot identified at delivery. A neonate was considered small-for-gestational age (SGA) when birth weight was less than the 10 th percentile 25. The neonatal outcome parameters analyzed were: stillbirth, admission to neonatal intensive care unit, 5-min Apgar score < 7, cord ph < 7.2 and birth weight < 3 rd percentile. Based on these outcome parameters the phenotype of maternal disease was categorized as mild pre-eclampsia, severe pre-eclampsia, HELLP syndrome, pre-eclampsia associated with acute kidney injury, pre-eclampsia superimposed on chronic hypertension or pre-eclampsia associated with SGA. Statistical analysis The Mann Whitney U-test and the Pearson χ 2 test were used to compare continuous and categorical variables, respectively. The statistical software package SPSS 14.0 (SPSS Inc., Chicago, IL, USA) was used for all data analysis and P < 0.05 was considered statistically significant. RESULTS Among the 2446 enrolled patients, 20 (0.8%) developed early pre-eclampsia. The outcomes observed in our population are described in Table 1. 65% of these patients met the criteria for severe pre-eclampsia, 20% for HELLP syndrome and 10% developed acute kidney injury. The mean gestational age at birth was 31 weeks, with 20% delivering before 28 weeks. There was a 5% rate of placental abruption and a 5% stillbirth rate. 45% of patients delivered an SGA neonate, 67% of them with a birth weight below the 3 rd percentile. Depending on the parameters required by the algorithm, 678 of the 2446 enrolled women met the applicability criteria for the six formulae. In these subsets the prevalence of early pre-eclampsia ranged from 0.8% to 1% (6 20 patients). The true- rate ranged from 40% to 83% and the false- rate between 5% and 17%. The distribution of enrollment characteristics stratified by true- and false- cases is presented in Table 2. s for two algorithms were more likely to be of African American ethnicity and have a history of hypertension with higher mean arterial blood pressures (Table 2). In true- cases the mean uterine artery pulsatility index (UtA-PI) was significantly higher for four algorithms while the lowest UtA-PI was higher for three algorithms. The distribution of biomarkers was similar in true- and false- cases (Table 2). The distribution of outcomes stratified by true-s and false-s is presented in Table 3. s were more likely to have severe pre-eclampsia as well as significantly higher mean arterial blood pressure during delivery. s for three algorithms were more likely to have pre-eclampsia superimposed on chronic hypertension. s for three algorithms had higher AST levels at delivery and overall a higher frequency of AST levels over 72 IU/mL. Owing to the small sample size of individual cells many of the differences in proportional distribution did not reach statistical significance (Tables 3 and 4).

3 288 Oliveira et al. Table 1 Disease characteristics of 20 pre-eclamptic patients requiring delivery before 34 weeks gestation n (%) or median (range) Phenotype Mild pre-eclampsia 7 (35) Severe pre-eclampsia 13 (65) HELLP syndrome 4 (20) Pre-eclampsia with renal injury 2 (10) Pre-eclampsia with SGA 9 (45) Chronic hypertension and 9 (45) superimposed pre-eclampsia Mean arterial blood pressure (mmhg) ( ) GA at delivery (weeks) 31.6 ( ) GA < 28 weeks at delivery 4 (20) Analytic parameters Hemoglobin (g/dl) 11.4 ( ) Platelets (1000/μL) 201 (46 531) AST (IU/L) 33 (17 622) ALT (IU/L) 25 (10 151) LDH (IU/mL) 292 ( ) Creatinine (mg/dl) 0.90 Proteinuria (g/day) 3.0 ( ) Neonatal outcomes Placental abruption 1 (5) Stillbirth 1 (5) Admission to NICU 14 (70) 5-min Apgar score < 7 3 (15) Cord ph < (15) Birth weight < 3 rd percentile 6 (30) ALT, alanine aminotransferase; AST, aspartate aminotransferase; GA, gestational age; LDH, lactic dehydrogenase; NICU, neonatal intensive care unit; SGA, small-for-gestational age. On evaluation of the distribution of disease characteristics for cumulative prediction categories for all six algorithms, there were 57 true- predictions and 40 false- predictions. This analysis showed that, at enrollment, African American ethnicity or prior hypertension were more common in those with a true- prediction while Caucasian ethnicity was more frequent in women with a false- prediction. At delivery severe pre-eclampsia characterized by higher systolic and diastolic blood pressure, creatinine elevation and pre-eclampsia superimposed on chronic hypertension were significantly more frequent in true- predictions. AST elevation, platelet count < /μL and HELLP syndrome were more frequent among false- predictions. The frequency of abnormal fetal and neonatal outcomes was similar between prediction categories (Table 5). DISCUSSION In a previous study, we evaluated the external validity of first-trimester pre-eclampsia prediction algorithms by applying them to a patient population that was prospectively enrolled in the first trimester 21. In this follow-up study we compare the clinical features of women who were correctly predicted to develop pre-eclampsia with those who were incorrectly predicted to be at low risk. At enrollment women who were more likely to be correctly predicted were African American and have a history of hypertension with measurably higher blood pressure and higher UtA resistance. These women appear to develop pre-eclampsia primarily superimposed on their pre-existing hypertension, exhibiting a phenotype that is characterized by severe hypertension and complicated by renal disease. In contrast, women incorrectly predicted to be at low risk for the development of pre-eclampsia were more likely to be Caucasian, and developed pre-eclampsia with hematologic and hepatic end-organ disease. Fetal and neonatal outcomes did not differ between these prediction categories. Several investigators have reported algorithms capable of detecting pre-eclampsia in the first trimester of pregnancy However, these algorithms focus on the prediction of pre-eclampsia requiring preterm delivery rather than the severity of the disease. Because of the association with maternal morbidity and mortality, the degree of blood pressure escalation, eclampsia, acute renal failure and HELLP syndrome are features that define severe pre-eclampsia 6. HELLP syndrome complicates up to 20% of severe cases of pre-eclampsia and can be associated with 49% morbidity and 24% mortality rates 4. The fact that severe pre-eclampsia tends to produce significant placental disease earlier in pregnancy is demonstrated by the higher rates of fetal growth restriction, placental abruption, stillbirth and prematurity associated with it 1,26. Accordingly, algorithms designed to predict pre-eclampsia requiring delivery before 34 weeks gestation are likely to identify a significant proportion of women exhibiting features of severe disease by virtue of the fact that these are more common earlier in gestation. However, our study suggests that on external validation these algorithms may underperform because of a limited ability to predict specific phenotypes of pre-eclampsia. Consistent with the literature 0.8% of our patients developed pre-eclampsia, two-thirds of whom had severe features and 20% met the criteria for HELLP syndrome. On individual assessment, five of six algorithms predicted pre-eclampsia that arose in African American women superimposed on chronic hypertension and where severity was defined by the degree of blood pressure escalation and renal involvement. All models were more likely to miss pre-eclampsia in Caucasian women and were unable to predict HELLP syndrome. The preferential ability to predict pre-eclampsia superimposed on chronic hypertension is most probably explained by the central importance of hypertension and mean arterial blood pressure in all algorithms and the high prevalence of these factors at enrollment in our population 27. The underperformance in the prediction of HELLP syndrome may occur because of a pathological mechanism that is less dependent on blood pressure than on the mechanism leading to renal failure. Other features of placental dysfunction such as stillbirth, placental abruption and the degree of fetal growth restriction were equally distributed among true s and false s. This is most probably explained by the lack of risk factors such as prior abruption or

4 First-trimester prediction of severity of pre-eclampsia 289 Table 2 Maternal characteristics at enrollment stratified by prediction category using six first-trimester prediction algorithms for pre-eclampsia (PE) Parra-Cordero 12 Scazzocchio 15 Poon 13 Poon 18 Odibo 17 Caradeux 20 (n = 4) (n = 13) (n = 11) (n = 9) (n = 5) (n = 1) Age (years) 29 (19 35) 29 (22 43) 29 (20 41) 33 (19 43) 29 (21 41) 32 (19 43) 28.5 (20 38) 32 (19 43) 31 (25 35) (22 38) 31.5 (19 43) Ethnicity White 0 1 (25) 0 2 (28.6) 0 2 (22.2) 0 2 (28.6) 0 1 (100) 1 (12.5) 1 (8.3) Black/African 8 (100) 1 (25)* 12 (92.3) 2 (28.6)* 10 (90.9) 4 (44.4) 11 (91.7) 2 (28.6)* 5 (100) 0 5 (62.5) 9 (75) South Asian 0 1 (25) 0 1 (14.3) 0 1 (11.1) 0 1 (14.3) (8.3) Indian/Pakistani (14.3) 0 1 (11.1) 0 1 (14.3) (8.3) South American 0 1 (25) 1 (7.7) 1 (14.3) 1 (9.1) 1 (11.1) 1 (8.3) 1 (14.3) (25) 0 History of chronic hypertension History of diabetes mellitus 5 (62.5) 1 (25) 9 (69.2) 0* 9 (81.8) 0 9 (75) 0* 5 (100) 0 6 (75) 3 (25) 4 (50) 2 (50) 6 (46.2) 1 (14.3) 5 (45.5) 2 (22.2) 4 (33.3) 2 (28.6) 3 (60) 0 3 (37.5) 4 (33.3) Obstetric history Nulliparous 4 (50) 3 (75) 7 (53.8) 5 (71.4) 6 (54.5) 6 (66.7) 7 (58.3) 5 (71.4) 2 (40) 0 5 (62.5) 7 (58.3) Parous with previous PE BMI (kg/m 2 ) 37.9 MAP (mmhg) 101 (84 110) Lowest UtA-PI 1.28 Mean UtA-PI 1.78 ( ) PAPP-A MoM 0.9 ( 1.8) PlGF MoM (0.2 1) PP-13 MoM ( 1.3) 1 (12.5) 0 1 (7.7) 1 (14.3) 1 (9.1) 1 (11.1) 1 (8.3) 1 (14.3) 1 (20) 0 1 (12.5) 1 (8.3) 24.8 (23 27)* 88 (83 111) 8 ( ) 0.95 ( )* 0.8 ( ) (87 111) ( ) 0.9 ( ) ( ) 0.9 ( 1.3) 24.5 (23 36) 88 (81 96) 0.96 ( ) 1.17 ( ) ( ) (0.4 ) (93 111) ( ) 1 ( ) ( ) ( 1.3) 27.4 (23 40) 88 (81 96)* 1.00 ( )* 1.24 ( )* ( ) (81 111) ( ) 1.0 ( ) ( ) ( 1.3) 27.4 (23 36) 89 (83 96)* 0.96 ( )* 1.17 ( )* (.2 2.6) (.2 1.4) ( ) ( ) 1.2 ( ) 0.5 ( ) ( (23 32) (88 111) (5 2.23) ( ) ( ) ( 1.3) 35.3 * 92 (81 110) 1.30 ( ) 1.65 ( ) 0.8 (0.2 1) ( 1) Only the first author of each study is given. Data shown as n (%) or median (range). *P < P < BMI, body mass index; MAP, mean arterial blood pressure; MoM, multiples of the median; PAPP-A, pregnancy-associated plasma protein-a; PlGF, placental growth factor; PP-13, placental-protein 13; UtA-PI, uterine artery pulsatility index.

5 290 Oliveira et al. Table 3 Distribution of maternal and neonatal outcomes according to prediction category using six first-trimester prediction algorithms for pre-eclampsia (PE) Parra-Cordero 12 Scazzocchio 15 Poon 13 Poon 18 Odibo 17 Caradeux 20 (n = 4) (n = 13) (n = 11) (n = 9) (n = 5) (n = 1) PE phenotype Mild 1 (12.5) 3 (75) 3 (23.1) 4 (57.1) 2 (18.2) 5 (55.6) 4 (33.3) 3 (42.9) (25) 5 (41.7) Severe 7 (87.5) 1 (25) 10 (76.9) 3 (42.9) 9 (81.8) 4 (44.4) 8 (66.7) 4 (57.1) 5 (100) 1 (100) 6 (75) 7 (58.3) HELLP 0 1 (25) 1 (7.7) 3 (42.9) 0 4 (44.4)* 0 4 (57.1)* 0 1 (100) 1 (12.5) 3 (25) Renal injury 1 (12.5) 0 2 (15.4) 0 2 (18.2) 0 2 (16.7) 0 1 (20) 0 1 (12.5) 1 (8.3) PE superimposed on chronic hypertension Highest MAP at delivery (mmhg) 5 (62.5) 1 (25) 9 (69.2) 0* 9 (81.8) 0 9 (75) 0* 5 (100) 0 6 (75) 3 (25) 131 Maternal lab. parameters at delivery Hb (g/dl) 11.0 (8 12.6) Platelets (1000/μL) 234 ( ) AST (IU/L) 37 (17 96) 110* ( ) (71 224) 45 (30 622) (8 12.8) 233 ( ) 29.5 (17 96) 111 ( (46 195)* 76 (30 622)* ALT (IU/L) 39 (12 87) 31 (10 77) 25 (10 87) 61 (10 151) LDH (IU/mL) 370 ( ) Cr (mg/dl) 0.95 (9 1.33) Proteinuria (g/day) 2+ (trace Delivery outcome GA at delivery (weeks) 29.7 ( ) (4 0.92) 2+ ( (3 32.3) (trace 30.9 ( ) (4 0.92) 2+ (trace 31.9 ( ) 135 ( ) 10.8 (8 12.8) 246 ( ) 27.5 (17 46) 111 (85 131)* (46 195) 76 (30 622)* 23 (10 70) 61 (10 151)* (trace 31.4 ( ) (4 0.94) 2+ ( ( ) (8 13) 224 ( ) 31 (17 46) ( ) (46 195)* 95 (30 622)* 24 (12 70) 80 (10 151)* (trace 31.1 ( ) (4 0.92) 3+ ( ( ) 141 ( ) 10.9 (8 12.6) 277 ( ) 28 (17 46) ( ) (8 12.8) (43 430) (20 287) 25 (12 70) (10 151) 358 ( ) 0.98 ( ) 2+ (trace 30.4 ( ) (4 1.31) (trace ( ) Delivery < 28 weeks 3 (37.5) 0 3 (23.1) 1 (14.3) 2 (18.2) 2 (22.2) 3 (25) 1 (14.3) 1 (20) 0 1 (12.5) 3 (25) Placental abruption 0 1 (25) 0 1 (14.3) 0 1 (11.1) 0 1 (14.3) (8.3) Stillbirth 1 (12.5) 0 1 (7.7) 0 1 (9.1) 0 1 (8.3) 0 1 (20) 0 1 (12.5) 0 5-min Apgar score < 7 3 (38) 0 3 (23.1) 0 3 (27.3) 0 3 (25) 0 2 (40) 0 2 (25) 1 (8.3) Cord ph < (37.5) 0 3 (23.1) 0 1 (9.1) 2 (22.2) 2 (16.7) 1 (14.3) 1 (20) (25) SGA < 10 th percentile 5 (62.5) 2 (50) 6 (46.2) 3 (42.9) 4 (36.4) 5 (55.6) 5 (41.7) 4 (57.1) 3 (60) 1 (100) 2 (25) 7 (58.3) SGA < 3 rd percentile 2 (25) 2 (50) 3 (23.1) 3 (42.9) 2 (18.2) 4 (44.4) 3 (25) 3 (42.9) 1 (20) 1 (100) 1 (12.5) 5 (41.7) Admission to NICU 7 (87.5) 4 (100) 10 (76.9) 4 (57.1) 8 (72.7) 6 (66.7) 8 (66.7) 5 (71.4) 4 (80) 1 (100) 5 (62.5) 9 (75) ( )* 191 (71 531) 35 (17 622) 40 (10 87) 324 ( ) ( ( ) Only the first author of each study is given. Data shown as n (%) or median (range). *P < P < ALT, alanine aminotransferase; AST, aspartate aminotransferase; Cr, creatinine; GA, gestational age; Hb, hemoglobin; lab., laboratory; LDH, lactic dehydrogenase; MAP, mean arterial blood pressure; NICU, neonatal intensive care unit; SGA, small-for-gestational age.

6 First-trimester prediction of severity of pre-eclampsia 291 Table 4 Pre-eclampsia characteristics stratified by prediction category using six first-trimester prediction algorithms for pre-eclampsia Parra-Cordero 12 Scazzocchio 15 Poon 13 Poon 18 Odibo 17 Caradeux 20 (n = 4) (n = 13) (n = 11) (n = 9) (n = 5) (n = 1) SBP 160 mmhg 7 (87.5) 1 (25) 10 (76.9) 3 (42.9) 9 (81.8) 4 (44.4) 8 (66.7) 4 (57.1) 5 (100) 1 (100) 6 (75) 7 (58.3) DBP 110 mmhg 4 (50) 0 5 (38.5) 1 (14.3) 5 (45.5) 1 (11.1) 5 (41.7) 1 (14.3) 3 (60) 0 3 (37.5) 3 (25) Creatinine > 1.1mg/dL 1 (12.5) 0 2 (15.4) 0 2 (18.2) 0 2 (16.7) 0 1 (20) 0 1 (12.5) 1 (8.3) AST 72 IU/L 2 (25) 1 (25) 1 (7.7) 3 (42.9) 0 4 (44.4)* 0 4 (57.1)* 0 1 (100) 1 (12.5) 3 (25) ALT 104 IU/L (14.3) 0 1 (11.1) 0 1 (14.3) (12.5) 0 Platelets < /μL 0 1 (25) 0 2 (28.6) 0 2 (22.2) 0 2 (28.6) 0 1 (100) 1 (12.5) 1 (8.3) Only the first author of each study is given. Data are presented as n (%). *P < 0.05, Fisher s exact test. ALT, alanine aminotransferase; AST, aspartate aminotransferase; DBP, diastolic blood pressure; SBP, systolic blood pressure. Table 5 Maternal and disease characteristics stratified by cumulative prediction categories - predictions (n = 57) - predictions (n = 40) P Enrollment characteristics African American 51 (89.5) 18 (45.0) < Caucasian 1 (1.8) 9 (22.5) History of hypertension 43 (75.4) 4 (10.0) < History of diabetes mellitus 25 (43.9) 11 (27.5) Nulliparous 31 (54.4) 26 (65.0) Parous with history of pre-eclampsia 6 (10.5) 4 (10.0) 1.0 Disease characteristics Mild pre-eclampsia 12 (21.1) 20 (50.0) Severe pre-eclampsia 45 (78.9) 20 (50.0) Chronic hypertension and superimposed pre-eclampsia 43 (75.4) 4 (10.0) < SBP 160 mmhg 45 (78.9) 20 (50.0) DBP 110 mmhg 25 (43.9) 6 (15.0) Creatinine > 1.1mg/dL 9 (15.8) 1 (2.5) AST 72 IU/mL 4 (7.0) 16 (40.0) < ALT 104 IU/mL 1 (1.8) 3 (7.5) Platelets < /μL 1 (1.8) 9 (22.5) HELLP 2 (3.5) 16 (40.0) < Placental abruption 0 5 (12.5) Stillbirth 6 (10.5) SGA < 10 th percentile 25 (43.9) 22 (55.0) SGA < 3 rd percentile 12 (21.1) 18 (45.0) min Apgar score < 7 15 (26.3) 1 (2.5) Cord ph < (17.5) 6 (15.0) Admission to NICU 42 (73.7) 29 (72.5) Data are presented as n (%). P-values reflect chi-square or, for cell sizes less than 5, Fisher s exact test. ALT, alanine aminotransferase; AST, aspartate aminotransferase; DBP, diastolic blood pressure; NICU, neonatal intensive care unit; SBP, systolic blood pressure; SGA, small-for-gestational age. prior stillbirth that are more specific for these complications in the prediction algorithms 28,29. Only models that included pregnancy-associated plasma protein-a, an independent marker for fetal growth disorders, were able to predict over 30 50% of cases of severe fetal growth restriction 17,18. These findings further suggest that algorithms that rely on multifactorial associations between markers and outcomes are likely to underperform in capturing the whole spectrum of complications, while inclusion of specific predictors is likely to perform better. It remains to be determined if one algorithm will be capable of predicting all pre-eclampsia-related complications or if algorithms for specific risk factors are required. The strength of our study is the detailed prospectively verified outcome information that allowed us to define accurately the pre-eclampsia phenotype. This allowed us to expand our prior external validation analysis to investigate performance with respect to organ-specific severity features. The limitation is the small sample size of individual complications, which arose because of the rarity of these complications. While we attempted to compensate by performing a distribution analysis on the cumulative prediction, this approach is not necessarily reflective of findings that may be observed in a large sample size. Despite these limitations our study emphasizes that further research to improve first-trimester algorithms is required. In their current form there is a preferential prediction of blood pressure-related events, with relative underperformance in identifying women at risk for other significant end-organ disease. Whether this

7 292 Oliveira et al. is owing to the mathematical modeling or differences in pathological mechanism requires further study. In conclusion, in an urban population with a high prevalence of chronic hypertension, patients who are correctly identified by first-trimester pre-eclampsia screening models are more likely to develop pre-eclampsia superimposed on chronic hypertension with severely elevated blood pressure and evidence of renal failure. In contrast, patients that are missed appear to be more likely to have HELLP syndrome. ACKNOWLEDGMENTS This study was supported by a grant from Diagnostic Technologies Limited and PerkinElmer. REFERENCES 1. Scazzocchio E, Figueras F. Contemporary prediction of preeclampsia. Curr Opin Obstet Gynecol 2011; 23: Publications Committee, Society for Maternal-Fetal Medicine, Sibai BM. Evaluation and management of severe preeclampsia before 34 weeks gestation. Am J Obstet Gynecol 2011; 205: Abilgaard U, Heimdal K. Pathogenesis of the syndrome of hemolysis, elevated liver enzymes, and low platelet count (HELLP): a review. Eur J Obstet Gynecol Reprod Biol 2013; 166: Baxter JK, Weinstein L. HELLP syndrome: the state of the art. Obstet Gynecol Surv 2004; 59: Haram K, Svendsen E, Abildgaard U. The HELLP syndrome: clinical issues and management. A review. BMC Pregnancy Childbirth 2009; 9: Arulkumaran N, Lightstone l. Severe pre-eclampsia and hypertensive crises. Best Pract Res Clin Obstet Gynaecol 2013; 27: Sibai BM. Diagnosis, controversies, and management of the syndrome of hemolysis, elevated liver enzimes, and low platelet count. Obstet Gynecol 2004; 103: Scholten RR, Hopman MT, Sweep FC, Van de Vlugt MJ, Van Dijk AP, Oyen WJ, Lotgering FK, Spaanderman ME. Co-occurrence of cardiovascular and prothrombotic risk factors in women with a history of preeclampsia. Obstet Gynecol 2013; 121: Akolekar R, Syngelaki A, Sarquis R, Zvanca M, Nicolaides KH. Prediction of early, intermediate and late pre-eclampsia from maternal factors, biophysical and biochemical markers at weeks. Prenat Diagn 2011; 31: Akolekar R, Syngelaki A, Poon L, Wright D, Nicolaides KH. Competing risks model in early screening for preeclampsia by biophysical and biochemical markers. Fetal Diagn Ther 2013; 33: Audibert F, Boucoiran I, An N, Aleksandrov N, Delvin E, Bujold E, Rey E. Screening for preeclampsia using first-trimester serum markers and uterine artery Doppler in nulliparous women. Am J Obstet Gynecol 2010; 203: 383.e Parra-Cordero M, Rodrigo R, Barja P, Bosco C, Rencoret G, Sepúlveda-Martinez A, Quezada S. Prediction of early and late pre-eclampsia from maternal characteristics, uterine artery Doppler and markers of vasculogenesis during first trimester of pregnancy. Ultrasound Obstet Gynecol 2013; 41: Poon L, Karagiannis G, Leal A, Romero XC, Nicolaides KH. Hypertensive disorders in pregnancy: screening by uterine artery Doppler imaging and blood pressure at weeks. Ultrasound Obstet Gynecol 2009; 34: Herraiz I, Arbués J, Camaño I, Gómez-Montes E, Grañeras A, Galindo A. Application of a first-trimester prediction model for pre-eclampsia based on uterine arteries and maternal history in high-risk pregnancies. Prenat Diagn 2009; 29: Scazzocchio E, Figueras F, Crispi F, Meler E, Masoller N, Mula R, Gratacos E. Performance of a first-trimester screening of preeclampsia in a routine care low-risk setting. Am J Obstet Gynecol 2013; 208: 203.e Baschat AA, Magder LS, Doye LE, Atlas RO, Jenkins CB, Blitzer MG. Prediction of preeclampsia utilizing the first trimester screening examination. Am J Obstet Gynecol 2014 Apr 15. pii: S (14) doi: /j.ajog [Epub ahead of print]. 17. Odibo AO, Zhong Y, Goetzinger KR, Odibo L, Bick JL, Bower CR, Nelson DM. First-trimester placental protein 13, PAPP-A, uterine artery Doppler and maternal characteristics in the prediction of pre-eclampsia. Placenta 2011; 32: Poon LC, Stratieva V, Piras S, Piri S, Nicolaides KH. Hypertensive disorders in pregnancy: combined screening by uterine artery Doppler, blood pressure and serum PAPP-A at weeks. Prenat Diagn 2010; 30: Kuc S, Koster M, Franx A, Schielen PC, Visser GH. Maternal characteristics, mean arterial pressure and serum markers in early prediction of preeclampsia. PLoS One 2013; 8: e Caradeux J, Serra R, Nien J, Pérez-Sepulveda A, Schepeler M, Guerra F, Gutiérrez J, Martínez J, Cabrera C, Figueroa-Diesel H, Soothill P, Illanes S. First trimester prediction of early onset preeclampsia using demographic, clinical, and sonographic data: a cohort study. Prenat Diagn 2013; 33: Oliveira N, Magder LS, Blitzer MG, Baschat AA. First-trimester prediction of pre-eclampsia: external validity of algorithms in a prospectively enrolled cohort. Ultrasound Obstet Gynecol 2014; 44: Block-Abraham DM, Turan OM, Doyle LE, Kopelman JN, Atlas RO, Jenkins CB, Blitzer MG, Baschat AA. First-trimester risk factors for preeclampsia development in women initiating aspirin by 16 weeks of gestation. Obstet Gynecol 2014; 123: Roberge S, Giguère Y, Villa P, Nicolaides K, Vainio M, Forest JC, von Dadelzen P, Vaiman D, Tapp S, Bujold E. Early administration of low-dose aspirin for the prevention of severe and mild preeclampsia: a systematic review and meta-analysis. Am J Perinatol 2012; 29: American College of Obstetricians and Gynecologists; Task Force on Hypertension in Pregnancy. Hypertension in pregnancy. Report of the American College of Obstetricians and Gynecologists Task Force on Hypertension in Pregnancy. Obstet Gynecol 2013; 122: Alexander GR, Kogan MD, Himes JH U.S. singleton birth weight percentiles for gestational age by race, Hispanic origin, and gender. Matern Child Health J 1999; 3: Gruslin A, Lemyre B. Pre-eclampsia: fetal assessment and neonatal outcomes. Best Pract Res Clin Obstet Gynaecol 2011; 25: Steegers EA, von Dadelszen P, Duvekot JJ, Pijnenborg R. Pre-eclampsia. Lancet 2010; 376: Akolekar R, Bower S, Flack N, Bilardo CM, Nicolaides KH. Prediction of miscarriage and stillbirth at weeks and the contribution of chorionic villus sampling. Prenat Diagn 2011; 31: Tikkanen M. Etiology, clinical manifestations, and prediction of placental abruption. Acta Obstet Gynecol Scand 2010; 89: