Objective
We sought to evaluate the effectiveness of an integrated first-trimester screening test to predict preeclampsia (PE).
Study Design
A prospective cohort of singleton pregnancies underwent routine first-trimester screening from 2009 through 2011 (n = 5759). A logistic regression-based predictive model for early- and late-onset PE was constructed based on: maternal characteristics; levels of pregnancy-associated plasma protein-A and free β-human chorionic gonadotropin at 8-12 weeks; and blood pressure and uterine artery Doppler at 11.0-13.6 weeks.
Results
Of the 5170 enrolled participants, 136 (2.6%) developed PE (early PE: 26 [0.5%]; late PE: 110 [2.1%]). At 5% and 10% false-positive rates, detection rates were 69.2% and 80.8% for early PE (area under the curve, 0.95; 95% confidence interval, 0.94–0.98) and 29.4% and 39.6% for late PE (area under the curve, 0.71; 95% confidence interval, 0.66–0.76), respectively.
Conclusion
First-trimester screening combining maternal factors with uterine artery Doppler, blood pressure, and pregnancy-associated plasma protein-A is useful to predict PE in a routine care setting.
Preeclampsia (PE) occurs in approximately 2-8% of pregnancies. In developed countries, PE is the primary cause of maternal admission to intensive care units and causes approximately 15% of all pregnancy-related deaths. Additionally, PE is associated with an increased risk of perinatal mortality and is the cause for approximately 10% of stillbirths and 15% of preterm births.
In recent years, the results of several studies indicate that a combination of maternal history, biochemical markers, and biophysical markers effectively predicts PE in the first trimester of pregnancy, a period in which prophylactic strategies with aspirin have been demonstrated to be more effective. The performance in the prediction of early-onset PE, which is the clinical form that contributes most significantly to adverse outcomes, is substantially higher than that for late forms of the disease. Thus detection rates (DR) for early PE ranged from 41% when using combinations of maternal history with pregnancy-associated plasma protein-A (PAPP-A), to 96%, when maternal factors, uterine artery (UtA) Doppler, and angiogenic factors were combined. In contrast, the DR for late PE ranges around 31–45%.
One gap in the current literature on the prediction of PE is that most studies have been performed in Anglo-Saxon populations and have been carried on under similar research settings. Another study carried out on a Mediterranean population, with smaller sample size, did not differentiate early and late PE. Thus there is a need to confirm the effectiveness of first-trimester screening for PE when applied under typical clinical conditions and to populations different from those of the original studies. The composition of the population under study may strongly influence maternal a priori risk. For instance, in south-European countries the proportion of black race and the rates of obesity and chronic hypertension are lower than in the United Kingdom.
In this study, we evaluated the effectiveness of an integrated first-trimester screening test for PE when performed under usual care conditions and in a south-European population. The screening strategy combined maternal history, blood pressure (BP), UtA Doppler, and biochemical markers (free β-human chorionic gonadotropin [fβ-HCG] and PAPP-A). The testing was conducted for 3 years in a clinical setting during routine first-trimester ultrasound and was performed by usual clinical staff. We evaluated the DR for early and late PE in 5759 patients.
Materials and Methods
Subjects
In this study, a prospective cohort composed of singleton pregnancies underwent routine first-trimester screening at the Department of Maternal-Fetal Medicine at Hospital Clinic Barcelona. The local ethics committee approved the study protocol and each patient provided written informed consent. Gestational age in all pregnancies was calculated based on the crown-rump length (CRL) at first-trimester ultrasound. Maternal characteristics and medical history was recorded and BP, UtA Doppler and plasmatic concentrations of PAPP-A, and fβ-HCG were measured in the first trimester.
From May 2009 through October 2011, a total of 5759 women underwent examination. Of these participants, a total of 589 (10.2%) were excluded for the following reasons (nonexclusively): missing outcome data (n = 525), major fetal defects or chromosomopathies (n = 25), miscarriage or fetal death <24 weeks (n = 80), and termination of pregnancy in the absence of medical indication (n = 21). After these participants were excluded, 5170 cases remained.
Predictive variables
Maternal characteristics and medical history were prospectively recorded at the time of first-trimester ultrasound (11.0-13.6 weeks) via a patient questionnaire. Characteristics recorded were: medical and obstetric history, maternal age, ethnicity, smoking status, parity, height, and weight.
A nurse measured BP automatically with a calibrated device (M6 Comfort; Omron Corp, Kyoto, Japan) in our outpatient clinics according to standard procedure. BP was measured in 1 arm (right or left) without distinction while women were seated and after a 5-minute rest. Mean arterial pressure (MAP) was calculated as: diastolic BP + (systolic – diastolic)/3.
UtA evaluation was performed transvaginally during the first-trimester ultrasound, as previously described. Both UtA-pulsatility indices (PI) were automatically measured and mean UtA-PI was calculated.
Maternal serum PAPP-A and fβ-HCG were measured using the DELFIA Xpress analyzer (Perkin-Elmer, Turku, Finland) between 8-12 weeks of gestation. Thereafter, these levels were converted to multiples of the expected normal median (MoM), which were corrected for CRL, maternal age, body mass index (BMI), smoking and diabetes status, and ethnicity according to local references.
Outcome measures
PE was defined as systolic BP ≥140 mm Hg and/or diastolic BP ≥90 mm Hg on at least 2 occasions 4 hours apart, developing >20 weeks of gestation in previously normotensive women, and proteinuria >300 mg in a 24-hour urine specimen. Early PE was defined as PE requiring delivery <34 weeks. Doctors who made the diagnosis were blinded to the study parameters obtained during the first trimester.
Statistical analysis
The Mann-Whitney U test and Pearson χ 2 test were performed to make univariate comparisons of quantitative and qualitative variables, respectively, between groups. A post hoc Bonferroni correction was conducted to maintain a type I error of 0.05 ( P = .025).
Logarithmic transformation was performed to normalize mean UtA-PI and MAP. In 52 cases, (1%) mean UtA-PI could not be calculated and these values were replaced by the average value of the whole cohort. Expected log values were calculated for all cases using a linear regression analysis of unaffected cases and included the following covariables: maternal age at first-trimester ultrasound (years), CRL at first-trimester ultrasound (millimeters), maternal height (centimeters) and weight (kilograms) at examination, parity (nulliparous vs multiparous), smoking status upon examination (0, 1-9, 10-19, and ≥20 cigarettes/d), and ethnicity (white European, South American, black, and other). Each individual observed value was expressed as a MoM of the expected value.
Logistic regression was used to estimate each woman’s a priori risk with respect to the following covariates: medical history of diabetes; chronic hypertension; renal or autoimmune diseases; congenital and acquired thrombophilic conditions; obstetric history of PE or intrauterine growth restriction; maternal age; BMI (kg/m 2 ); smoking (cigarettes/d); and ethnicity and parity.
Logistic regression analysis was performed to estimate the individual risks for early and late PE with respect to the following covariables: a priori risk (log transformed), log MoM mean UtA-PI, log MoM MAP, log MoM PAPP-A, and log MoM fβ-HCG. Receiver operating characteristic curves were performed to analyze model performance, which was expressed as DR for different cutoffs of false-positive rates (FPRs).
In all regression models, stepwise forward algorithms were performed to select variables at a P value cutoff of .05. Goodness-of-fit models were assessed by calculating Nagelkerke R 2 .
The statistical package SPSS 18.0 (IBM Corp, Armonk, NY) was used to conduct all the statistical analyses and graphs were generated with MedCalc (MedCalc Software, Mariakerke, Belgium).
Results
Among the 5170 women included in the study, 136 (2.6%) developed PE, including 110 (2.1%) cases of late PE and 26 (0.5%) cases of early PE. Table 1 shows the epidemiological and clinical characteristics of the population by study group.
| Variable | Unaffected (n = 5034) | Late PE (n = 110) | Early PE (n = 26) |
|---|---|---|---|
| Age, y, median (IQR) | 32 (28–35.4) | 33.2 (29–36.3) | 31.3 (29.9–36.5) |
| BMI, kg/m 2 , median (IQR) | 24 (22.7–24.7) | 24.6 (23.5–26.4) a | 24.4 (22.7–28) |
| Ethnicity, n (%) | |||
| White European | 3757 (74.6) | 73 (66.4) | 15 (57.7) |
| Black | 22 (0.4) | 1 (0.9) | 1 (3.8) |
| South American | 784 (15.6) | 28 (25.5) | 6 (23.1) |
| Other | 471 (9.4) | 8 (7.3) | 4 (15.4) |
| Smoking status, cigarettes/d, n (%) | |||
| 0 | 4637 (92.1) | 100 (90.9) | 24 (92.3) |
| <10 | 107 (2.1) | 4 (3.6) | 0 (0) |
| 10-20 | 245 (4.9) | 4 (3.6) | 1 (3.8) |
| >20 | 45 (0.9) | 2 (1.8) | 1 (3.8) |
| Medical history, n (%) | |||
| Chronic hypertension | 48 (1) | 10 (9.1) a | 4 (15.4) b |
| Diabetes mellitus | 88 (1.7) | 7 (6.4) a | 0 (0) |
| Renal disease | 6 (0.1) | 0 | 3 (11.5) b , c |
| Autoimmune disease | 68 (1.4) | 4 (3.6) | 1 (3.8) |
| Coagulation disorders | 40 (0.8) | 4 (3.6) a | 0 (0) |
| Obstetric history, n (%) | |||
| Nulliparous | 2971 (59) | 70 (63.6) | 14 (53.8) |
| Previous PE | 28 (0.6) | 10 (9.1) a | 5 (19.2) b |
| Previous IUGR d | 28 (0.6) | 1 (0.9) | 3 (11.5) b , c |
| Mean BP, mmHg, median (IQR) | 78.5 (74.1–83.1) | 79.4 (74.9–84.1) | 85.7 (80–89.7) b , c |
| Mean UtA-PI, median (IQR) | 1.67 (0.53–1.25) | 1.68 (1.54–1.84) | 2.23 (1.75–3) b , c |
| Maternal serum biochemistry (MoM), median (IQR) | |||
| PAPP-A | 1.06 (0.53–1.25) | 0.55 (0.28–1.05) a | 0.87 (0.44–1.24) |
| β-HCG | 1 (0.63–1.16) | 0.96 (0.55–1.15) | 0.92 (0.5–1.04) |
a Significant comparison between unaffected and late PE;
b Significant comparison between unaffected and early PE;
c Significant comparison between late and early PE;
Scatterplots showing mean UtA-PI and MAP (in log MoM values) against gestational age at delivery and the correlation between UtA-PI and MAP are provided in the Supplemental Figures in the Appendix .
The following formula best fit the expected log mean UtA-PI: 0.668018 – (0.002772 * CRL) – (0.001536 * height) – (0.001151 * maternal age); R 2 = 4.6%.
The following formula best fit the expected log MAP: 1.803485 + (0.002990 * BMI) + (0.000645 * maternal age) – (0.00421 if South American); R 2 = 13.1%.
Figure 1 shows the distribution in log MoM of mean UtA-PI, MAP, and PAPP-A among the different study groups. The log MoM mean UtA-PI was significantly greater in the early PE group compared to unaffected ( P < .001) and late PE ( P = .001) groups. Similarly, log MoM MAP was significantly higher in the early PE group than in the unaffected ( P = .001) and late PE ( P < .001) groups. Log MoM PAPP-A was significantly lower in the late PE group compared to the unaffected group ( P < .001).
The following model best fit the a priori risk for late and early PE [a priori risk = e y /(1 + e y )]:
Late PE Y = –6.135 + (2.124 if previous PE) + (1.571 if chronic hypertension) + (0.958 if diabetes mellitus) + (1.416 if thrombophilic condition) – (0.487 if multipara) + (0.093 * BMI); R 2 = 7.9%.
Early PE Y = –7.703 + (0.086 * BMI) + (1.708 if chronic hypertension) + (4.033 if renal disease) + (1.931 if parous, previous PE) + (0.005 if parous, no previous PE); R 2 = 13%.
The following models best fit the patient-specific a posteriori risk (a posteriori risk = e y /[1 + e y ]):
Late PE Y = 0.328 + (2.205 * log a priori risk) – (1.307 * log MoM PAPP-A); R 2 = 10.1%
Early PE Y = –0.320 + (2.681 * log a priori risk) + (13.132 * log MoM mean UtA-PI) + (25.733 * log MoM MAP); R 2 = 36.8%.
One example of the application of these models is a 35-year-old woman with a prothrombin gene mutation, but no other medical conditions, who underwent her first pregnancy. At the time of the first-trimester ultrasound (CRL, 65 mm), her height was 165 cm and weight was 65 kg (BMI 23.8 kg/m 2 ). She had a mean UtA-PI of 1.85, a MAP of 90 mm Hg, and a PAPP-A of 0.87 MoM.
The expected log mean UtA-PI is: 0.668018 – (0.002772 * 65) – (0.001536 * 165) – (0.001151 * 35) = 0.194.
The log MoM mean UtA-PI is: log (1.85) – 0.194 = 0.073.
The expected log MAP is: 1.803485 + (0.00299 * 23.8) + (0.000645 * 35) = 1.897.
The log MoM MAP is: log (90) – 1.897 = 0.057.
The a priori odds for early PE is: Y = –7.703 + (0.086 * 23.8) = –5.656.
The a priori risk = e –5.656 /(1 + e –5.656 ) = 0.0034
The a posteriori odds for early PE is: Y = –0.320 + (2.681 * log(0.0034)) + (13.132 * 0.073) + (25.733 * 0.057) = –4.4945.
The a posteriori risk = e –4.4945 /(1 + e –4.4945 ) = 0.011 = 1/91.
The same woman with a mean UtA-PI of 2.5 would have a risk of 1/18 for early PE.
Figure 2 shows the receiver operating characteristic curve for late PE (area under the curve, 0.710; 95% confidence interval, 0.658–0.763) and early PE (area under the curve, 0.96; 95% confidence interval, 0.94–0.98) models. The diagnostic performance for late and early PE for FPRs of 5%, 10%, and 15% is presented in Table 2 . Table 3 shows the DR for early PE for a 5% and 10% FPR of each individual predictor and their combinations.
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
