Study population

Samples were collected as part of the Prenatal Exposures and Preeclampsia Prevention 3 Study, a longitudinal cohort study of preeclampsia mechanisms in obese and overweight women that was approved by the University of Pittsburgh institutional review board; all women gave informed consent. Lean women were recruited in smaller numbers to compare any findings in overweight and obese women with and without preeclampsia to findings in normal weight women. Women with preexisting hypertension, diabetes mellitus, renal disease, other medical complications, or multiple gestations were excluded. Women were recruited in early pregnancy from the outpatient clinics of Magee-Womens Hospital in Pittsburgh, PA, and had body composition assessed by bioelectrical impedance in the first, second, and third trimesters (at approximately 10, 20, and 35 weeks of gestation). The outpatient clinics serve primarily low-income, uninsured, unmarried, black, or biracial women; 373 women had complete early-pregnancy (first trimester) data and were eligible for inclusion in this study.

Measurements

Women completed a questionnaire regarding their health behaviors, reproductive history, and demographic characteristics. Standing height and waist and hip circumferences were measured twice for accuracy, and the mean of the 2 values was used. Waist circumference was measured at the natural waist with the center of the navel as a physical landmark. Hip circumference was measured just below the bony prominence of the anterior superior iliac spine. Early pregnancy BMI was calculated from weight and height measurements at the first visit (at 10.3 ± 2.9 weeks of gestation).

Resistance and reactance were measured with a Quantum IV Bioelectrical Impedance Analyzer (RJL Systems, Clinton Township, MI). Measurements were taken with the patient lying supine with arms at a 30-degree angle from the body and with the legs not touching so as not to disrupt the electrical circuit. Electrodes were attached in a tetrapolar arrangement, with 2 electrodes on the dorsal surface of the right foot and 2 electrodes on the dorsal surface of the right hand, 1 proximally and 1 distally. The distal electrodes act as the generating electrodes that transmit a small, painless electrical current; the proximal electrodes receive the electric current and measure the voltage drop between the right hand and right foot.

Body composition calculations

BIA theory estimates total body water (TBW) based on the resistance and reactance measured by a BIA machine and the patient’s height, weight, abdominal circumference, and hematocrit level. TBW during pregnancy was calculated with the equation determined by Lukaski et al ( Table 1 ). TBW was then used to estimate the weight of body fat. We derived an equation for weight of fat mass at any gestational age based on the equations provided by van Raaij et al. Water content of fat free mass was calculated using two separate equations, one for 0 to 10 weeks of gestation and one for 10 to 40 weeks of gestation, that were derived from Figure 1 of van Raaij et al. These equations were validated against deuterium dilution spaces and underwater weighing.

Hematocrit level was measured in blood samples that were obtained by venipuncture.

Preeclampsia definition

We used 2 definitions of preeclampsia. The first matches the current American College of Obstetrics and Gynecology definition when we began the study, in which a woman with previously normal blood pressure has a blood pressure ≥140 and/or ≥90 mm Hg after 20 weeks of gestation and proteinuria; we will refer to this as the “clinical definition” of preeclampsia in the results. In previous studies, we found that a research definition (adds to the American College of Obstetricians and Gynecologists clinical definition an increase of >30 systolic and/or >15 diastolic above blood pressure at <20 weeks of gestation and hyperuricemia that is defined as 1 standard deviation of uric acid concentration above the mean for gestational age) defines a more severe and homogeneous preeclampsia population. Thirty women in the study cohort met the American College of Obstetricians and Gynecologists criteria for “clinical preeclampsia” that was defined earlier; 14 of these women also met the more restrictive “research” definition.

Blood pressure was determined by the average of 5 pressures taken after hospital admission for delivery and before the administration of any medications that would alter blood pressure. Proteinuria was defined as >0.3 g of protein in a 24-hour urine collection, 2+ protein measured by dipstick in a random urine sample, a catheterized urine sample with 1+ protein, or a protein-creatinine ratio >0.3. A jury reviewed the abstracted medical records to determine that criteria for preeclampsia had been satisfied.

Statistical methods

Baseline data were described with the mean ± standard deviation for continuous variables and percentages for categoric variables in the total population and separately by preeclampsia status. Potential differences between women with normal pregnancies or with preeclampsia were evaluated with the use of t tests for continuous variables (equal variances unless otherwise called for; unequal variances test used where appropriate) and χ ² tests for categoric differences (Fisher exact test in cases where expected cell counts were <5). This study focused on early pregnancy BMI and body fat percentage as predictors of preeclampsia outcomes; therefore, first-trimester measurements of BMI and body fat were used in all primary analyses. Body fat percentage was examined as a function of BMI; the Pearson correlation coefficient is presented to assess the linear relationship. Lacking sufficient sample size to test appropriately for interaction between BMI and percent body fat, we instead assessed the relationship between body fat and preeclampsia by testing for differences in percent body fat between women with preeclampsia and healthy control subjects within each of the World Health Organization BMI classifications using t tests (a test with unequal variance where appropriate). Logistic regression models were constructed to analyze BMI and body fat percentage as continuous variables and allow adjustment for a limited selection of potential confounders. Because the Prenatal Exposures and Preeclampsia Prevention 3 study was designed to compare obese women who did or did not experience preeclampsia, we initially limited our logistic regression models only to obese women. We also performed a secondary analysis that included all participants because of the surprisingly high rates of preeclampsia in the lean and overweight women (by the clinical definition, 7.69% of lean women and 6.31% of overweight women experienced preeclampsia). All statistical analyses were performed with SAS software (version 9.4; SAS Institute, Cary, NC); probability values < .05 were considered statistically significant.