Cohort design, participant recruitment, and blood collection

This nested case-control study originated from separate Canadian cohort studies of pregnant women recruited in Halifax, Nova Scotia, and in Quebec City, Quebec. Blood samples were collected <20 weeks of gestation prior to diagnosis of PE.

In Halifax, Nova Scotia, pregnant women attending the blood collection services laboratory for routine prenatal blood screening were invited to participate <20 weeks of gestation, based on self-report and if they were planning on delivering at the Izaak Walton Killam Health Centre as previously described. Gestational age was confirmed using last menstrual period or ultrasound estimates. Blood samples were collected from October 2002 through July 2005. Recruitment into the Quebec City, Quebec, cohort spanned March 2005 through April 2010 at the first routine visit to the hospital between week 14-17 (2005 through 2008) or at the dating ultrasound between week 8-12 (2008 through 2010). In both cohorts, blood samples were sent to the clinical laboratories for procurement and then to the research laboratory for storage at −80°C.

Selection of cases and controls

The study was designed to examine several outcomes, and all cases of pregnancy loss, gestational diabetes, PE, preterm delivery, or having an infant who was low birthweight or small for gestational age were included. For the present analysis, cases were defined as women with PE, defined according to the Canadian Hypertension Society as gestational-onset hypertension with proteinuria (blood pressure >140/90 mm Hg after 20 weeks’ gestation with urinary protein >300 mg/d or urinary dipstick ≥2+).

Noncases were randomly selected from the women without any of the case outcomes with frequency matching to the combined case group on the following factors: study site (Halifax, Nova Scotia, or Quebec City, Quebec), gestational week of recruitment (<10, 10-13, 14-15, 16-17, 18-19), season (October through December, January through March, April through September), and year of blood sample collection, to ensure similar conditions at the time of blood collection between cases and controls. For the present analysis, controls were defined as the noncases plus participants randomly selected from those with the other case conditions at the frequency of these conditions within the entire cohort. Specifically, 219 subjects with a case condition (eg, gestational diabetes, low birthweight, preterm birth, or small for gestational age) were added to the control group (representing 11% of the controls). This was in contrast to 41% of the cases who had 1 of these other conditions. Women with preexisting hypertension were excluded.

After excluding withdrawals and participants with missing essential outcome information (no gestational age at enrollment), the total number of cohort participants delivering live births was 2036 in Halifax, Nova Scotia, and 7184 in Quebec City, Quebec. Women with multiple gestations were subsequently excluded. The final number of participants in both cohorts with available information on 25(OH)D concentration for the current analysis was 2144: 169 cases with PE and 1975 controls.

Questionnaires and medical chart reviews

At both sites, exposure and covariate information was surveyed including maternal age, education level, family income, prepregnancy weight, height, smoking habits, chronic medical conditions, pregnancy history, physical activity, and caffeine intake in early pregnancy. In Halifax, Nova Scotia, the questionnaire was completed during the 20th week of pregnancy and in Quebec City, Quebec, between 18-28 weeks. After delivery, medical charts were reviewed for all participants by trained personnel to collect data on prenatal ultrasound dating, maternal blood pressure readings and status, urinary protein findings, use of medications for hypertension before and during pregnancy, maternal weight at the time of delivery, antenatal hospital admissions, gestational age at delivery, and infant birthweight, sex, and outcome.

Measurement of vitamin D status

Serum samples were shipped on dry ice to McGill University and stored at –80°C until analysis of total 25(OH)D concentration using an automated chemiluminescence immunoassay (Liaison; DiaSorin, Stillwater, MN). Internal quality control measures included duplicate measure of the high and low kit controls and a pooled serum sample from nonpregnant healthy adults. The laboratory participated in the vitamin D external quality assessment scheme program and obtained a certificate of proficiency for 2011 through 2012 and 2012 through 2013, which reflects that ≥80% of the reported results fell within 30% of the all-laboratory trimmed mean. The laboratory also received a quality assurance certified value from the National Institute for Standards and Technology (NIST). Accuracy using NIST vitamin D controls indicated a 3.3% difference of the all-laboratory trimmed mean from the NIST reference measurement procedure in October 2012 and a 6.3% difference in January 2013. The interassay coefficient of variation was 5.7% and intraassay coefficient of variation ranged from 1.4–7.6% with a mean of 3.9% across runs. In addition, 25(OH)D analyzed in 40 samples with both the chemiluminescence immunoassay and the liquid chromatography tandem-mass spectroscopy method were compared yielding a bias of 3.0 ± 12.4 nmol/L.

Statistical analyses

Covariates assessed for confounding were selected based on factors noted in previous research and related to PE and/or vitamin D : maternal age (<25, 25-<30, 30-<35, ≥35 years); maternal prepregnancy body mass index (BMI), categorized in 3 groups of underweight and normal weight (<24.9 kg/m ² ), overweight (25-29.9 kg/m ² ), and obese (≥30 kg/m ² ); marital status (married/common-law, single); maternal highest education qualification (≤high school, some or all college, technical or some or all university); annual family income (<$40,000 and ≥$40,000 Canadian); smoking (nonsmoker and/or ex-smoker who quit before pregnancy vs smoked at some time during pregnancy); parity (nulliparous vs primiparous/multiparous); and physical activity (number of times done for 20-30 minutes in the first half of pregnancy) classified in 6 groups (none, 1 time/mo, 2-3 times/mo, 1 time/wk, 2-3 times/wk, ≥4 times/wk). Other potential confounders included infant sex (female or male); caffeine consumption in early pregnancy, calculated based on the daily number of cups of coffee or tea (0, 0-149, 150-299, and ≥300 mg/d); residence (rural or urban); season of blood collection (spring, summer, autumn, or winter); and year of blood collection (2002 through 2010). Ethnicity of parents, dairy product consumption, and prenatal supplements were only available for the Quebec City, Quebec, cohort and were, therefore, not included in the analyses.

Characteristics of mothers who developed PE and mothers who did not were compared using χ ² or Fisher exact test for categorical variables and Student t test for continuous variables. Logistic regression analyses adjusting for potential confounding factors were used to estimate the risk (adjusted odds ratios [aOR] with 95% confidence intervals [CI]) of PE according to the exposure variable, serum 25(OH)D concentrations <20 weeks of gestation. The cut-offs established by the IOM in 2011 were used to categorize 25(OH)D: <30, 30-49.9, and ≥50 nmol/L; with ≥50 nmol/L set as the referent category. Subgroup analysis based on 25(OH)D >75 nmol/L was not possible as only 9% of controls and 6% of cases were in this category. The following potential prespecified variables were forced in the model: maternal age, smoking, parity, prepregnancy BMI, season and year of blood collection, gestational age at blood collection, and study site (Halifax, Nova Scotia, or Quebec City, Quebec). Other covariates that were associated with PE with a P value < .2 in unadjusted analysis were assessed to determine whether they confounded the relationship between 25(OH)D and PE: if removing the covariate did not change the odds ratio (OR) for the association between 25(OH)D and PE by >5%, it was removed from the adjustment model. Effect modification of vitamin D status with prepregnancy BMI and with smoking was tested using the likelihood ratio test. Then, spline regressions were developed to assess the dose-response relationship between 25(OH)D concentration and PE risk. All analyses were performed with Statistical Analysis Software, Version 9.3 (SAS Institute, Cary, NC).

Ethical approval

The study was approved by the research ethics boards of the Izaak Walton Killam Health Centre in Halifax, Nova Scotia; the Center Hospitalier Universitaire de Québec; and McGill University in Montreal, Quebec. All participants signed informed written consent.

Participants’ characteristics

Of the number of cohort participants (9220: 2036 from Halifax, Nova Scotia, and 7184 from Quebec City, Quebec), 169 (1.8%) developed PE. Table 1 presents the characteristics of the study population by case-control status. There were no significant differences between mothers who developed PE and mothers who did not develop PE in terms of age, marital status, education, family income, infant sex, smoking in pregnancy, caffeine consumption, physical activity, and living in an urban or rural area. There was a higher proportion of obesity, preexisting diabetes, and nulliparity in cases compared to controls. Likewise, while not a potential confounder, women with PE had a higher proportion of caesarean delivery at a lower mean gestational age.

Blood collection information

As shown in Table 2 , at a mean gestational age of 14 weeks, the mean maternal serum 25(OH)D concentration was significantly lower in the group of women who later developed PE than in the control group (47.2 ± 17.7 vs 52.3 ± 17.2, P < .0001) and a higher proportion had 25(OH)D <30 nmol/L.

Stay updated, free articles. Join our Telegram channel

Join