Materials and Methods

The present study is part of the Amsterdam Born Children and their Development (ABCD) study. The ABCD study is a prospective birth cohort study examining the associations between pregnancy conditions and the child’s health in a multiethnic population ( www.abcd-study.nl ). Pregnant women from Amsterdam were approached for their participation between January 2003 and March 2004 when they first visited an obstetric care provider. All (12,373) received a questionnaire by post, which was filled out by 8266 women (67% response rate). These data were completed with information on pregnancy outcome from Youth Health Care Registration and the Dutch Perinatal Registration (PRN).

For inclusion in the current study, women had to participate in the ABCD biomarker study (focus group of 4389 women), births had to be singleton live births, gestational age had to be over 22 weeks, and data on prenatal depression status and birthweight had to be available.

The study protocol of the ABCD study was approved by the Central Committee on Research involving Human Subjects in the Netherlands, the Medical Ethical Committees of participating hospitals, and the Registration Committee of the Municipality of Amsterdam. All participants signed an informed consent.

The independent variable of prenatal depressive symptoms was included in the pregnancy questionnaire, which was filled out around the 16th (interquartile range [IQR], 14th to 18th) week of pregnancy. Depressive symptoms were assessed by the validated Dutch version of the 20-item Center for Epidemiological Studies Depression Scale (CES-D). Three categories were defined: no depressive symptoms (<16), mild depressive symptoms (16-22), and major depressive symptoms (≥23).

The use of categories is widespread in epidemiology, although there is no agreement with regard to definition. We chose to use the same cutoffs that Chung et al used because the first marker of 16 points is very common and because Chung et al stated that these categories are also suitable for a low-income population. The internal consistency of the CES-D scale did not differ between ethnic minority groups (overall Cronbach’s alpha = 0.90). A previous study showed good psychometric properties of the CES-D, translated into Moroccan-Arabic and Turkish, in a Dutch group of Moroccan, Turkish, and Dutch elderly people. The 1% of our population using antidepressants was equally presented in each of the CES-D categories.

Serum folate status was determined when women visited an obstetric care provider for the first time (median at 13 weeks of gestation; IQR, 10.3–15.7). The blood sample was taken in a Vacuette plastic specimen collection tube (Greiner Bio-One BV, Alphen aan de Rijn, The Netherlands), processed at the regional laboratory, and sent to Dr Stein and Colleagues Medical Laboratories (Maastricht, The Netherlands) at –80°C. The samples were sent by overnight mail or courier, enabling processing within 28 hours. This delay did not compromise the validity of the measurement.

Serum folate concentrations were measured with chemiluminescence detection using the Advia Centaur automated instrument (Folate BA method; Bayer Diagnostics, Tarrytown, NY) with an intraassay coefficient of variation of 7% and interassay coefficient of variation of 6%. This method of serum assay has been compared with other available methods by Owen and Roberts.

Serum folate values were standardized for pregnancy duration at the time of blood collection because folate concentration changes during pregnancy. In this study cohort, a regression coefficient describing the progression of serum folate concentrations by pregnancy week was calculated, separately for women who did and did not report taking folic acid supplements. Then, based on the pregnancy duration at the time of blood collection, it was added to or subtracted from the values of women coming in earlier or later than median, resulting in values below zero in some cases. Hereafter these folate values were divided into quintiles.

Gestational age, birthweight, and onset of delivery were available from the Youth Health Care Registration and the PRN systems. Gestational age was based on ultrasound by the obstetric care provider or, when unavailable (<10%), on the first day of the last menstrual period. Onset of delivery encompassed 3 categories: spontaneous, induction, and primary cesarean section. These data were available from the PRN system.

The covariates of maternal age, prepregnancy body mass index (BMI) (kilograms per square meter), educational level, primiparity, ethnicity, smoking, and alcohol consumption were available from the pregnancy questionnaire and thus, were all self-reported. Missing values on maternal weight (9.5%) and height (3.4%) were imputed by means of a random imputation method using linear regression, which accounted for the differences between the ethnic groups. Educational level was included as years of education after primary school, as a measure of socioeconomic status. Ethnicity was based on the mother’s country of birth (definition of ethnicity by the Dutch Central Bureau of Statistics). Primiparity, smoking, and alcohol consumption were dichotomized.

The distinction between preexistent and pregnancy-induced hypertension was made by combining data from the pregnancy questionnaire and the PRN. Preexistent hypertension was defined by either an indication in the PRN or self-reported high blood pressure and/or use of medication before the 20th week of pregnancy. Pregnancy-induced hypertension was defined by either an indication of pregnancy related hypertension/eclampsia/preeclampsia in the PRN or self-reported high blood pressure and/or use of medication after the 20th week of pregnancy. This classification was in accordance with the guidelines of the International Society for the Study of Hypertension in Pregnancy.

Differences in maternal and birth characteristics among the CES-D categories were tested using 1-way analysis of variance with post hoc Tukey tests and χ ² tests. The associations were examined using linear regression analysis (SPSS 17.0; SPSS Inc, Chicago, IL).

The associations of depressive symptoms and folate status with gestational age were first explored univariately (crude model). Subsequently depressive symptoms and folate status were added to a model together (combined model).

Then this model was adjusted for covariates maternal age, prepregnancy BMI, education after primary school, primiparity, ethnicity, smoking, alcohol consumption, hypertension, and onset of delivery (adjusted model). The combined effect of depressive symptoms and folate status was explored by adding an interaction term. The procedure was similar for birthweight analysis, although the confounder gestational age (linear and quadratic) was also added here.

Next, to illustrate the combined effect of both mild and major depressive symptoms and the lowest folate status quintile, 4 new categories were defined (combination of depressive symptoms and the lowest folate quintile; depressive symptoms only; the lowest folate quintile only; and no depressive symptoms and not in the lowest folate quintile [reference]).

Mean gestational age and birthweight values for these groups were first determined using simple descriptive statistics. The adjusted means were calculated from the unianova regression equations (here, the independent variable was the new 4 category variable combining depressive symptoms and folate status). We used average population characteristics for the continuous covariates and the assumptions: no smoking, no alcohol, primiparity, Dutch ethnicity, no hypertensive disorders, and spontaneous onset of birth for the categorical covariates.