Study design and population
The study design is that of a longitudinal cohort study with prospectively collected data originating from a population-based obstetric database. The information gathered regard data on maternal, delivery and infant characteristics from all antenatal, ultrasound, delivery, and postnatal care units in the counties of Stockholm and Gotland in Sweden. Information about maternal reproductive history, lifestyle habits, height, weight, and state of health are usually recorded by midwives at the first antenatal visit and during pregnancy. Biometric measurements used in the study were collected from all ultrasound units in the region, and information about pre-gestational maternal diabetes and chronic hypertension, was obtained from standard delivery charts as well as diagnoses at discharge from the delivery hospital.
Singleton pregnancies (72 309) from the population area of the study with available data on biometric measurements in the first and early second trimester between January 1st, 2008 and October 22nd, 2014 were included in the study sample. Pregnancies with congenital malformations (n= 2 495) [International Classification of Diseases tenth revision (ICD-10) codes Q00-Q99)] or stillbirths (n=197) were excluded, resulting in a final study population of 69 617 singleton pregnancies.
The study exposure regarded differences in fetal biometric parameters, such as the biparietal diameter (BPD), measured by ultrasound scan during the first half of pregnancy. The Swedish Association for Obstetrics and Gynecology (SFOG) dictates that fetal BPD should be measured from the outer edge of the proximal parietal bone to the inner edge of the distal parietal bone at the level of thalami and septum pellucidum and performed by specially trained midwives according to a standardized protocol (8). Firstly, we estimated gestational age based on biometrical measurements obtained at first trimester ultrasound scan (i.e. the combined ultrasound and biochemical screening test, CUB). Next, we related the observed fetal size at the second scan to the expected size estimated on the basis of the first trimester scan. At both first trimester (11+0 to 13+6 weeks of gestation) and second trimester (14+0 to 21+0 weeks of gestation) ultrasound examinations, gestational age was extrapolated from the formula by Selbing (gestational age= 58.65 + 1.07 x BPD + 0.0138 x ((BPD)2). Expected gestational age at second trimester scan was calculated by adding number of days between the two examinations to the observed gestational age at first trimester scan. The difference between observed and expected gestational age at second trimester scan, was expressed in z scores (z = , where μ equals the mean and σ equals the standard deviation). The individual z scores were transformed into centiles and fetuses with growth more than the 90th percentile were defined as having accelerated growth, while fetuses with growth between the 10th and 90th percentile were considered as appropriate and used as the reference group. Fetuses with growth smaller than expected (<10th percentile) were considered to be of delayed growth.
The outcome measure was overall prematurity (i.e. birth at < 37+0 weeks). Furthermore, we analyzed the odds for very preterm birth (birth at <32+0 weeks) and moderate preterm birth (birth between 32+0 and 37+0 weeks). Preterm births with clinician-initiated obstetric interventions, as opposed to spontaneous preterm births, were defined as medically induced preterm births and included prelabor C-sections and induced labors.
Pregnancy characteristics collected from the woman’s medical records regarded maternal age at first antenatal visit (15-35 years or 36-55 years), maternal height (130-154 cm or 155-200 cm), BMI at early pregnancy (< or ≥ 30 kg/m2), primiparity (yes/no), smoking at early pregnancy (yes/no), use of in vitro fertilization (IVF) for the current pregnancy (yes/no), pre-gestational diabetes mellitus (yes/no), chronic hypertension (yes/no) and male fetal gender (yes/no). Pre-pregnancy diabetes mellitus was defined by the ICD-10 diagnosis codes O240, O241, while chronic hypertension was defined as treatment with antihypertensive medication at first antenatal visit and/or ICD-10 diagnosis code indicating chronic hypertension (O10, O11) developed at any time point before the 20th gestational week. ICD codes were provided by the responsible doctor at discharge from the hospital after delivery, while information regarding blood pressure measurements, proteinuria, and medication was provided by midwives at antenatal care or at the hospital before delivery. Information on fetal gender was collected from delivery charts. Maternal country of birth was divided into Sweden, other Nordic countries (i.e., Norway, Denmark, Finland, and Iceland), and non-Nordic countries. Information on paternal characteristics was not available in the database.
The statistical software package SAS 9.4 (version 6.1; SAS, Cary, NC, USA) was used for the statistical analyses and a two-sided p-value below 0.05 was considered statistically significant. Pregnancy characteristics in categorical form were cross-tabulated with both exposure and outcome and compared with the use of chi square test. In order to quantify the difference in growth in the three exposure groups the mean and dispersion index of the difference between observed and expected gestational age at second trimester scan for each of the three groups was calculated. The odds for accelerated growth (>90th percentile) or delayed growth (<10th percentile) were calculated based on discrepancy in fetal growth at early second trimester scan, using fetuses with appropriate growth (10th to 90th percentile) as the reference group. The odds of preterm birth were estimated by binary logistic regression analysis and expressed as crude and adjusted odds ratios (aORs) with 95% confidence intervals (CIs). Adjustments were made for maternal age (< or ≥ 35 years), maternal height (< or ≥ 155 cm), BMI (< or ≥ 30 kg/m2), non-Nordic origin (yes/no), primiparity (yes/no), smoking (yes/no), IVF (yes/no), pre-gestational diabetes mellitus (yes/no), chronic hypertension (yes/no) and male fetal gender. The selection of potential confounders recorded at the first antenatal visit, was based on directed acyclic graphs (DAGs) (Figure S1). Lastly, the multivariable risks and their distribution due to the presence of quantitative covariates were calculated for all categories of preterm birth (i.e. mean predictive value and standard deviation) based on a regression model for delayed, appropriate and accelerated fetal growth.
The study was approved by the Regional Ethical Review Authority in Stockholm, Sweden (Dnr 2014/2179-31/1). The need for written or oral informed consent for participation in the study was weaved since all collected data were depersonalized prior to the analysis.
Patient and public involvement
There was no patient or public involvement in the design, conduct, reporting or dissemination plans of our research study.