Study participants
We used data from the Avon Longitudinal Study of Parents and Children (ALSPAC), a prospective birth cohort study based in Southwest England [
23‐
25]. The women invited to participate in this study were pregnant with an expected delivery date between April 1, 1991, and December 31, 1992, and living in one of the three Bristol-based health districts. A detailed description of this study is available elsewhere [
23‐
25].
The initial number of pregnancies enrolled was 14,541. When the oldest children were approximately 7 years of age, an attempt was made to increase the initial sample with eligible children who did not join the study originally. This resulted in an additional 913 children being enrolled. Therefore, the total sample size was 15,454 pregnancies. Of these 14,901 were alive at 1 year of age. In the three decades since enrolment, ALSPAC has used questionnaires completed by both parents and children, routine medical data and research clinics as methods of follow-up. The clinics took place when the participants were 7, 9, 10, 11, 13, 15, 17 and 25 years old [
23‐
27].
Ethical approval for the ALSPAC study was obtained from the ALSPAC Law and Ethics Committee and Local Research Ethics Committees. Informed consent for the use of data collected via questionnaires and clinics was obtained from participants following the recommendations of the ALSPAC Law and Ethics Committee at the time. The study website contains details of all the data that is available through a fully searchable data dictionary and variable search tool:
http://www.bristol.ac.uk/alspac/researchers/our-data/ [
28].
Covariates
We selected potential confounders a priori and used a directed acyclic graph (DAG) to encode this causal knowledge of this research question. In summary, we have included only covariates in our model, which we believe to be common causes of the exposure and outcome and have excluded any variables that might be potential mediators of the association [
38]. Therefore, we considered the following as possible confounders of the association between aPHV and cardiovascular structure and function at age 25 years; maternal age, gestational age at birth, household social class, maternal education, mother’s partner’s education, breastfeeding of baby until 3 months, parity, birthweight, maternal body mass index (BMI), maternal marital status, maternal smoking status during first 3 months of pregnancy, and height and DXA-determined fat mass at age 9.
Maternal age was reported in the mother’s antenatal questionnaires. Gestational age at birth was estimated from clinical records. Household social class was measured as the highest of the mother’s or her partner’s occupational social class using data on job title and details of occupation collected about the mother and her partner from the mother’s questionnaire at 32 weeks gestation. Social class was derived using the standard occupational classification (SOC) codes developed by the UK Office of Population Census and Surveys and classified as I professional, II managerial and technical, IIINM non-manual, IIIM manual and IV&V part skilled occupations and unskilled occupations.
A questionnaire at 32 weeks gestation asked mothers to report on educational attainment, which was categorised as below O-Level (Ordinary Level; exams taken in different subjects usually at age 15–16 at the completion of legally required school attendance, equivalent to today’s UK General Certificate of Secondary Education), O-Level only, A-Level (Advanced-Level; exams taken in different subjects usually at age 18) or university degree or above.
Breastfeeding information was collected via questionnaires administered at 4 weeks, 6 months and 15 months. Parity was defined as the number of previous pregnancies that had resulted in a live or stillborn infant collected at 18 weeks gestation. Birthweight was extracted from medical records. Maternal height and weight data were self-reported from a questionnaire administered at 12 weeks gestation; these were used to calculate maternal BMI. Marital status was obtained from antenatal questionnaires and classified as never married, married and widowed, divorced, or separated. Smoking in the first trimester of pregnancy was self-reported by mothers at 18 weeks gestation; responses to smoking any tobacco (cigarettes, cigars, pipes, or other) were grouped as follows: no smoking, < 10 per day, 10–19 per day or greater than 19 per day, and were re-categorised as a dichotomous variable (smoking: yes/no).
Height and fat mass of offspring was measured at clinics at age 9 years. Standing height was measured to the last complete mm using the Harpenden Stadiometer. Fat mass (in kg, less head) was derived from whole body DXA scans performed using a GE Lunar Prodigy (Madison, WI, USA) narrow fan beam densitometer.
Statistical analysis
Statistical analysis was performed using Stata MP 14.2. Linearity of association between aPHV and cardiovascular structure and function in males and females was assessed by comparing fit of models regressing carotid intima-media thickness, left ventricular mass index and relative wall thickness, pulse wave velocity, and systolic blood pressure on fourths of aPHV (treated as a continuous exposure) to models regressing carotid intima-media thickness, left ventricular mass index and relative wall thickness, pulse wave velocity and systolic blood pressure on fourths of aPHV (treated as a categorical exposure); models were then formally compared using a likelihood ratio test. Data on aPHV and all cardiovascular outcomes were normally distributed. We used linear regression to examine the association between aPHV and each measure of cardiac structure and function. All analyses were stratified by sex to examine whether associations of aPHV and each outcome differed for males and females.
Dealing with missing data
There were 1197–2193 participants with complete data for aPHV, each outcome and all covariates. Cardiac structure and function were also measured at the 18 year ALSPAC clinic. To increase efficiency and minimise selection bias, we used multivariate multiple imputation to impute missing data on covariates and outcomes in all participants that had a measure of aPHV and had a measure of the outcome at the age 18 year or age 25 year research clinic (
N = 4339 for carotid intima-media thickness,
N = 2752 for left ventricular mass index,
N = 2776 for relative wall thickness,
N = 3964 for pulse wave velocity and
N = 4571 for systolic blood pressure). We carried out 20 cycles of regression switching and generated 20 imputation datasets [
39]. We then examined associations of aPHV and outcomes in these multiple imputed datasets; results are averaged across the results from each of these 20 datasets using Rubin’s rules, taking account of uncertainty in the imputation so that the standard errors for any regression coefficients (used to calculate 95% confidence intervals) take account of uncertainty in the imputations and uncertainty in the estimate [
39]. See Additional file
1: Table S1 for a list of the variables included in the multiple imputation models and how they were entered into the models.
We repeated the main adjusted analysis in the observed dataset, among participants with complete-case data on exposure, outcome and covariates (N = 1199 for carotid intima-media thickness, N = 1197 for left ventricular mass index, N = 1203 for relative wall thickness, N = 1394 for pulse wave velocity and N = 2193 for systolic blood pressure).