Methods
The current study draws on data collected as part of a prospective
cohort study (STOP), which aims to develop screening tests to identify
adverse pregnancy outcomes. Healthy nulliparous pregnant women were
recruited at two major maternity hospitals, the Lyell McEwin Hospital,
the tertiary hospital serving a lower socio-economic community in
Adelaide’s Northern suburbs and the Women’s and Children’s Hospital, the
primary tertiary maternity hospital for complex care in metropolitan
Adelaide, South Australia. Between March 2015 and December 2017,
nulliparous women with a singleton pregnancy attending their first
antenatal clinic between 9+0 and
16+0 weeks’ gestation were enrolled as part of the
prospective cohort study described elsewhere.19, 20Women were excluded if they were considered already at high risk of
pregnancy complications at screening (i.e. experienced three or more
previous miscarriages or with pre-existing hypertension or diabetes).
Participants were followed prospectively, with vaccination, pregnancy,
and birth outcome data collected by research midwives. As pertussis
vaccination was recommended to be administered between 28-32 weeks’
gestation, we restricted the
analyses to data from women whose pregnancies reached at least 32 weeks’
gestation to allow all women to have had the opportunity to receive the
pertussis vaccine.
The exposure of interest was pertussis-containing vaccine
(tetanus-diphtheria-acellular pertussis: dTpa). A research midwife
interviewed and collected maternal vaccination status of the women
during their study visit interview at 32-36 weeks’ gestation. Maternal
vaccination date and gestation at administration were recorded.
Following delivery, a research midwife verified final vaccination status
by reviewing Pregnancy-Hand-Held-Records and interviewing women.
Pregnancy-Hand-Held-Records are the primary medical record of pregnancy
care in South Australia.
Pregnancy outcomes assessed were gestational hypertension (GH),
preeclampsia (PE), chorioamnionitis and/or funisitis, pre-delivery
hospitalisation due to acute respiratory infections or influenza-like
illness, premature rupture of membranes (PPROM), placental abruption,
spontaneous preterm birth and preterm birth. Birth outcomes included
small for gestational age (SGA), low birthweight (< 2500g)
(LBW), LBW at term (≥ 37 weeks’ gestation), Apgar scores at 1 and 5 min,
neonatal care unit admissions, respiratory distress and mechanical
ventilation.
Pregnancy and birth complications were diagnosed using the Brighton
Collaboration consensus list of terms, 21 and The
Global Alignment of Immunization Safety Assessment in Pregnancy (GAIA)
project.22 GH was defined as hypertension [systolic
BP (SBP) ≥ 140 mmHg or diastolic BP (DBP) ≥ 90 mmHg] after 20 weeks of
gestation in previously normotensive women. PE was defined as GH with
proteinuria (24 h urinary protein ≥ 300 mg or spot urine protein:
creatinine ratio ≥ 30 mg/mmol creatinine or urine dipstick protein ≥ 2+)
or any multi-organ complication of PE, including SGA age infant.
Suspected Chorioamnionitis was considered to be present only with a
physician’s diagnosis, which was dependent on maternal fever ≥ 38 °C,
with at least two of the following: maternal tachycardia, fetal
tachycardia, uterine tenderness, foul odour of amniotic fluid, or
maternal leucocytosis or increased CRP. Funisitis is a histopathologic
diagnosis, and it is the extension of infection or inflammation to the
umbilical cord. Preterm birth (PTB) was defined as any birth after
20+0 and before 37+0 weeks of
gestation. SGA was defined as neonates with a birthweight <
10th percentile customized for maternal factors
including maternal height, booking weight, ethnicity and gestational age
at delivery.
During the first study visit at 9–16 weeks’ gestation, information was
obtained regarding baseline socio-demographic, lifestyle and clinical
characteristics including age, ethnicity, level of education, household
income, employment, exercise, smoking, supplement use, intake of alcohol
and recreational drugs, medical and obstetric history, and complications
during the current pregnancy. Participants also completed questionnaires
assessing stress levels in the past month (Perceived Stress Scale
(PSS-10)) 23, current anxiety symptoms (short form of
the Spielberger State–Trait Anxiety Inventory (STAI))24 and depressive symptoms during pregnancy (Edinburgh
Postnatal Depression Scale (EPDS)). 25
Demographic, lifestyle and clinical characteristics of participants were
summarized descriptively by pertussis vaccination exposure during
pregnancy. Continuous variables were summarized as mean with standard
deviation (SD) or median with interquartile range (IQR), as appropriate,
while counts and percentages were used to summarize categorical
variables. To assess if there was an association between maternal
pertussis vaccination status and each of the outcome variables, we
initially conducted independent samples t-tests and Mann-Whitney U
tests, as appropriate, for continuous variables and chi-square tests of
association for binary and categorical variables.
The timing for vaccination exposures and time-at-risk windows were
calculated for each time-sensitive pregnancy and birth outcome,
accounting for the temporal nature of each outcome of interest. For
instance, women were at risk for PTB from 20+0 until
36+6 weeks of gestation but had to attain at least 32
weeks’ gestation for inclusion in the analysis data set. Cox
proportional-hazards models with gestational age in weeks as the
underlying time metric were used to derive hazard ratios (HRs) that
compared the hazard rates between vaccinated and unvaccinated women for
time-sensitive outcomes. Vaccination status was treated as a
time-varying exposure in these models, in that each vaccinated woman’s
pregnancy was divided into unvaccinated and vaccinated exposure periods.
Thus, a woman who did not receive the vaccine during pregnancy was
classified into the unvaccinated group in any risk set, whereas a woman
who received the vaccine at some point during her pregnancy was
initially classified as unvaccinated and then classified as vaccinated
from the time at vaccination onwards.
We used log-binomial models to estimate risk ratios (RR) and adjusted
risk ratios (aRR) comparing the risk of late onset or early postpartum
adverse pregnancy outcomes and time-independent birth outcomes between
vaccinated and unvaccinated mothers. A multivariable linear regression
model was applied to compare the difference in mean gestational age at
delivery and mean birthweight by maternal pertussis vaccination status.
For all multivariable models, age, level of education, ethnicity,
gravidity, annual household income, alcohol intake, recreational drug
use, smoking, pre-pregnancy body mass index (BMI), use of micronutrient
supplements, asthma, assisted reproductive treatment and current
psychological states were amongst the variables selected as potential
confounders based on evidence in the literature11-14guided by directed acyclic graphs. Sensitivity analyses were conducted
in all multivariable models to evaluate whether the effects of maternal
pertussis vaccination on pregnancy and birth outcomes were maintained
after adjustment for receipt of maternal influenza vaccination. The
overall missing covariate data at baseline was <5% and
therefore all available data were used in the analyses of all
pre-specified outcomes. For all analyses, p values < 0.05 were
considered statistically significant. We did not correct for multiple
comparisons to minimize the risk of Type II errors. Data were recorded
in a REDCap 26, 27 online database and all statistical
analyses were conducted using Stata version 15. Written informed consent
was obtained from all women. Personal identifying information in the
study database was eliminated to ensure that confidentiality of all
patients’ records was maintained. The STOP study protocol was approved
by the Human Research Committee of the Women’s and Children’s Hospital
Adelaide Australia (HREC/14/WCHN/90).19