Methods:
The familial risk of PA was estimated through this retrospective cohort
study using the Utah Population Database (UPDB), which is a genealogic
database of over 11 million individuals spanning up to 17 generations.
The UPDB contains individual-level medical and demographic information,
linked to official statewide birth and death records dating back to the
1900s. 17 This study was approved by the Institutional
Review Board at the University of Utah and the Utah Resource for Genetic
and Epidemiologic Research, which oversees studies performed using UPDB
data.
Cases of PA and controls were ascertained from birth certificates (1939
– 2020), maternal death certificates (1904 – 2020), and fetal death
certificates (1978 – 2020), inpatient medical records, and
International Classification of Diseases (ICD) codes (1996 – 2020) from
the University of Utah Health Sciences Center. While PA can occur at any
time point in pregnancy, this diagnosis is often confirmed at the time
of delivery and more mild cases of placental abruption are often missed
on official medical records. Diagnosing PA antepartum is challenging
because of the poor sensitivity of ultrasound imaging and uncertainty on
the clinicians’ part to confirm the diagnosis prior to delivery.
Frequently, a suspected diagnosis of PA will be based on recurrent
antepartum bleeding in the absence of a more obvious explanation.
Therefore, many cases are not captured within our data sources. For this
reason, we included “antepartum bleeding” in the case definition.
Pregnancies affected by placenta previa, multifetal gestations, cases
with inadequate demographic information allowing pedigree linkage, or
cases that were listed as gestational carriers were excluded. Controls
were 3:1 matched to cases based on age, parity, number of eligible
female relatives in the UPDB, and availability of eligible first-degree
relatives (FDRs), second-degree relatives (SDRs), or third-degree
relatives (TDRs) for analyses.
Statistical analysis included relatives of all PA cases and controls.
Demographic and clinical variables were compared using Pearson’s
Chi-squared test for categorical variables and Wilcoxon rank sum test
for continuous variables. Clinical variables described were restricted
to the index pregnancy, i.e., the pregnancy affected by PA, for cases.
Familial risk of PA was estimated using generalized linear mixed-effect
regression for FDRs, SDRs, and TDRs, clustering around cases and
controls to account for non-independence of observations within
families. The analysis included all cases and controls with at least one
eligible relative in the UPDB. Family members of controls were used as
the reference group in all analyses. Variables adjusted for in the
primary analysis included age, year of pregnancy, race, ethnicity,
education level, and parity. A stratified analysis was performed
estimating familial risk of PA stratified by the number of pregnancies
affected by placental abruption (at least 1 vs. 2 or more vs. 0). A
sensitivity analysis was performed excluding cases coded as “antepartum
bleeding,” which excluded 50.4% of cases.
A secondary analysis was performed, which included cases and controls
with available data in the medical record to adjust for clinical risk
factors. This analysis was performed using an alternative familial risk
statistical method, referred to as an “ego-driven” familial analysis
as opposed to “relative-driven”. Ego-driven familial risk analyses
estimate the risk of PA for a case and control depending on whether a
family member was affected. The method allows for adjusting for
important clinical risk factors and medical comorbidities because they
are only required to be available for the case or control, as opposed to
all the relatives in the analysis. Familial risk of PA using this method
was estimated using conditional logistic regression, adjusting for all
demographic characteristics and potentially confounding clinical risk
factors complicating the pregnancy affected by PA. These risk factors
included smoking, drug use, diabetes, uterine anomaly, hypertension, or
pre-labor rupture of membranes (PROM) complicating the pregnancy
affected by PA, and a history of cesarean delivery in prior pregnancies.
Maternal drug use was noted to be co-linear with smoking and was not
adjusted for in multivariable analysis. Clinical risk factors were
selected based on prior published work evaluating risk factors for
PA1,5,7,8 and our univariate analyses. Statistical
analyses were carried out using R version 4.2.2 [R Core Team, Vienna,
Austria].18