Methods
Development cohort
We identified potentially eligible participants from the Oxford
Pregnancy Biobank between April 2009 and April 2010 and examined them
against our inclusion criteria to obtain an objectively healthy
population of pregnant women from which to estimate RIs. The Oxford
Pregnancy Biobank was a prospective study of pregnant women in Oxford,
UK (REC 07/H0607/74), in which pregnancy-specific data were recorded in
each trimester and after delivery. A maternal blood sample was taken and
stored from the first, second and third trimesters at
10-14+6, 18-23+6, and
31-38+6 weeks, respectively.
The inclusion criteria for our study were adult pregnant women (≥18
years old) with uncomplicated pregnancies. We excluded women with
chronic hypertension or diabetes mellitus, and women whose index
pregnancy was complicated by gestational hypertension (blood pressure
≥140/90), pre-eclampsia (hypertension with proteinuria), HELLP syndrome
(haemolysis, elevated liver enzymes, low platelets), acute fatty liver
of pregnancy, gestational diabetes (abnormal oral glucose tolerance
test) or obstetric cholestasis (pruritis with bile acids
>14 µmol/L).
Analysis
Plasma CRP was measured using an immunoturbimetric method on the Abbott
Architect C16000 analyser (Abbott Laboratories Ltd, Maidenhead, UK)
after a single freeze-thaw cycle (-80°C). Inter-assay imprecision
expressed as CV, % was 5.4% at 1.6 mg/L, and 1.0% at 53.6 mg/L,
imprecision was 1.02%. The lower detection limit was 0.2 mg/L.
Statistics
Data on CRP in each trimester were approximated to a normal distribution
using the natural logarithmic transformation:
\begin{equation}
f\left(x\right)=ln(x-k)\nonumber \\
\end{equation}where k was calculated to minimise skewness. Outliers were
identified using Horn’s13 interpretation of the method
described by Hoaglin, et al. 14 Data were binned
into groups of maternal age and BMI, and participants were excluded in
whom CRP was more extreme than 1.5 times the interquartile range above
or below the 3rd or 1st quartiles,
respectively. The transformations were reapplied after excluding
outliers and Shapiro-Wilk tests were used to confirm the distributions.
RIs were calculated for each trimester, equating to the
2.5th and 97.5th percentiles derived
from the mean and standard deviation, and the reference limits were
presented with their respective 90% confidence
intervals.15 One-way ANOVA or Student’s t-tests were
used to investigate differences in CRP according to groups of maternal
age, BMI, mean arterial pressure (at booking), ethnicity, and fetal sex.
Participants with missing data in any variable were excluded from
subgroup analyses as the frequency of missing data was
minimal.16 We ran a multivariate linear regression
model to investigate the association between CRP and BMI, and
investigated partitioning the RIs using the method described by Lahti,et al .17 Summary statistics for normally
distributed data are presented as the mean and standard deviation, or
otherwise as frequencies and proportions. Statistical significance was
assumed at the 5% level, or at the 1% level for subgroup analyses.
Evaluation cohort
We evaluated the diagnostic accuracy of the new upper reference limit
for CRP for detecting infection, using chorioamnionitis as a
well-defined example of infection. We retrospectively identified a
subset of 50 consecutive women who had undergone amniocentesis for
suspected infection, and extracted the maternal serum CRP to calculate
the sensitivities, specificities, positive and negative predictive
values, and areas under the receiving operator curves for infection
using two reference limits:
- CRP ≤7 mg/L (derived from current guidance on suspected sepsis in
pregnancy)10;
- The gestation-specific reference limit estimated from the development
cohort.
The overall diagnostic accuracy of the two reference standards was
compared using receiving operator curve (ROC) analysis. Chorioamnionitis
was defined as microbiological growth or microscopic evidence of
organisms in amniotic fluid, growth on a placental swab, and/or a
histological chorioamnionitis/funisitis if delivery occurred within one
week. Women with uterine contractions or evidence of another infection
were excluded. Where possible, CRP results were extracted from the same
day as the amniocentesis, or within 24 hours.