Statistics
For each of the ten biomarkers the data was log-transformed using the
natural logarithm in order to obtain normal distribution. The mode of
delivery was stratified into three groups; VD, pre-labour CS and
in-labour CS, as classified by obstetrician at the time of birth.
Two-way ANOVA was used to test for overall difference between birth
type, where GA was included in the model as well as the interaction, and
to test for the overall difference between genders, where birth type was
included in the model as well as the interaction. A t-test with
un-pooled variance was used to detect pairwise significant differences
with Holms method (28) to adjust for multiple testing.
One-way ANOVA was used to test for the overall difference of the
biomarkers between gestational age. A pairwise t-test with un-pooled
variance, adjusted for multiple testing using Holms method, was used to
test for differences between the weeks.
Maternal age, BMI, and neonatal birth weight and age at sampling were
initially included in the models, but did not contribute to the
explanation of the variation, and these parameters were therefore
excluded from the final models (data not shown). The birth weight and GA
was lower in the pre-labour CS group. Birth weight was included in the
first calculations, but did not make any differences as long as GA was
included, and thus were excluded in the final calculation.
No difference was found in the use of anaesthetics between pre-labour
and in-labour CS (data not shown), thus we did not correct for this in
the analysis. The analysis was re-run excluding the cases with
pre-labour rupture of the membranes (PROM). No difference in the result
was found (data not shown). Only 0.8% of the infants were registered
with an infection the first 4 days after birth, thus this was not
adjusted for in the statistics.
All statistical analyses and figures were performed using the dplyr (29)
and ggplot2 (30) packages in R software version 3.5.2.