Statistical analysis
The
study was initially designed to detect an 8% absolute difference in PE
rate with 80% power and a 5% type I error
rate.18 We assume the
incidence of PE was 4% in the LDA group, while in the control group the
rate was 12%. Accounting for a potential 20% loss to follow-up,
therefore, a sample size of 424 (212 in the LDA group, 212 in the
Control group) was needed. However, the actual loss to follow-up
(including 47 subjects met twin-to-twin transfusion syndrome) of 12.90%
(n=138) was less than the 20% expected and we finally followed up 932
patients (277 in the LDA group and 655 in the control group).
To
adjust the unmatched baseline characteristics, a 1:1 propensity score
matching (PSM) was applied between the LDA group and the control group
to simulate a randomized controlled trial (RCT) of the variables. The
matching algorithm was the nearest-neighbor matching and the caliper was
set at 0.1. The standardized mean difference (SMD) was calculated by
using R (version 2.15.3, the R Foundation for Statistical Computing,
Tsinghua University, China; downloaded at
http://www.R-project.org/). The
value of SMD≤0.1 was considered to be a negligible difference. We also
performed a sensitivity analysis to evaluate the biases caused by
propensity score matching. SPSS version 22.0 software (SPSS Inc,
Chicago, USA) was used for all the statistical
analyses.
We dealt with data via an independent t-test and described as mean ±
standard deviation if variables were in accordance with a normal
distribution. Otherwise, variables were described as mean ± quartile and
examined by the Kruskal–Wallis test. Differences in the classified
variables were presented by percentage and evaluated by the McNemar’s
test. P < 0.05 or SMD>0.1 was considered as
significant. A conditional logistic regression model was used in paired
data (after PSM), while an unconditional logistic model was applied in
the sensitivity analysis. Confounding factors including age, BMI, and
those which were previously reported to be connected with preeclampsia
or were unmatched between LDA and normal groups.