Statistical analysis
The primary outcome was live birth rate (LBR). Secondary outcomes of the study included clinical pregnancy rate(PR), miscarriage rate (MR) and multiple pregnancy rate (MPR).
The baseline characteristics were compared between the two groups. The normality of continuous variables was examined by normality plots and Shapiro-Wilk test. Since none of the Continuous variables studied demonstrated normal distribution by both tests, they are presented as medians (first quartile, third quartile), while categorical variables are presented as n (%). Continuous variables were analyzed by Mann-Whitney U test, and categorical variables were analyzed using Chi-square test or Fisher’s exact test. All values were two tailed and P<0.05 was considered to be significant. All analysises were performed by using SPSS (version 22, IBM).
As two groups were not randomly assigned in clinical practice, potential confounders and selection biases were accounted for by propensity score matching 15. Propensity scores were calculated using logistic regression based on potential variables related to the outcome16. The variables included maternal age, paternal age, maternal BMI, parity, gravidity, duration of infertility, cause of infertility, baseline FSH, antral follicle count (AFC), ovarian stimulation protocol, insemination methods, endometrial preparation protocol, endometrial thickness, number of blastocyst vitrified, cycles of ET, day of blastocyst transferred and the proportion of using top quality blastocysts. A one-to-one nearest neighbor matching method without replacement was performed to match data between group G and group GP with a caliper width equal to 0.03 17. In order to investigate the effect of group GP in women aged 35 and over and in women received at least 3 cycles of ET, two groups were stratified by age and ET order, and PS matching of each subgroup was performed separately. PS matching was performed by using MatchIt package in R software.
A generalized estimating equations (GEE) model was conducted to evaluate the association between the effect of an additional PQE and outcomes due to including patients contributing multiple cycles 8. To further verify the results, multivariate GEE models were performed using pre-matching data to adjust for aforementioned confounders.