Methods
Study population
Records were reviewed for all women who underwent fresh or frozen
autologous IVF/intracytoplasmic sperm injection (ICSI) cycles at the
Assisted Reproduction Center of Northwest Women’s and Children’s
Hospital, Xi’an, China. A cohort of women with positive serum β- human
chorionic gonadotropin (β-hCG) after embryo transfer between May 2014
and April 2019 were included. Demographical traits, cycle
characteristics, clinical and laboratory data were extracted from
electronic medical records. If a clinical intrauterine pregnancy
resulted, the maximum number of fetal sacs in early pregnancy, the
pregnancy and perinatal outcome were included. Figure 1 shows the
flowchart of patient selection and an overview of the treatment.
The diagnosis of PCOS was re-evaluated according to the Rotterdam
criteria with satisfying at least two of the three
criteria:11 oligo-/anovulation (a cycle length
>35 days or variation between consecutive menstrual cycles
of >10 days); clinical or biochemical hyperandrogenism;
ultrasound diagnosis of polycystic ovary morphology.12Patients with other causes of hyperandrogenism and ovulation dysfunction
(congenital adrenal hyperplasia, Cushing’s syndrome and
androgenic-secreting tumors) were excluded. Exclusion criteria also
included patients with recurrent pregnancy loss, uterine malformations,
treatment with preimplantation genetic testing (PGT) and those involving
donor sperms and oocytes.
The baseline characteristics were the female age, body mass index (BMI),
maternal underlying medical conditions, gravidity and parity, history of
prior spontaneous pregnancy loss as well as basal hormone levels. Age
(years) and BMI (kg/m2) were also categorized for the
clarity of data analysis. Three age subgroups were formed: age
<30 years, 30-34 years and ≥35 years. BMI subgroups were:
<25 kg/m2 (normal), 25-29.9
kg/m2 (overweight), ≥30 kg/m2(obesity). Infertility diagnosis (ovulation dysfunction, male and tubal
factor, unexplained and multiple diagnoses) were included as categorical
variables. Cycle characteristics included the insemination type
(conventional IVF, ICSI or a combination of IVF and ICSI), fresh versus
frozen cycle, the number and quality of transferred embryos.
Stimulation and embryos transfer protocol
For a full description of the IVF protocols, luteal phase support, and
laboratory procedures please refer to our previous
publication.13 Embryos were cultured to day 3 or day 5
depending on the number of embryos of good morphological quality on day
3. The strategy of the number of embryos transferred changed gradually
during the study period. From 2014 to July 2018, one or two of the best
quality embryos were transferred into the uterus on day 3 or 5. Since
Sep 2018, offering transfer of single-embryo is the routine in clinical
care. If two embryos were transferred, the quality of the best embryo
was used for analysis. Embryonic cleavage and morphologic appearance
were assessed as described previously.14 In case of
high risk of OHSS or those with elevated serum progesterone levels on
the day of ovulation trigger embryos were electively cryopreserved at
the physician’s discretion and after discussion with the patient.
Embryos that were cryopreserved in ‘freeze-all’ cycles and supernumerary
embryos which were vitrified, were transferred in artificially
supplemented cycles or in natural cycles. The vitrification, warming
procedure, endometrial preparation and embryos transfer procedures was
done according to standard protocols.15 If pregnancy
was achieved, luteal phase support was continued until 10 weeks’
gestation.
Pregnancy assessment and outcomes
Pregnancy was defined as a serum β-hCG level greater than 20 IU/L 14
days after cleavage embryo transfer or 12 days after blastocyst
transfer. If the β-hCG assay yielded a positive result, the patient
underwent ultrasonographic monitoring to determine the number of
gestational sacs and fetal viability at the
6th-7th week of gestation. A
biochemical pregnancy loss was defined as a pregnancy without the
intrauterine gestational sac that resolved spontaneously. Clinical
pregnancy was defined as an intrauterine gestational sac visible by
means of transvaginal ultrasound coincident with a positive serum β-hCG
concentration. The ongoing pregnancy was defined by presence of fetal
heart beat on ultrasound scan at 12 weeks’ gestation. Live birth was
defined as delivery of a live-born infant after 24 weeks’ gestational
age.
The primary endpoint of interest was clinical pregnancy loss which was
defined as a pregnancy ending before 24 weeks of gestation, which were
further categorized based on gestation length: early pregnancy loss (≤13
weeks), late pregnancy loss (13-24 weeks).16,17 We
also examined the rate of pregnancy loss stratified by plurality of the
pregnancy, defined as the number of fetal sacs on early ultrasound.
Pregnancies with two and more than two fetal sacs in the ultrasound were
combined in the analysis because of low numbers. Vanishing twins was
defined as pregnancy with two intrauterine gestational sacs at 6-7
weeks’ gestation but that eventually delivered one infant. Fetuses dying
after ≥24 gestational weeks are registered as stillborn. Ectopic
pregnancies and hydatid moles were excluded from the pregnancy loss
analyses due to their different etiology.
Maternal complications in the analysis included gestational diabetes
mellitus (GDM) diagnosed via the 75 g 2-hour oral glucose tolerance
test,18 hypertensive disorders of pregnancy, including
gestational hypertension or pre-eclampsia, placental abruption and
placenta previa, premature rupture of the membranes (PROM), macrosomia
(birth weight >4,000g). Neonatal outcome variables included
gestational age at delivery, preterm birth (PTB; <32 and
<37 weeks), low birthweight (<1,500 and
<2,500 g) and macrosomia (>4,000 g). With the
intention to eliminate the impact of multiple pregnancies on maternal
and newborn outcomes, we restricted the analysis to only singletons.
Statistical analysis
All statistical analyses were performed using SPSS version 21.0 (IBM
Corp., USA). Categorical data were presented by the number of cases and
corresponding percentage and continuous data were presented as the mean
value ± SD. Categorical data and continuous data that did not show a
normal distribution were analyzed by Pearson’s chi-squared test/Fisher’s
exact test or Kruskal–Wallis test as appropriate. Binary logistic
regression models were used to calculate odds ratios (ORs) and 95%
confidence intervals (CI) of pregnancy loss and to evaluate the effect
of potential confounders. Female age and BMI were recorded as either
continuous or categorical variables. Female age, BMI, number of embryo
transferred, a history of previous pregnancy loss and comorbidities
(hypertension and diabetes), were considered as a covariate potential
confounders for the hypothesized relationships. P -values
<0.05 were considered to indicate statistically significance.