Results
Of the 21820 charts identified with a positive β-hCG, 2357 (10.8%) subjects had PCOS, and 19463 (89.2%) controls did not. After excluding the biochemical pregnancies (n=1382), ectopic pregnancies (n =130) and molar pregnancies (n =3), the study population consisted of 20305 clinical pregnancies (2224 cases with PCOS; 18081 controls without) (Figure 1). Among 2224 intrauterine pregnancies with PCOS, 394 ended in pregnancy loss (17.7 %). 275 (12.4% of total pregnancies) were early pregnancy losses, 119 (5.4%) were late pregnancy losses and 5 (2.7‰) were stillbirths. In pregnancies without PCOS (n=18081), there were 2861 pregnancy losses (12.8% in early stage and 3.1% in late stage) and 18 (1.2 ‰) stillbirths (Figure 2A).
Patient characteristics
Differences in women characteristics were observed between patients with and without PCOS (Table 1). Specifically, women with PCOS were more likely to be younger than the non-PCOS women (29.3±3.3 vs 30.5±4.2). PCOS patients were more often nulliparous than controls (64.7% vs 57.7%, P <.001) and overweight (BMI ≥25kg/m2) and obese (BMI ≥30 kg/m2) were more common in women with than without PCOS (33.3% vs 17.0%, and 9.2% vs 1.7%, P <.001). The prevalence of hypertensive disease and diabetes mellitus were significantly higher (5.9% and 1.2%) in women with PCOS as compared with women with non-PCOS (2.7% and 0.2%). Anovulation was recorded as the indication for fertility treatment in 930 patients (39.5%) with PCOS and male factor being mentioned as an indication in 423 (17.9%). Whereas, tubal and male factor infertility were major causes in women without PCOS (47.8% and 23.9%, respectively). Basal serum FSH level was lower in the those with PCOS as presumed, alternatively, number of antral follicle count and levels of total testosterone and LH were significantly higher adjusting for age. There was no significant difference in having a history of prior pregnancy loss (P = .060) between women with and those without PCOS. In this cohort women, more PCOS cases were treated by conventional IVF compared with of controls (81.2 vs 71.4%). Rate of high quality of transferred blastocysts was slightly higher in the PCOS group than in the control group (55.8 vs 50.0%), however, no significant difference was seen in the quality of embryo in day 3 between two groups
Female age was found to be effect modifier, so we stratified the analysis. Figure 2B displays the age-stratified association between the PCOS and pregnancy loss. We found that regardless of age and PCOS status, the risks of having pregnancy loss were higher in the first trimester. In PCOS group, the risk of pregnancy loss decreased from 12.4 percent in the first trimester to 5.4 percent during the secondary trimester. We observed that younger women (<35 years) with PCOS had higher odds of late pregnancy loss (>13 weeks) when compared with same age group without PCOS (<25 years: 6.9 vs. 2.2 percent; 25-35 years: 5.3 vs. 3.2 percent, PCOS vs. Non-PCOS, respectively). However, no significant difference was seen among those over 35 years of age (4.3 vs. 2.8 percent, P =.337). As expected, in both of the two groups, the risk of early pregnancy loss increased with increasing maternal age (P =.046 andP <.001). However, the change related to maternal age became less significant during the secondary trimester. No significant differences in late pregnancy losses were seen among three age groups (Table 2).
When data were analyzed according to the type of embryo transfer procedure, increased risk of late pregnancy losses was also seen in women with PCOS using either freshly or thawed fertilized embryos (4.6 vs. 2.7 percent, OR 1.71, 95%CI, 1.21-2.40, P =.02; 5.9 vs. 3.3 percent, OR 1.81, 95%CI, 1.41-2.34, P <.001, freshly and thawed fertilized, respectively) (Table 3). No significant difference in early pregnancy loss was found between women with and without PCOS, irrespective of the type of embryo transfer procedure (Figure 2C).
We calculated the risk of early and late pregnancy loss by pregnancy plurality on early ultrasound for the 20305 clinical pregnancies. The number of pregnancy loss, partial pregnancy losses (vanishing co-twins), singletons and twin deliveries are shown in Table 4. The results indicated that there was no significant difference in early pregnancy loss between the PCOS and non-PCOS group (one gestational sac: 15.7% vs 16.0%, OR 0.98, 95%CI, .85-1.14, P=.794; ≥ two gestational sacs: 5.1 vs. 4.8 percent, OR 1.06, 95%CI, .74-1.52, P=.738, PCOS vs. Non-PCOS, respectively). However, after 13 weeks’ gestation, women with PCOS had significantly higher risk of pregnancy loss. For pregnancies with one fetal sac on ultrasound, the risk of loss was increased significantly for patients with PCOS (4.1% vs 2.7%, OR 1.56, 95%CI,1.18-2.05,P =.002). Among pregnancies with two or more fetal sacs, there was a tendency towards a higher incidence of pregnancy losses during the whole gestation in the PCOS group compared with non-PCOS group (93/703 (13.2%) vs 464/5215 (8.9%), OR 1.56, 95%CI,1.23-1.98, P<.001); Women with PCOS had a doubled risk of late pregnancy loss compared with women without PCOS (57/703 (8.1%) vs 212/5215 (4.1%), OR 2.08, 95%CI,1.54-2.82, P <.001).
Logistic regression to assess risk of early and late pregnancy loss
Early pregnancy loss
PCOS status did not influence the risk of early pregnancy loss, not even after adjusting for age and the other confounders (Table 5). Compared with women who were <35 years of age, women aged ≥35 years had higher early pregnancy loss risk (P <.001, 95% CI, 1.65-2.39). Overweight, comorbidities including gestational diabetes and hypertension and single-embryo transfer (SET) were also positively associated with early pregnancy loss. Previous pregnancy loss was also not related to early pregnancy loss after adjusting for cofounders.
Late pregnancy loss
The diagnosis of PCOS was associated with late pregnancy loss in the univariate analysis (OR 1.79, 95% CI, 1.79 (1.46-2.20,P <.001) and in a multivariable regression Model 1 (adjusted OR 1.60, 95% CI, 1.13–2.28, P =.009) (Table 6). However, the effect was reduced into marginal after adjusted for BMI (Model 2), (adjusted OR 1.42, 95% CI, .99-2.03, P = .059). BMI was independently associated with late pregnancy loss in the overweight groups (adjusted OR 1.65, 95% CI, 1.26-2.17, P <.001). Regardless of PCOS, preexisting medical conditions have a significant effect on pregnancy loss rate, either before or after adjustment for confounders. Contrary to the early stage, lower risk of late pregnancy loss was seen in SET when compared to double-embryo transfer (DET) (P =.009, 95% CI, 1.06-1.49), but the effect was reduced to a non-significant level after adjustment for other confounders (P =.152, 95% CI, 0.93-1.56). As gestation developed, maternal age had no significant negative effect on late pregnancy loss (adjusted OR 1.03, 95% CI, .99-1.06, P =.109). Again, a history of prior pregnancy loss was not independently associated with late pregnancy loss.
Obstetric and Perinatal Complications
Among the 1356 live-born singletons in patients with PCOS, 1218 singletons were from singleton gestations and 138 (10.2%) were survivors of a twin conception in early weeks. In pregnancies without PCOS, 10456 singletons resulted from singleton gestations and 1245 (10.6%) were derived from vanishing twins. There is no significant difference in the rate of vanishing twins between groups (P =.600) (Table 7). To facilitate comparison of obstetric and neonatal outcome parameters between pregnancies with PCOS and without PCOS, data analysis was restricted to singleton pregnancies with confirmed outcome. Women with a diagnosis of PCOS had a higher risk of developing hypertensive disorder of pregnancy or pre-eclampsia than women with no such diagnosis (P =.001, P <.001, respectively). PCOS cases had a nearly doubled risk of GMD compared with those without PCOS (12.8 vs 6.9, P <.001). Infants born to mothers with a diagnosis of PCOS were more often suffered from preterm (10.9 vs 8.7,P =.008) and very preterm birth (1.8 vs 1.0, P =.009), but were not at increased risk of LBW and very LBW (5.9 vs 5.3,P =.266, 1.0 vs 0.7, P =.205). The average birthweight in children with PCOS mothers was greater than in controls (3.34 ± 0.57 vs 3.30 ± 0.52, P =.023) and the risk of macrosomia was also significantly increased in former group (11.7 vs 8.4,P <.001).