Distinct pregnancy challenges in PCOS women undergoing IVF/ICSI with singleton and twin births: a propensity-matched analysis.

In general, 1017 women with PCOS and 5863 women with tubal factors (control group) met the criteria, respectively (Fig.  1 ). After excluding women with missing data, 921 (90.56%) PCOS women and 5361 (91.44%) controls were included in the 1:2 PSM analysis. Eventually, 696 PCOS women and 1392 controls with singleton birth, and 201 PCOS women and 402 controls with twin births were matched for the final analysis. All baseline covariates had a SMD below 0.1 after matching, indicating a well effectiveness of PSM (Fig.  2 and Table S1). Fig. 2 Covariate balance before and after propensity score matching (PSM). Love plots visually summarized the covariate balance for women with ( A ) singleton births and ( B ) twin births. Abbreviations: APS, antiphospholipid syndrome; ET, embryo transfer; SLE, systemic lupus erythematosus Covariate balance before and after propensity score matching (PSM). Love plots visually summarized the covariate balance for women with ( A ) singleton births and ( B ) twin births. Abbreviations: APS, antiphospholipid syndrome; ET, embryo transfer; SLE, systemic lupus erythematosus Table  1 provided a rather concise overview of population demographics and baseline clinical data, aligning with previous studies or clinical practice. Women with PCOS irrespective of singleton or twin births, demonstrated higher percentages of nulligravida, nullipara, metabolic and endocrine abnormalities than their counterparts. ART characteristics were shown in Table S2a (fresh ET) and S2b (frozen ET), where GnRH antagonist protocol, half ICSI fertilization, and ovarian hyperstimulation syndrome were more common in PCOS women than controls, except for population with twin births following fresh ET. Additionally, among women receiving frozen ET, blastocyst transfers and artificial cycles for endometrial preparation were more prevalent in PCOS population. Table 1 Baseline demographic and clinical characteristics Singleton birth Twin birth PCOS Control P PCOS Control P n  = 696 n  = 1392 n  = 201 n  = 402 Maternal characteristics Ethnicity, n (%) 0.219 0.362  Han 681 (97.84%) 1348 (96.84%) 194 (96.52%) 391 (97.26%)  Others 15 (2.16%) 44 (3.16%) 7 (3.48%) 11 (2.74%) Education, n (%) < 0.001 0.096  High school or lower 387 (55.60%) 901 (64.73%) 116 (57.71%) 255 (63.43%)  University or higher 309 (44.40%) 491 (35.27%) 85 (42.29%) 147 (36.57%) BMI, kg/m 2 22.43 (20.33–24.40) 21.09 (19.49–23.01) < 0.001 22.21 (20.32–24.16) 21.20 (19.52–23.07) < 0.001 Smoking, n (%) 2 (0.29%) 116 (8.33%) < 0.001 0 (0) 44 (10.95%) N/A Hypertension, n (%) 10 (1.44%) 10 (0.72%) 0.077 0 (0) 2 (0.50%) N/A DM, n (%) 14 (2.01%) 10 (0.72%) < 0.001 9 (4.48%) 3 (0.75%) 0.001 Insulin resistance, n (%) 122 (17.53%) 25 (1.80%) < 0.001 45 (22.39%) 5 (1.24%) < 0.001 Hypothyroidism, n (%) 20 (2.87%) 19 (1.36%) 0.038 5 (2.49%) 8 (1.99%) 0.560 Nulligravida, n (%) 418 (60.06%) 582 (41.81%) < 0.001 143 (71.14%) 197 (49.00%) < 0.001 Nullipara, n (%) 640 (91.95%) 1153 (82.83%) < 0.001 191 (95.02%) 361 (89.80%) 0.029 Irregular menstruation, n (%) 563 (80.89%) 185 (13.29%) < 0.001 164 (81.59%) 53 (13.18%) < 0.001 Duration of infertility, year 3 (2–4) 2 (1–4) 0.114 3 (2–4) 2 (1–4) 0.014 Family history of hypertension, n (%) 119 (17.1%) 175 (12.57%) 0.001 35 (17.41%) 53 (13.18%) 0.048 Family history of DM, n (%) 50 (7.18%) 62 (4.45%) 0.001 13 (6.47%) 27 (6.72%) 0.878 Basal reproductive endocrinology  FSH, IU/L 5.80 (5.01–6.73) 6.40 (5.48–7.54) < 0.001 5.70 (5.13–6.65) 6.36 (5.38–7.43) < 0.001  LH, IU/L 8.73 (5.37–13.68) 4.87 (3.64–6.41) < 0.001 8.06 (5.68–12.95) 4.86 (3.74–6.38) < 0.001  LH/FSH ratio 1.46 (0.99–2.20) 0.74 (0.56–0.99) < 0.001 1.35 (0.93–1.90) 0.76 (0.57–1.00) < 0.001  Estradiol, pmol/L 117.40 (85.12–155.25) 104.85 (73.49–141.15) < 0.001 112.20 (80.70–148.80) 96.98 (67.41–129.60) < 0.001  Progestogen, nmol/L 1.19 (0.78–1.76) 1.34 (0.94–1.91) < 0.001 1.16 (0.70–1.62) 1.35 (0.93–1.92) < 0.001 Paternal characteristics Ethnicity, n (%) 0.434 N/A  Han 684 (98.28%) 1373 (98.64%) 201 (100%) 396 (98.51%)  Others 12 (1.72%) 19 (1.36%) 0 (0) 6 (1.49%) Education, n (%) < 0.001 0.242  High school or lower 387 (55.60%) 902 (64.80%) 118 (58.71%) 252 (62.69%)  University or higher 309 (44.40%) 490 (35.20%) 83 (41.29%) 150 (37.31%) BMI, kg/m 2 24.21 (22.18–26.42) 23.66 (21.63–25.95) 0.060 24.00 (21.91–25.95) 23.83 (21.63–26.12) 0.654 Smoking, n (%) 196 (28.16%) 370 (26.58%) 0.394 63 (31.34%) 107 (26.62%) 0.122 Data are presented as number (percentage) or median (interquartile range) Abbreviations : DM Diabetes mellitus, FSH Follicle stimulating hormone, LH Luteinizing hormone, N/A Not applicable Baseline demographic and clinical characteristics Data are presented as number (percentage) or median (interquartile range) Abbreviations : DM Diabetes mellitus, FSH Follicle stimulating hormone, LH Luteinizing hormone, N/A Not applicable Pregnancy and peripartum outcomes were compared between PCOS women and their counterparts, where disparities existed when distinguishing twin or singleton births. Consistency was observed when integrating crude and adjusted results for a prudent interpretation. Among women with singleton births (Table  2 ), PCOS was associated with a higher prevalence of HDP, PE, CIC, and chronic DM (about 40% of cases newly identified during pregnancy). Nonetheless, in addition to inherent adverse outcomes accompanied by twin delivery (Table S3), those women with PCOS demonstrated an elevated incidence of VPTB and GDM than the controls did, along with a lower percentage of placental abruption and small for gestational age (Table  3 ); there was also a tendency for PTB and low birthweight in the PCOS-twin group. The seemingly paradoxical coexistence of the lower percentage of small for gestational age and the tendency towards low birthweight could probably attribute to the occurrence of preterm pregnancy termination among the PCOS-twin population. No significant difference was observed in PPROM between the groups. Table 2 Maternal and neonatal outcomes of women with singleton births Crude model Adjusted OR (95% CI) PCOS Control OR (95% CI) P Model 1 Model 2 Model 3 Maternal outcomes n   = 696 n   = 1392 PTB (< 37 weeks) 87 (12.5%) 158 (11.35%) 1.11 (0.81–1.54) 0.606 1.08 (0.80–1.46) 1.08 (0.80–1.45) 1.02 (0.76–1.36) VPTB (< 32 weeks) 15 (2.16%) 22 (1.58%) 1.39 (0.81–2.39) 0.369 1.35 (0.77–2.36) 1.34 (0.77–2.35) 1.42 (0.80–2.52) Caesarean section 403 (57.9%) 809 (58.12%) 0.99 (0.83–1.19) 0.921 0.98 (0.81–1.19) 0.97 (0.79–1.19) 0.92 (0.74–1.15) Stillbirth 10 (1.44%) 9 (0.65%) 2.35 (0.82–6.73) 0.214 1.94 (0.72–5.24) 1.96 (0.73–5.24) 1.54 (0.65–3.65) HDP 106 (15.23%) 120 (8.62%) 1.91 (1.51–2.42) < 0.001 1.58 (1.24–2.02) 1.58 (1.23–2.01) 1.46 (1.11–1.91)  Chronic hypertension a 27 (3.88%) 18 (1.29%) 3.00 (1.98–5.55) < 0.001 2.70 (1.81–4.02) 2.62 (1.72–3.99) 2.33 (1.40–3.87)  Gestational hypertension 39 (5.60%) 48 (3.45%) 1.67 (1.11–2.50) 0.033 1.35 (0.90–2.01) 1.34 (0.89–2.00) 1.36 (0.91–2.04)  PE 46 (6.61%) 56 (4.02%) 1.67 (1.25–2.23) 0.005 1.41 (1.05–1.89) 1.43 (1.07–1.91) 1.23 (0.88–1.72) DM 153 (21.98%) 270 (19.4%) 1.17 (0.97–1.42) 0.212 1.15 (0.94–1.41) 1.14 (0.93–1.40) 1.13 (0.88–1.44)  Chronic DM a 23 (3.3%) 13 (0.93%) 3.54 (2.08–6.02) < 0.001 3.96 (2.46–6.38) 3.97 (2.47–6.39) 3.73 (2.18–6.37)  GDM 130 (18.68%) 257 (18.46%) 1.01 (0.82–1.25) 0.921 0.99 (0.79–1.23) 0.98 (0.79–1.22) 0.98 (0.77–1.24) Postpartum hemorrhage 63 (9.05%) 98 (7.04%) 1.31 (0.98–1.74) 0.144 1.26 (0.98–1.62) 1.27 (1.00–1.63) 1.25 (0.97–1.61) Placental abruption 16 (2.30%) 24 (1.72%) 1.33 (0.79–2.24) 0.384 1.63 (0.97–2.73) 1.65 (0.98–2.76) 1.69 (1.01–2.85) CIC 3.84 (2.16–6.83) < 0.001 4.86 (2.50–9.46) 4.92 (2.52–9.58) 4.05 (1.90–8.62)  Yes 20 (2.87%) 11 (0.79%)  No 655 (94.11%) 1339 (96.19%)  N/A 21 (3.02%) 42 (3.02%) PPROM 1.44 (0.65–3.18) 0.489 1.12 (0.55–2.29) 1.13 (0.56–2.31) 1.10 (0.59–2.06)  Yes 20 (2.87%) 29 (2.08%)  No 655 (94.11%) 1321 (94.9%)  N/A 21 (3.02%) 42 (3.02%) Neonatal outcomes b n   = 687 n   = 1360 Sex  Female 308 (44.83%) 600 (44.12%)  Male 379 (55.17%) 760 (55.88%) Low birthweight 53 (7.71%) 98 (7.21%) 1.09 (0.81–1.47) 0.639 1.02 (0.80–1.31) 1.03 (0.81–1.32) 0.92 (0.71–1.18) Macrosomia 46 (6.70%) 75 (5.51%) 1.23 (0.85–1.77) 0.384 1.27 (0.89–1.82) 1.22 (0.87–1.72) 1.14 (0.75–1.73) Small for gestational age 32 (4.66%) 84 (6.18%) 0.73 (0.47–1.14) 0.278 0.79 (0.47–1.32) 0.78 (0.46–1.32) 0.77 (0.75–1.73) Large for gestational age 134 (19.51%) 230 (16.91%) 1.18 (0.97–1.44) 0.212 1.22 (0.47–1.32) 1.20 (0.96–1.49) 1.17 (0.95–1.45) a Those diagnosed before and after conception were both included b Neonates with stillbirth or deliveries before 28 gestational weeks were excluded Data are presented as number (percentage) or OR (95% CI). OR (95% CI) was calculated with cluster-robust standard errors on hospital levels. P was adjusted for multiple comparisons using Benjamini-Hochberg method. Model 1 was adjusted for maternal education, smoking, nulligravida, nullipara, duration of infertility, family history of hypertension, family history of DM, male factor infertility, BMI < 18.5 kg/m 2 , BMI ≥ 28 kg/m 2 , hypothyroidism, intrauterine adhesions, genital tract malformations, endometriosis, and uterine adenomyosis. Model 2 was additionally adjusted for preconception hypertension and DM based on model 1, except for outcomes of hypertensive diseases and DM during pregnancy, respectively. Model 3 was additionally adjusted for insulin resistance based on model 2 Abbreviations : CI Confidence interval, CIC Cervical incompetence, DM Diabetes mellitus, GDM Gestational diabetes mellitus, HDP Hypertensive diseases of pregnancy, N/A Not applicable, OR Odds ratio, PE Pre-eclampsia, PTB Preterm birth, PPROM Preterm premature rupture of membranes, VPTB Very preterm birth Maternal and neonatal outcomes of women with singleton births a Those diagnosed before and after conception were both included b Neonates with stillbirth or deliveries before 28 gestational weeks were excluded Data are presented as number (percentage) or OR (95% CI). OR (95% CI) was calculated with cluster-robust standard errors on hospital levels. P was adjusted for multiple comparisons using Benjamini-Hochberg method. Model 1 was adjusted for maternal education, smoking, nulligravida, nullipara, duration of infertility, family history of hypertension, family history of DM, male factor infertility, BMI < 18.5 kg/m 2 , BMI ≥ 28 kg/m 2 , hypothyroidism, intrauterine adhesions, genital tract malformations, endometriosis, and uterine adenomyosis. Model 2 was additionally adjusted for preconception hypertension and DM based on model 1, except for outcomes of hypertensive diseases and DM during pregnancy, respectively. Model 3 was additionally adjusted for insulin resistance based on model 2 Abbreviations : CI Confidence interval, CIC Cervical incompetence, DM Diabetes mellitus, GDM Gestational diabetes mellitus, HDP Hypertensive diseases of pregnancy, N/A Not applicable, OR Odds ratio, PE Pre-eclampsia, PTB Preterm birth, PPROM Preterm premature rupture of membranes, VPTB Very preterm birth Table 3 Maternal and neonatal outcomes of women with twin births PCOS Control Crude model Adjusted OR (95% CI) OR (95% CI) P Model 1 Model 2 Model 3 Maternal outcomes n   = 201 n   = 402 PTB (< 37 weeks) 143 (71.14%) 251 (62.44%) 1.51 (1.03–2.22) 0.079 1.30 (0.93–1.80) 1.26 (0.90–1.76) 1.27 (0.86–1.87) VPTB (< 32 weeks) 23 (11.44%) 23 (5.72%) 2.08 (1.32–3.26) 0.008 1.97 (1.16–3.37) 2.00 (1.22–3.27) 1.57 (0.92–2.68) Caesarean section 186 (92.54%) 367 (91.29%) 1.21 (0.71–2.04) 0.603 1.23 (0.68–2.22) 1.14 (0.63–2.05) 1.16 (0.67–2.00) Stillbirth 0 (0) 4 (1.00%) N/A N/A 0.97 (0.68–1.39) 0.92 (0.67–1.25) 0.92 (0.67–1.25) HDP 36 (17.91%) 65 (16.17%) 1.14 (0.83–1.55) 0.541 1.01 (0.14–7.20) 0.88 (0.11–7.32) 0.69 (0.14–3.34)  Chronic hypertension a 4 (1.99%) 6 (1.49%) 1.33 (0.21–8.63) 0.808 1.03 (0.52–2.04) 1.06 (0.53–2.11) 1.24 (0.71–2.19)  Gestational hypertension 8 (3.98%) 18 (4.48%) 0.89 (0.49–1.59) 0.744 1.01 (0.62–1.63) 0.92 (0.61–1.40) 0.90 (0.60–1.34)  PE 26 (12.94%) 42 (10.45%) 1.28 (0.84–1.94) 0.384 2.23 (1.69–2.95) 2.27 (1.75–2.96) 2.03 (1.63–2.54) DM 66 (32.84%) 71 (17.66%) 2.41 (1.86–3.11) < 0.001 2.26 (0.93–5.48) 2.85 (1.21–6.72) 2.01 (0.85–4.76)  Chronic DM a 10 (4.98%) 7 (1.74%) 2.86 (1.25–6.51) 0.033 2.09 (1.65–2.65) 2.05 (1.61–2.61) 1.96 (1.56–2.46)  GDM 56 (27.86%) 64 (15.92%) 2.23 (1.70–2.92) < 0.001 0.53 (0.30–0.95) 0.56 (0.31–1.01) 0.52 (0.26–1.03) Postpartum hemorrhage 12 (5.97%) 47 (11.69%) 0.50 (0.31–0.80) 0.013 0.20 (0.06–0.62) 0.21 (0.07–0.60) 0.08 (0.00–5.48) Placental abruption 1 (0.50%) 10 (2.49%) 0.20 (0.08–0.53) 0.005 2.63 (0.49–13.90) 2.89 (0.60–13.9) 3.23 (0.71–14.6) CIC 3.32 (0.60–18.33) 0.278 1.30 (0.93–1.80) 1.26 (0.90–1.76) 1.27 (0.86–1.87)  Yes 12 (5.97%) 7 (1.74%)  No 180 (89.55%) 382 (95.02%)  N/A 9 (4.48%) 13 (3.23%) PPROM 1.23 (0.65–2.31) 0.606 1.33 (0.73–2.45) 1.31 (0.70–2.45) 1.41 (0.79–2.50)  Yes 26 (12.94%) 45 (11.19%)  No 166 (82.59%) 344 (85.57%) N/A 9 (4.48%) 13 (3.23%) Neonatal outcomes b n   = 392 n   = 789 Sex  Female 184 (46.94%) 364 (46.13%)  Male 208 (53.06%) 425 (53.87%) Low birthweight 220 (56.12%) 402 (50.95%) 1.18 (1.05–1.46) 0.033 1.15 (0.98–1.36) 1.14 (0.96–1.35) 1.11 (0.93–1.31) Macrosomia 0 (0) 2 (0.25%) N/A N/A 0.75 (0.58–0.98) 0.76 (0.58–1.00) 0.71 (0.55–0.92) Small for gestational age 46 (11.73%) 123 (15.59%) 0.73 (0.51–0.88) 0.013 0.67 (0.35–1.31) 0.70 (0.35–1.37) 0.67 (0.34–1.34) Large for gestational age 11 (2.81%) 30 (3.80%) 0.73 (0.40–1.15) 0.270 1.15 (0.98–1.36) 1.14 (0.96–1.35) 1.11 (0.93–1.31) a Those diagnosed before and after conception were both included b Neonates with stillbirth or deliveries before 28 gestational weeks were excluded Data are presented as number (percentage) or OR (95% CI). OR (95% CI) was calculated with cluster-robust standard errors on hospital levels. P was adjusted for multiple comparisons using Benjamini-Hochberg method. Model 1 was adjusted for maternal education, smoking, nulligravida, nullipara, duration of infertility, family history of hypertension, family history of DM, male factor infertility, BMI < 18.5 kg/m 2 , BMI ≥ 28 kg/m 2 , hypothyroidism, intrauterine adhesions, genital tract malformations, endometriosis, and uterine adenomyosis. Model 2 was additionally adjusted for preconception hypertension and DM based on model 1, except for outcomes of hypertensive diseases and DM during pregnancy, respectively. Model 3 was additionally adjusted for insulin resistance based on model 2  Abbreviations : CI Confidence interval, CIC Cervical incompetence, DM Diabetes mellitus, GDM Gestational diabetes mellitus, HDP Hypertensive diseases of pregnancy, N/A Not applicable, OR Odds ratio, PE Pre-eclampsia, PTB Preterm birth, PPROM Preterm premature rupture of membranes, VPTB Very preterm birth Maternal and neonatal outcomes of women with twin births a Those diagnosed before and after conception were both included b Neonates with stillbirth or deliveries before 28 gestational weeks were excluded Data are presented as number (percentage) or OR (95% CI). OR (95% CI) was calculated with cluster-robust standard errors on hospital levels. P was adjusted for multiple comparisons using Benjamini-Hochberg method. Model 1 was adjusted for maternal education, smoking, nulligravida, nullipara, duration of infertility, family history of hypertension, family history of DM, male factor infertility, BMI < 18.5 kg/m 2 , BMI ≥ 28 kg/m 2 , hypothyroidism, intrauterine adhesions, genital tract malformations, endometriosis, and uterine adenomyosis. Model 2 was additionally adjusted for preconception hypertension and DM based on model 1, except for outcomes of hypertensive diseases and DM during pregnancy, respectively. Model 3 was additionally adjusted for insulin resistance based on model 2  Abbreviations : CI Confidence interval, CIC Cervical incompetence, DM Diabetes mellitus, GDM Gestational diabetes mellitus, HDP Hypertensive diseases of pregnancy, N/A Not applicable, OR Odds ratio, PE Pre-eclampsia, PTB Preterm birth, PPROM Preterm premature rupture of membranes, VPTB Very preterm birth The pattern of the main findings was reflected in subgroup analyses stratified by maternal transfer age, nulligravida, and fresh/frozen ET, with no significant between-subgroup interactions (Table S4a and S4b), alongside four sensitivity analyses (Table S5a-d) on the impact of missing data imputation, preconception chronic diseases, other infertility etiologies (also see Table S6), or women with twin pregnancies but ultimately singleton births. These results subsequently reinforced the robustness of the main findings. Figure 3 summarized the detailed results of mediation analyses in Table S7a and S7b. Overall, the mediation analyses supported the significant total effect of PCOS on adverse pregnancy outcomes ( P  < 0.05). Preconception BMI mediated roughly one third of the total effects on HDP and PE in women with singleton births, and around 10% of the effects on GDM and DM during pregnancy (chronic DM and GDM) in those with twin births. Additionally, IR accounted for 17.5% and 35.1% of the increased risks of HDP and PE for singleton births, respectively. Minimal but significant contribution of preconception DM was observed on the pathway from PCOS to PE in singleton births (-2.2%, P  = 0.020, Table S7a), as well as that of preconception hypertension on the pathway from PCOS to GDM (0.9%, P  < 0.001) and to VPTB (0.7%, P  < 0.001) in twin births (Table S7b). Fig. 3 Mediation analysis of the relationship between PCOS and adverse pregnancy outcomes. Mediation effects of preconception BMI and IR were depicted in women with ( A-D ) singleton births and ( E , F ) twin births. Abbreviations: DM, diabetes mellitus; HDP, hypertensive diseases of pregnancy; IR, insulin resistance; PE, pre-eclampsia; GDM, gestational diabetes mellitus Mediation analysis of the relationship between PCOS and adverse pregnancy outcomes. Mediation effects of preconception BMI and IR were depicted in women with ( A-D ) singleton births and ( E , F ) twin births. Abbreviations: DM, diabetes mellitus; HDP, hypertensive diseases of pregnancy; IR, insulin resistance; PE, pre-eclampsia; GDM, gestational diabetes mellitus A correlation matrix of the PCOS population was displayed in Fig. 4 to preliminarily explore the association between preconception prognostic factors and adverse outcomes aforementioned. It predominantly showed weak correlations as represented by light-coloured cells. Metabolic disorders in deep-coloured cells indicated obvious correlations to pregnancy complications. In addition to traditional clinical indicators like BMI and IR, this matrix exhibited potential interactions of composite metabolic indicators (TyG, TG/HDL, and TyG-BMI) with increased obstetric risks (especially HDP in PCOS-singleton group). Among women in the PCOS-singleton group, CIC was associated probably with a history of cervical surgery, and possibly with Hb, fasting plasma glucose, and frozen ET. Fig. 4 Bubble heatmap showing potential correlations between preconception parameters and adverse pregnancy outcomes of PCOS population. Bubble size represented -log 10 transformed value of P . Color intensity displayed the Spearman correlation coefficients, with red for a positive correlation and blue for a negative one. Abbreviations: ALT, alanine aminotransferase; ART, assisted reproductive technology; AST, aspartate transaminase; CIC, cervical incompetence; DM, diabetes mellitus; FSH, follicle stimulating hormone; GDM, gestational diabetes mellitus; HDL, high-density lipoprotein; HDP, hypertensive diseases of pregnancy; IR, insulin resistance; LH, luteinizing hormone; PE, pre-eclampsia; TG, triglycerides; TyG, triglyceride-glucose index; TyG-BMI, triglyceride-glucose body mass index; VPTB, very preterm birth Bubble heatmap showing potential correlations between preconception parameters and adverse pregnancy outcomes of PCOS population. Bubble size represented -log 10 transformed value of P . Color intensity displayed the Spearman correlation coefficients, with red for a positive correlation and blue for a negative one. Abbreviations: ALT, alanine aminotransferase; ART, assisted reproductive technology; AST, aspartate transaminase; CIC, cervical incompetence; DM, diabetes mellitus; FSH, follicle stimulating hormone; GDM, gestational diabetes mellitus; HDL, high-density lipoprotein; HDP, hypertensive diseases of pregnancy; IR, insulin resistance; LH, luteinizing hormone; PE, pre-eclampsia; TG, triglycerides; TyG, triglyceride-glucose index; TyG-BMI, triglyceride-glucose body mass index; VPTB, very preterm birth

This retrospective cohort study included women with PCOS or tubal factors who underwent IVF/ICSI at Women’s Hospital, Zhejiang University School of Medicine from 1 June 2017 to 31 May 2022 (Fig.  1 ). Detailed follow-ups were conducted before, during, and after IVF/ICSI. Obstetric outcomes after 20 weeks of gestation were collected from 95 hospitals by 30 June 2023. To minimize confounding interference, 1:2 PSM analysis was performed in PCOS group and controls, among women with singleton and twin births, respectively. Obstetric outcomes were compared between the groups. Mediation analysis and an exploratory correlation analysis were further performed. Fig. 1 Flowchart of the participants through the study Flowchart of the participants through the study We enrolled women who underwent IVF/ICSI at our hospital and maintained pregnancy beyond 20 weeks from Zhejiang, China. PCOS was diagnosed according to Rotterdam criteria (2003), which requires at least two of three features: polycystic ovaries on ultrasound, hyperandrogenism, or anovulatory irregular cycles, alongside exclusion of related disorders. All included PCOS women met at least two of the three features, with hyperandrogenism clinically or biochemically confirmed. Non-PCOS participants with tubal infertility were arranged as controls. Exclusion criteria included chromosomal or gene abnormalities of one or both spouses, maternal age older than 40 years at enrollment, pre-existing severe medications (heart, renal or liver dysfunction, malignant tumors, and epilepsy), and refusal to consent. Participants with preconception diabetes mellitus (DM), insulin resistance (IR), and hypertension were not excluded, considering the fact that PCOS is a syndrome with highly heterogeneous endocrine-metabolic manifestations in the real world. For multiple ART cycles meeting the criteria, only the first cycle was considered. After enrollment, detailed follow-ups were arranged throughout ART cycles, pregnancy, and peripartum. The specific procedure is described as follows. Before ART, participants received a standard pre-pregnancy exam at reproductive health clinics. Demographic and clinical data of both the participants and their husbands were recorded. Several emerging composite indicators correlating to glucose and lipid metabolism were calculated, namely triglyceride-glucose (TyG) index = ln[triglycerides (TG) (mg/dL)×fasting plasma glucose (mg/dL)/2] [ 16 ], TG/high-density lipoprotein (HDL), and triglyceride-glucose body mass index (TyG-BMI) = TyG×BMI (kg/m 2 ) [ 17 ]. After up-front assessments and treatments, ovulation was inducted, and oocytes were retrieved under transvaginal ultrasound guidance. Embryos were fertilized via IVF, ICSI, or half ICSI, and graded according to Istanbul consensus before fresh or frozen ET. Biochemical and clinical pregnancies were confirmed through human chorionic gonadotropin levels at 2 weeks post-transfer and B-ultrasound at 3 months post-transfer. Fertility counseling, weight management, and psychological support were also provided during the visits. Detailed maternal and neonatal outcomes were obtained by specialists from 95 hospitals in Zhejiang province, via extracting medical records or telephoning participants. Pregnancy complications included PTB, HDP (including chronic hypertension, gestational hypertension, and PE), diabetes mellitus (DM; including chronic DM and GDM), CIC, PPROM, and so on. Neonatal outcomes referred to neonatal sex and birthweight-related outcomes. Across the hospitals, all diagnosis was made on the basis of national and international guidelines. As for chronic DM, women diagnosed before or after conception were both included, given their similar pregnancy management. Grouping was focused on births rather than pregnancies, since most abortions in this study occurred at early gestation. After excluding population with missing data, PSM was employed to balance the confounding effect within singleton and twin births, respectively. This robust statistical method matched samples with one or more controls based on propensity scores. The analysis was performed using R software (version 4.2.2) relying on the MatchIt package and MSTATA software. The PCOS population were matched to the controls considering delivery districts in a 1:2 ratio using greedy nearest neighbor matching without replacement. The caliper width was set at 0.2 of the standard deviation of the logit-transformed propensity score. Standardized mean differences (SMD) before and after PSM were calculated to evaluate the effectiveness of matching, where SMD < 0.1 were considered relatively balanced. Matching variables were selected based on prior knowledge and the literature, namely, maternal age ≥ 35 years at embryo transfer (ET), preconception antiphospholipid syndrome (APS) or systemic lupus erythematosus (SLE), as well as fresh versus frozen ET owing to its significant impact on pregnancy outcomes and multifactorial infertility causes well beyond PCOS [ 18 ]. Preconception PCOS features like obesity, IR, DM, and hypertension [ 2 ] were excluded from matching to avoid blocking the mediated effects. Other statistical analyses were performed using Stata version 19.0 (Stata Corp.), SPSS version 29.0 (IBM Corp.), or R software (version 4.2.2), with p  < 0.05 (two-tailed) as the significance threshold. No data points were missing in primary outcomes and ART parameters. The missing rate for baseline demographic and clinical characteristics was lower than 10%, and was treated with multiple imputations by chained equations with five imputations. In crude analysis after PSM, non-Gaussian continuous variables and categorical variables were compared using Wilcoxon rank sum test and conditional logistic regression models, respectively. Unconditional regression adjustments post PSM were applied as a doubly robust approach to further confirm the results [ 19 – 21 ]. In other analyses, non-Gaussian continuous variables and ordinal categorical variables were compared using the Mann–Whitney U test, while nominal categorical variables using the Pearson’s chi-square test. Benjamini-Hochberg method was applied to account for multiple comparisons in the main analysis, and results were considered statistically significant when the false discovery rate was below 0.05. All regression estimates were adjusted for cluster-robust standard errors (CRSEs) to address the potential intra-hospital variations in clinical practice and patient referral biases. Subgroup analyses were performed stratified by maternal transfer age, nulligravida, and fresh/frozen ET. Four sensitivity analyses were conducted to minimize the potential confounding, namely, using different data imputations, excluding participants with preconception chronic diseases, excluding those with other infertility etiologies, or only including those with singleton pregnancies and ultimately singleton births. Moreover, mediation analyses were applied to decompose the total effect of PCOS on high-rate adverse pregnancy outcomes through preconception mediators, with a nonparametric bootstrap method of 1000 simulations. To preliminary explore risk factors of adverse outcomes of PCOS women, correlation analysis was conducted involving parameters of clinical characteristics, laboratory tests, and ART cycles, using Spearman method with p  < 0.10 as the significance threshold.

Existing literature presents a continuous debate regarding the impact of PCOS on pregnancy complications. In this longitudinal cohort study, we posited that the number of foetuses may serve as an essential covariate. We evaluated maternal and neonatal outcomes in women with PCOS who underwent IVF/ICSI and maintained pregnancy until 20 weeks of gestation, comparing to those of the controls with tubal infertility. Our findings indicated that PCOS was associated with HDP, PE, and CIC in singleton births, while with GDM and VPTB in twin births; in either singleton or twin births, there was a higher prevalence of chronic DM in the PCOS population compared to their counterparts, but no significant between-group differences in PPROM. This study accumulated evidence for the differential impact of PCOS on pregnancy outcomes between singleton and twin pregnancy, highlighting the necessity of individual preconception counseling, reproductive strategies, and obstetric management in clinical practice. The principal strength of this study lied in the relatively comprehensive and detailed follow-ups and data extraction. A broad number of clinical and laboratory parameters from the participating couples were extracted at multiple follow-up intervals. Various common pregnancy and peripartum outcomes were gathered from up to 95 hospitals. Additionally, PSM technique was utilized to mitigate selection biases of several established covariates of PE or GDM. Unlike previous research, covariates characteristic of PCOS were deliberately excluded from the matching variables to prevent overcorrection. Considerable number of subsequent analyses conducted as well to validate the primary findings. Lastly, the correlation analysis indicated some preconception prognostic factors for adverse pregnancy outcomes among PCOS population, particularly emerging composite factors linked to cardiovascular and metabolic diseases, which may serve as a tentative reference for future research. Several limitations should be considered to interpret this study. First, individuals were included with metabolic conditions like preconception obesity, DM, and hypertension, to improve the approximation to clinical practice that PCOS is a complex syndrome frequently accompanied by metabolic disorders; meanwhile, these conditions also posed potential confounders. Although considerable amounts of adjustments and sensitivity analyses were employed, the consistency of our findings across these analyses mitigated, but did not entirely eliminate residual confounding. Second, due to limited clinical adoption during the research period, we were unable to incorporate PCOS-related biomarkers, including anti-Müllerian hormone and sex hormone-binding globulin, into the analyses. While these factors might be associated with ART success and miscarriage [ 22 ], their relationship with obstetric risks were unclear. Recent research reported different subtypes based on these biomarkers and metabolic profiles [ 1 ]. Further research is warranted to reveal the differential pregnancy outcomes of stratified PCOS population. Third, no comparator was completely neutral in ART population. We chose tubal factor infertility as one of the most established and frequently utilized control [ 23 – 26 ] to effectively exclude severe confounders from other infertility causes like autoimmune diseases and genital tract malformations [ 27 , 28 ]. However, we acknowledged that the underlying inflammatory or infectious differences between tubal and ovarian pathologies should not be neglected. Fourth, data were retrospectively collected from a single-center cohort in China, necessitating further validation through more powerful designs in more diverse populations. This study revealed that PCOS women could face different types of higher obstetric risks than controls based on different feotal numbers, which aligned with existing evidence [ 3 , 4 , 10 ]. It could be reasoned that multiple pregnancy per se suffered from various pathophysiological mechanisms such as increased intrauterine pressure, excessive stretching of uterine muscle fibers, enlarged placental attachment, inherent placental insufficiency, increased production of angiogenic substances, and early termination of pregnancy [ 29 – 31 ], incurring profound risk for PTB, hypertension, GDM, postpartum haemorrhage, placental abruption, CIC, and small for gestational age (Table S3) [ 15 , 32 , 33 ]. As a result, the contribution of PCOS would be substantially overshadowed or amplified. Moreover, women with pre-existing comorbidities are recommended to receive single ETs or may face foetal reduction during early pregnancy, to avoid the severe risks linked to twin gestations [ 34 ]. Given that PCOS is frequently accompanied by endocrine and metabolic abnormalities, the disparity in comorbidity incidence between PCOS women and controls would be markedly attenuated in the twin-birth cohort compared to the singleton-birth cohort (e.g., BMI ≥ 28 kg/m 2 and hypothyroidism in Table S6). Consequently, outcomes of the twin-birth cohort should be skewed in favor of the PCOS group relative to the control group. This further strengthened the robustness of the increased risks of DM and VPTB observed in PCOS-twin population. Among women with singleton births, those with PCOS demonstrated an apparently higher percentage of hypertension (including PE) than those controls did. This finding was in good agreement with previous reports [ 5 , 7 , 10 – 12 , 35 , 36 ], which could attribute to PCOS-related features like obesity, IR, metabolic issues, and hyperandrogenism [ 7 , 8 , 35 , 37 ]. Our mediation analyses further identified the critical role of preconception BMI and IR on hypertensive disorders during pregnancy, reportedly relating to pathophysiology mechanisms like vascular endothelial dysfunction, systemic inflammation, oxidative stress, sympathetic nervous system activation [ 38 , 39 ]. In addition, our exploratory analysis indicated several potential correlators of HDP and PE, such as metabolic indicators (discussed in the later paragraph) and platelet counts. Interestingly, women with PCOS did present elevated platelet counts [ 40 ] before and after conception. However, emerging evidence primarily related high platelet counts to hypertension in non-pregnant populations [ 41 , 42 ], or during pregnancy [ 43 ]. Our finding offers initial observation for future research to investigate the correlation between preconception platelet count on hypertension post conception. Consistent with prior observations [ 5 , 6 , 13 ], we observed a 3-fold OR increase in CIC associated with PCOS women with singleton births. Current evidence often attributes CIC in PCOS to hyperandrogenemia [ 7 , 44 , 45 ], which may induce premature cervical ripening by accelerating collagen fiber dissolvement via IL-8 recruitment, promoting mechanical stretching via estradiol synthesis and inflammatory mediators, and inducing hyaluronic acid production, as well as impair cervical perfusion via promoting vasoconstriction [ 46 – 49 ]. Notably, our subgroup and correlation analyses suggested that frozen ET may also be related to CIC in the PCOS-singleton population. We reasoned that those receiving frozen ET were commonly complicated by endometrial or cervical abnormalities, and consequently, a history of multiple trans-cervical procedures, collectively elevating their risk of developing CIC [ 50 ]. Furthermore, most frozen ETs were carried out in artificial cycles, where estrogen fluctuations may influence cervical remodeling [ 51 , 52 ]. A better understanding of the pathogenesis underlying PCOS-related CIC and its prophylactic interventions is necessary. For both PCOS women with singleton and twin pregnancies, a modest but significant correlation was exhibited between preconception metabolic factors and obstetric risks (Fig.  4 ). As an important phenotype of PCOS, metabolic disturbances could provoke oxidative stress, inflammatory responses, and endothelial dysfunction, and subsequently interfere with essential placental development mechanisms like trophoblast invasion and spiral artery remodeling [ 53 ]. Interestingly, we found a potentially stronger correlation of obstetric risks with these emerging composite parameters compared to traditional single ones (e.g. BMI and IR). Growing evidence has emerged supporting the predictive efficacy of these composite indices during early pregnancy for subsequent outcomes such as GDM, gestational hypertension, PE and low birthweight [ 53 – 57 ]. Our findings may serve as a tentative reference for future investigations into the relationships between pregnancy outcomes and preconception metabolic profiles. With the evolving objective of ART, favourable obstetric outcomes have become as essential as a successful pregnancy. We promote elective single ET as the predominant approach to balance the objective of preventing multiple gestation and achieving a successful pregnancy [ 34 , 58 ]. Furthermore, we emphasize the importance of sufficient discussions and thorough information dissemination concerning potential obstetric outcomes and preconception risk factors before ART initiation. PCOS women should be aware of the risks of hypertension and CIC during pregnancy, alongside their potential risk factors like preconception metabolic disturbances and a history of cervical surgery. In certain situations [ 34 ] where PCOS women receive multiple ET after comprehensive evaluations, in addition to general adverse outcomes of twin pregnancies, elevated risks of VPTB and GDM should be particularly highlighted. Detailed preconception assessments, thorough information dissemination, tailored pregnancy management, and adequate psychological support are crucial for PCOS women in achieving a successful pregnancy and favourable obstetric outcomes [ 59 ].

In this study, women with PCOS who had singleton births demonstrated a higher incidence of hypertension and CIC compared to those with tubal factors. Among women with twin births, increased risks of GDM and VPTB were noted in addition to the inherent risks of twin gestation. Our findings accumulated evidence for the distinct pregnancy outcomes in singleton and twin births of PCOS, and underscored the necessity of individual reproductive and obstetric management accordingly.

Polycystic ovary syndrome (PCOS) is a prevalent metabolic disorder affecting 11% to 13% of women worldwide [ 1 ]. It is marked by irregular menstruation, polycystic ovaries, hyperandrogenism, and abnormal metabolism [ 2 ]. As a leading cause of infertility [ 3 ], PCOS necessitates the use of assisted reproductive technologies (ARTs) in an increasing number of affected women to achieve pregnancy. Pregnant women with PCOS also face increased risks of complications like gestational diabetes mellitus (GDM), hypertensive diseases of pregnancy (HDP), and cervical incompetence (CIC) [ 4 – 6 ]. Consequently, PCOS presents significant health and economic challenges for individuals and society. In addition to these acknowledged obstetric risks mentioned above, PCOS might be associated with other severe pregnancy outcomes like pre-eclampsia (PE) [ 7 – 10 ] and preterm birth (PTB) [ 3 , 8 , 9 , 11 , 12 ] in the majority of existing literature. However, the other studies observed dissimilar results [ 3 , 5 , 12 ]. Interestingly, we found inconsistency of singleton and twin gestations included in the prior studies as a potential covariate. For instance, a Swedish cohort focusing on twin births found no significant difference in PE occurrence between PCOS and non-PCOS population [ 3 ]. Conversely, another nationwide Swedish cohort reported an increased risk for PE in PCOS women with singleton births than their counterparts [ 10 ]. Similar discrepancy existed in other outcomes like PTB, small for gestational age, and preterm premature rupture of membranes (PPROM) [ 4 , 5 , 13 ]. These variations may be attributed to fundamentally different maternal physiology and perinatal complications based on the number of foetuses [ 14 , 15 ], which could either obscure or amplify the effects of PCOS [ 4 ]. Nonetheless, there is limited and debated evidence examining the effects of PCOS on singleton versus twin gestations, which requires further assessments in current clinical practice. To investigate the pregnancy outcomes between these two populations, we carried out a retrospective cohort study spanning 2017–2022. We included women who underwent in vitro fertilization (IVF) or intracytoplasmic sperm injection (ICSI) and maintained pregnancy beyond 20 weeks of gestation, comparing those with PCOS to their counterparts with tubal infertility. Propensity score matching (PSM) was used to minimize selection biases within singleton and twin delivery groups, respectively. Pregnancy and neonatal outcomes were collected from up to 95 hospitals and analysed using several models and adjustments. Mediation analysis was conducted to decompose the total effect of PCOS on several pregnancy risks through preconception mediators. An exploratory correlation analysis was performed to indicate preconception factors potentially associated with several pregnancy outcomes in the PCOS population.

Supplementary Material 1 Supplementary Material 1 Supplementary Material 2 Supplementary Material 2