Impact of race and ethnicity on in vitro fertilization outcomes in infertile women with polycystic ovary syndrome in the United States.

Of the 256,018 patient records obtained, 128,703 women met the inclusion criteria. Of the 128,703 patients, 21,866 (16.9%) met the definition of PCOS. The other 106,837 patients (83%) were, therefore, controls. In the PCOS group, 15,643 women (71.5%) were white, 3,504 (16%) were Asian, 1,601 (7.3%) were Hispanic, and 1,118 (5.1%) were African American. In the control group, 72,297 (67.7%) were white, 18,803 (17.6%) were Asian, 8,403 (7.9%) were Hispanic, and 7,334 (6.9%) were African American ( P < .001, testing equality of all groups) ( Supplemental Fig. 1 , available online). Unadjusted comparison of patients’ demographics in women with infertility with and without PCOS undergoing IVF is shown in Supplemental Table 1 . Patients in the PCOS group were significantly younger than those in the non-PCOS group ( P < .001). Patients with PCOS had a significantly higher BMI, significantly higher AMH levels, moderate and severe ovarian hyperstimulation, and a significantly higher number of oocytes retrieved than those in the non-PCOS group ( P < .001). Non-PCOS patients were significantly more likely to have had prior pregnancies, be smokers, and have had used higher FSH gonadotropin doses during their treatment cycle ( P < .001). Non-PCOS women were significantly more likely to present with male factor infertility, endometriosis, diminished ovarian reserve, tubal and uterine infertility factors, and unexplained infertility than those in the PCOS group ( P < .001). Unadjusted comparison of pregnancy and neonatal outcomes according to PCOS and non-PCOS statuses is shown in Supplemental Table 2 . The non-PCOS group had significantly more embryos transferred to the uterus than the PCOS group (0.94 vs. 0.87, respectively; P < .001). However, the PCOS group had a significantly higher implantation rate (44.07 vs. 38.34, P < .001), pregnancy rate (52.5% vs. 47.6%, P < .001), and LBR (47.3% vs. 42.6%, P < .001) than the non-PCOS group. In addition, the multiple pregnancy rate was significantly higher in the PCOS group than in the non-PCOS group (13.7% vs. 11.5%, respectively; P < .001). The overall pregnancy loss rate was not significantly different between the PCOS and non-PCOS groups (9.3% vs. 9.6%, respectively; P = .390). The PCOS group also demonstrated a significantly lower neonatal birth weight (2,832 vs. 2,890.56 g, P < .001) and a significantly higher neonatal death rate than the non-PCOS group (1.4% vs. 0.9%, P < .001). Patients’ demographics by race and ethnicity in both the PCOS and non-PCOS groups are shown in Supplemental Tables 3 and 4 . African American women in both the PCOS and non-PCOS groups had a significantly higher BMI and percentage of prior pregnancies and tubal factor infertility than the rest of the women ( P < .001, testing equality in both groups). White women in the non-PCOS group were significantly more likely to be smokers than the rest of the women in the non-PCOS group ( P < .001). Gonadotropin-releasing hormone antagonist suppression was significantly more likely used in Asian women than the rest of the women in both the PCOS and non-PCOS groups ( P < .001). The AMH levels in the PCOS group were significantly higher in African American women ( P < .001), and in the non-PCOS group, these were significantly higher in both white and African American women than in the other women ( P < .001). The number of oocytes retrieved was significantly higher in African American women in both the PCOS and non-PCOS groups than in the other women ( P < .001). Unadjusted results for pregnancy and neonatal outcomes by race and ethnicity in both the PCOS and non-PCOS groups are shown in Tables 1 and 2 . For the primary outcome, the LBR in the PCOS group was highest in white women (49.5%), compared with those in Hispanic (42.7%), Asian (41.6%), and African American (36%) women ( P < .001). The LBR in the non-PCOS group was highest in white women (45.1%), compared with Hispanic (40.5%), Asian (35.4%), and African American (34.3%) women ( P < .001). For the secondary outcomes, the implantation and pregnancy rates were significantly lower in African American and Asian women in both the PCOS and non-PCOS groups ( P < .001). The pregnancy loss rate was significantly higher in African American women in both the PCOS (19.5%) and non-PCOS (13.4%) groups ( P < .001). Furthermore, the pregnancy loss rate significantly increased in Hispanic women in the PCOS group (15.3%; P < .001). By way of comparison, the pregnancy loss rates in white women in the PCOS and non-PCOS groups were 8.0% and 9.3%, respectively. The C-section rate significantly increased, and consequently, the vaginal delivery rate significantly decreased in African American women in the non-PCOS group ( P < .001). African American women had neonates with significantly lower birth weights in both the PCOS and non-PCOS groups ( P < .001). Neonatal deaths differed in both the PCOS and non-PCOS groups, with Hispanic women having significantly higher neonatal deaths in the PCOS group (3.9%; P < .001) but African American women having significantly higher neonatal deaths in the non-PCOS group (2.3%; P < .001). Table 1 Pregnancy and neonatal outcomes by race and ethnicity in women with infertility with polycystic ovary syndrome. Patient variables White (n = 15,643) Asian (n = 3,504) Hispanic (n = 1,601) African American (n = 1,118) P value Embryos to the uterus (N, mean/SD) 0.90 (0.91) 0.68 (0.86) 0.93 (0.99) 0.88 (0.96) <.001 Implantation rate (mean/SD) 45.66 (46.65) 39.11 (46.41) 41.22 (44.52) 37.51 (45.07) <.001 Pregnancy rate (N, %) 4,692 (54.2%) 706 (46.1%) 436 (52.0%) 265 (46.1%) <.001 Live birth rate (N, %) 4,284 (49.5%) 638 (41.6%) 358 (42.7%) 207 (36.0%) <.001 Pregnancy loss rate (N, %) 386 (8.2%) 62 (8.8%) 67 (15.3%) 52 (19.5%) <.001 Multiple pregnancy rate (N, %) 1,228 (14.2%) 144 (9.4%) 140 (16.7%) 82 (14.3%) <.001 C-section rate (N, %) 511 (44.2%) 71 (40.1%) 46 (49.5%) 38 (53.5%) .196 Vaginal delivery rate (N, %) 645 (55.8%) 106 (59.9%) 47 (50.5%) 33 (46.5%) .196 Patient variables White (n = 5,336) Asian (n = 762) Hispanic (n = 471) African American (n = 278) P value Weight (g, mean/SD) 2,873.48 2,806.65 2,637.76 2,434.53 <.001 Neonatal deaths (N, %) 58 (1.1%) 11 (1.5%) 18 (3.9%) 10 (3.6%) <.001 Congenital defects (N, %) 34 (0.6%) 1 (0.1%) 3 (0.6%) 0 (0.0%) .192 Note: Statistically significant P values are shown in bold style. C-section = cesarean section; SD = standard deviation. Table 2 Pregnancy and neonatal outcomes by race and ethnicity in women with infertility without polycystic ovary syndrome. Patient variables White (n = 15,643) Asian (n = 3,504) Hispanic (n = 1,601) African American (n = 1,118) P value Embryos to the uterus (N, mean/SD) 0.97 (0.95) 0.73 (0.94) 1.10 (1.02) 1.05 (0.97) <.001 Implantation rate (mean/SD) 40.75 (45.24) 31.76 (43.60) 34.41 (42.57) 32.40 (42.73) <.001 Pregnancy rate (N, %) 21,036 (50.2%) 3,219 (39.2%) 2,298 (45.3%) 1,833 (40.9%) <.001 Live birth rate (N, %) 18,919 (45.1%) 2,910 (35.4%) 2,054 (40.5%) 1,542 (34.4%) <.001 Pregnancy loss rate (N, %) 1,953 (9.3%) 324 (10.1%) 217 (9.4%) 245 (13.4%) <.001 Multiple pregnancy rate (N, %) 5,173 (12.3%) 640 (7.8%) 619 (12.2%) 454 (10.1%) <.001 C-section rate (N, %) 2,401 (43.4%) 383 (43.2%) 314 (49.8%) 254 (54.2%) <.001 Vaginal delivery rate 3,127 (56.6%) 503 (56.8%) 316 (50.2%) 215 (45.8%) <.001 Patient variables White (n = 5,336) Asian (n = 762) Hispanic (n = 471) African American (n = 278) P value Weight (g, mean/SD) 2,921.51 2,857.33 2,814.77 2,669.81 <.001 Neonatal deaths (N, %) 179 (0.8%) 20 (0.6%) 38 (1.5%) 43 (2.3%) <.001 Congenital defects (N, %) 122 (0.5%) 17 (0.5%) 15 (0.6%) 14 (0.7%) .646 Note: Statistically significant P values are shown in bold style. C-section = cesarean section; SD = standard deviation. Pregnancy and neonatal outcomes by race and ethnicity in women with infertility with polycystic ovary syndrome. Note: Statistically significant P values are shown in bold style. C-section = cesarean section; SD = standard deviation. Pregnancy and neonatal outcomes by race and ethnicity in women with infertility without polycystic ovary syndrome. Note: Statistically significant P values are shown in bold style. C-section = cesarean section; SD = standard deviation. Adjusted results using logistic and linear regression are shown in Figures 1 and 2 , Supplemental Figures 2–6 , and Supplemental Table 5 . Adjusted odds ratios (AORs) were calculated for each interaction between each race and ethnic group and our referent group (i.e., non-PCOS white women). For the primary outcome, African American and Asian women in the PCOS group were significantly less likely to have a live birth (AORs, 0.72 and 0.81, respectively; P < .001) ( Fig. 1 ). For the secondary outcomes, Asian women in the PCOS group were significantly less likely to have an implanted embryo (AORs, 0.90; P = .003) ( Supplemental Fig. 2 ). Similarly, the pregnancy rate was significantly lower in Asian women in the PCOS group (AOR, 0.78; P < .001) ( Supplemental Fig. 3 ). The pregnancy loss rate was significantly higher in African American and Hispanic women with PCOS (AORs, 2.47 and 1.80, respectively; P < .001) ( Fig. 2 ). White women in the PCOS group were significantly more likely to have a multiple pregnancy (AOR, 1.10; P = .008), whereas Asian women in the PCOS group were significantly less likely to have a multiple pregnancy (AOR, 0.75; P = .007) ( Supplemental Fig. 4 ). Neonatal weight was significantly lower in all women with PCOS (white, Asian, Hispanic, and African American) than in white non-PCOS women ( P < .001) ( Supplemental Fig. 5 ). Among those women who had a live birth, the likelihood of a neonatal death was significantly higher in African American, Asian, and Hispanic women (AORs, 4.42, 2.35, and 5.21, respectively; P < .01) ( Supplemental Fig. 6 ). In addition, we derived odds ratios for contrasts between women with PCOS of each racial and ethnic group and white non-PCOS women, adjusted for age, BMI, smoking status, and number of transferred embryos ( Supplemental Table 5 ). We found that white women with PCOS actually had a significantly higher likelihood of a live birth than white women without PCOS (AOR, 1.114; P < .001), whereas Asian and African American women had a significantly lower chance of a live birth (AORs, 0.819 and 0.729, respectively; P < .001). We also derived odds ratios for contrasts between women with PCOS of each racial and ethnic group and white women with PCOS, adjusted for age, BMI, smoking status, and number of transferred embryos ( Supplemental Table 6 ) and confirmed the significant impact of race (Asian, Hispanic, and African American) in reducing the LBRs ( P < .05). Figure 1 Live birth rate (LBR). Calculated using the number of women who delivered a live born divided by the number of women who had an embryo transfer. Each measurement included the 95% confidence interval. Comparisons within the same race and ethnicity showed that white and Asian women with polycystic ovary syndrome (PCOS) had significantly higher LBRs than non-PCOS women ( P < .05). A logistic regression model was also used to calculate the LBR outcomes on the basis of PCOS status and their interaction with different races and ethnicities, adjusting for patient’s age, body mass index, smoking status, and number of embryos transferred. White non-PCOS women were used as the referent group (details shown in the Results section). Figure 2 Pregnancy loss rate. Calculated using the number of miscarriages divided by the number of clinical pregnancies. Each measurement included the 95% confidence interval. Comparisons within the same race and ethnicity showed that Hispanic and African American women with polycystic ovary syndrome (PCOS) had significantly higher pregnancy loss rates than non-PCOS women ( P < .05). A logistic regression model was also used to calculate the loss rate outcomes on the basis of PCOS status and their interaction with different races and ethnicities, adjusting for patient’s age, body mass index, smoking status, and number of embryos transferred. White non-PCOS women were used as the referent group (details shown in the Results section). Live birth rate (LBR). Calculated using the number of women who delivered a live born divided by the number of women who had an embryo transfer. Each measurement included the 95% confidence interval. Comparisons within the same race and ethnicity showed that white and Asian women with polycystic ovary syndrome (PCOS) had significantly higher LBRs than non-PCOS women ( P < .05). A logistic regression model was also used to calculate the LBR outcomes on the basis of PCOS status and their interaction with different races and ethnicities, adjusting for patient’s age, body mass index, smoking status, and number of embryos transferred. White non-PCOS women were used as the referent group (details shown in the Results section). Pregnancy loss rate. Calculated using the number of miscarriages divided by the number of clinical pregnancies. Each measurement included the 95% confidence interval. Comparisons within the same race and ethnicity showed that Hispanic and African American women with polycystic ovary syndrome (PCOS) had significantly higher pregnancy loss rates than non-PCOS women ( P < .05). A logistic regression model was also used to calculate the loss rate outcomes on the basis of PCOS status and their interaction with different races and ethnicities, adjusting for patient’s age, body mass index, smoking status, and number of embryos transferred. White non-PCOS women were used as the referent group (details shown in the Results section).

This retrospective cohort study used a data set from the SART CORS database that included 256,018 patient records from 2014 to 2017. Of those, 128,703 (50%) met our study inclusion criteria, consisting of the following: women between the ages of 21 and 40 years with infertility; those with reported race and/or ethnicity; those with known infertility cause; and those undergoing their first autologous fresh IVF cycle. Of the 127,315 excluded cycles, 49,507 women (39%) were not undergoing their first IVF cycle, 76,165 patients (60%) did not report race, and 1,643 women (1.3%) selected multiple races. This study was approved by Institutional Review Board at the University of Illinois at Chicago (UIC IRB Protocol #2019-1441). Race and ethnicity were self-reported by the patient as African American, Asian, Hispanic, or white. Those who did not select any race and/or ethnicity were excluded. For race and ethnicity classification, we used the Inclusive Language section of the AMA Manual of Style ( 23 ). Polycystic ovary syndrome was defined as oligo-ovulation (<6 cycles/year), or anovulation, and “polycystic”-appearing ovaries. Controls were women with infertility who did not have PCOS. In addition to PCOS, infertility was classified as male factor, endometriosis, diminished ovarian reserve, tubal ligation (not reversed), tubal hydrosalpinx, and other tubal and uterine conditions. If no factor was identified, the patient was classified as having unexplained infertility. Male factor infertility was defined as abnormal semen parameters or function. Endometriosis included all stages, whether treated or not. Diminished ovarian reserve was defined as a single high follicle-stimulating hormone (FSH) level, high estradiol level (defined as the upper limit of laboratory’s normal range) measured in the early follicular phase or during a clomiphene challenge test, or reduced ovarian volume due to congenital, medical, surgical, or other causes. Other tubal conditions included any other tubal disease, including, but not limited to, pelvic or peritubal adhesive disease, prior tubal surgery, prior ectopic pregnancy, or tubal occlusion (partial or complete without hydrosalpinx). Uterine conditions included myomas, diethylstilbestrol exposure, intrauterine adhesions, and congenital anomalies (SART CORS Data Dictionary, www.sartcorsonline.com ; 9/24/2019 Update). Autologous IVF cycle was defined as using the patient’s own oocytes, fertilized with either her partner’s or donor sperm. Fresh embryo transfer was defined as transfer of the patient’s embryos generated from oocytes retrieved during the IVF cycle. Only fresh embryo transfer cycles were selected because they constituted the majority of the cases in the database during the 2014–2017 period. Demographic data included patients’ age, BMI, smoking status, and prior pregnancy status. Specific IVF cycle parameters included gonadotropin-releasing hormone medication and FSH dosage, AMH level, days of ovarian stimulation, number of oocytes retrieved, number of embryos transferred, implantation rate, pregnancy rate, LBR, pregnancy loss rate, multiple pregnancy rate, cesarean section (C-section) and vaginal delivery rates, birth weight, neonatal death rate, and congenital defects. Live birth rate was defined as the number of women who delivered a live born divided by the number of women who had an embryo transfer. Implantation rate was calculated using the number of women who had a fetal heartbeat on ultrasound (US) divided by the number of embryos transferred. Clinical pregnancy rate was calculated using the number of women who had a fetal heartbeat on US divided by the number of women who had an embryo transfer. Pregnancy loss rate was calculated using the number of miscarriages divided by the number of clinical pregnancies. Multiple pregnancy rate was defined as the number of multiple pregnancies (2 or more fetal heartbeats on US) divided by the number of women with embryo transfer. Neonatal death was calculated as the number of infant deaths divided by the number of live births. The C-section and vaginal delivery rates were calculated as the number of C-sections and vaginal deliveries, respectively, divided by the number of women who had a live birth. The SART CORS database contains comprehensive data from >90% of all clinics performing ART cycles in the United States. The data were collected through voluntary submission, verified by SART, and then reported to the Centers for Disease Control and Prevention in compliance with the Fertility Clinic Success Rate and Certification Act of 1992 (Public Law 102-493). The Society for Assisted Reproductive Technology maintains Health Insurance Portability and Accountability Act–compliant business associate agreements with reporting clinics. In 2004, after a contract change with the Centers for Disease Control and Prevention, The Society for Assisted Reproductive Technology gained access to the SART CORS data system for the purposes of conducting research. The Society makes data available for research purposes to entities that have agreed to comply with SART research guidelines. Patients undergoing treatment at SART member clinics sign clinical consent forms that include permission to use their data for research with appropriate provisions for safeguarding confidentiality. Data are submitted by individual clinics and verified by the medical director of each clinic. The data in the SART CORS are validated annually with select clinics having on-site visits for chart review on the basis of an algorithm for clinic selection. During each visit, data reported by the clinic were verified with information recorded in patients’ charts. Approximately 10% of clinics are audited each year to validate the accuracy of reported data. During each audit visit, data reported by the clinic are compared with information recorded in the medical record; 10 of 11 data fields selected for validation were found to have discrepancy rates of <5%. The exception was the diagnosis field, which, depending on the diagnosis, had a discrepancy rate between 2.1% and 9.2% ( 24 ). Using R ( 25 ), we compared means of continuous outcomes using 2-sided t-tests and analyses of variance and distributions of categorical outcomes with the chi-square tests. We fitted logistic regression models to the likelihoods of a live birth (primary outcome), clinical pregnancy, pregnancy loss, multiple pregnancy, and neonatal death (secondary outcomes). We fitted a Poisson rate model to the implantation rate and a linear regression model to birth weight. These models examined outcomes on the basis of PCOS status and their interaction with different races and ethnicities, adjusting for patient’s age, BMI, smoking status, and number of embryos transferred. We also compared race and ethnicity within each of the 2 groups, without adjusting for the confounders. For the regression models, the white women with infertility without PCOS ( Supplemental Table 5 , available online) and white women with infertility with PCOS ( Supplemental Table 6 ) were used as the referent groups. The figures include 95% confidence intervals around the model estimates. A P value of <.05 was considered statistically significant.

This is the largest IVF study to date assessing the impact of maternal race and ethnicity on pregnancy and neonatal outcomes in women with infertility with and without PCOS in the United States. Significant racial and ethnic disparities in reproductive outcomes were observed in this large PCOS cohort, which need to be confirmed by future large prospective studies. This is the largest study to date assessing the impact of race and ethnicity on pregnancy results and neonatal outcomes in women with infertility with and without polycystic ovary syndrome undergoing in vitro fertilization in the United States. These results suggest that there are significant racial and ethnic outcome disparities in women with infertility undergoing in vitro fertilization, with or without polycystic ovary syndrome. Further large prospective studies are urgently needed to confirm our study findings.

This study found significantly different LBRs and pregnancy loss rates in the racial minority groups with PCOS compared with those in white women with infertility with PCOS undergoing IVF in the United States. We also confirmed the previous findings by Stern et al. ( 17 ) that a higher proportion of white mothers in the United States underwent IVF treatment than other racial groups. Prior studies showed that all groups of minority women had lower LBRs when undergoing IVF than white women ( 13 , 16 , 26 , 27 , 28 ). This study went further by comparing race and ethnicity in women with infertility with and without PCOS. Our study demonstrated that African American and Asian women with PCOS were significantly less likely to have a live birth compared with white non-PCOS women, after adjusting for the patient's age, BMI, smoking status, and number of embryos transferred; this finding has not been previously reported in the literature. Another interesting finding in our study was that white women with PCOS had a significantly increased likelihood of having a live birth compared with white non-PCOS women. A previous meta-analysis of women with infertility with PCOS, without racial or ethnic group differentiation, found a significantly higher clinical pregnancy rate in women with PCOS than in non-PCOS women ( 29 ). Our results also confirm the finding in previous studies demonstrating a negative interaction between higher BMI and poor pregnancy and neonatal outcomes ( 4 , 6 , 12 ). In our study, African American women had a higher BMI and lower LBR when analyzed in each PCOS and non-PCOS group, separately. As described in other studies, obesity can affect IVF outcomes via multiple mechanisms (e.g., a systemic inflammatory state, oxidative stress, and hormonal dysregulation) ( 6 , 12 ) and could have played a factor in the adverse pregnancy outcomes noted in our study. The pregnancy loss rate in our study was significantly higher in African American and Hispanic women with PCOS than in the non-PCOS group. We hypothesize that the reason for LBRs differing by race and ethnicity may be, in part, due to the observed differences in pregnancy loss rates. Previous studies have confirmed that PCOS alone increased the risk of miscarriage compared with that in non-PCOS patients ( 29 ). Shapiro et al. ( 16 ) also showed that African American and Hispanic women with infertility had a higher chance of a miscarriage than those with other races and ethnicities after IVF. Sha et al. ( 29 ) reported significantly increased clinical pregnancy rates in patients with PCOS compared with those in non-PCOS patients. However, Fujimoto et al. ( 26 ) reported a lower clinical pregnancy rate in Asian women than in white women. Our study confirmed this result, with Asian women with PCOS being significantly less likely to have a clinical pregnancy compared with non-PCOS white women. The likelihood of having a multiple pregnancy was significantly higher in white women with PCOS than in our referent group (white non-PCOS women). Studying each group separately with and without PCOS, Hispanic women in both groups had more multiple pregnancies than the other women. No previous studies have shown similar results. An increase in stillbirth rates when comparing PCOS with non-PCOS women has been reported in previous studies ( 30 ). Our study found an overall significantly increased risk of neonatal deaths in patients with PCOS compared with that in non-PCOS patients, with the addition that, specifically, African American, Asian, and Hispanic women with PCOS were at increased risk of neonatal death compared with white non-PCOS women. However, our study did not differentiate between neonatal death and stillbirth. Our study showed a significantly lower birth weight in neonates of women of all races with PCOS than in those of white PCOS and non-PCOS women. Farland et al. ( 31 ) demonstrated in their study that women with PCOS also had neonates with a lower birth weight and were at increased risk of small for gestational age, as well as an increased risk of preterm delivery. This last finding could account for the increased neonatal death rate observed in our study. However, further studies would need to be performed to include gestational age at delivery because our data set lacked this information. Despite the large number of patients in this study (128,703 cycles), there are some limitations. Our study data were based on retrospectively collected information, and this could increase the risk of confounding factors. Race and ethnicity were self-reported by patients, which may have increased the risk of selection bias. In our initial cohort, 76,165 patients undergoing their first IVF cycle did not report race, and 1,643 selected multiple races, which resulted in their exclusion from our analysis. Neonatal outcomes were also self-reported by the mothers, which could have increased the risk of selection bias. Patients in both the PCOS and non-PCOS groups had multiple causes of infertility, which were not evenly distributed and could have biased our results. Another limitation in our study is that we only evaluated fresh embryo transfers between 2014 and 2017. Current IVF practice includes the freeze-all strategy as an alternative to fresh embryo transfers to minimize ovarian hyperstimulation and allow for preimplantation genetic testing ( 32 ). It is known that African American women demonstrate significantly higher serum estradiol levels across the menstrual cycle due to increased ovarian aromatase activity ( 33 ). We hypothesize that some of the negative IVF outcomes in African American women found in our study may be related to a suboptimal hormonal milieu during controlled ovarian hyperstimulation and fresh embryo transfer. Future studies should evaluate whether IVF outcomes in African American women are improved during freeze-all of embryos with subsequent frozen embryo transfer.

G.M.C. has nothing to disclose. F.D. has nothing to disclose. A.S. has nothing to disclose. B.S. has nothing to disclose.