Ovarian hyperstimulation syndrome after assisted reproductive technologies: trends, predictors, and pregnancy outcomes.

Data used in this study were obtained from the NASS, which is maintained by the Centers for Disease Control and Prevention. Reporting to NASS is federally mandated (Fertility Clinic Success Rate and Certification Act of 1992, Public Law No. 102–493, October 24, 1992), and therefore it includes data from an estimated 97% of IVF cycles performed in the United States ( 23 ). NASS collects information on patient characteristics, medical and obstetric history, infertility diagnosis, treatment parameters, and outcomes of IVF cycles performed in the United States. Annual data validation, during which data for selected variables are compared with information recorded in medical records, was performed in ~7%–10% of reporting clinics during the period under study (2000–2015). In general, discrepancy rates for 2015 ART cycle data were low (<4%) ( 24 ). Inclusion and exclusion criteria for cycles varied based on the type of analysis being performed. To examine trends in the rate of OHSS, we included all fresh cycles (autologous and donor) and embryo-banking cycles combined that were performed from 2000 to 2015 (n = 1,833,430 cycles). Data regarding trends in GnRH antagonist use were limited to cycles performed from 2004 to 2015 in which embryo banking was not performed, because GnRH antagonist use was not collected before 2004 and is not collected for embryo-banking cycles. We calculated the rates of moderate OHSS, severe OHSS, any OHSS, and GnRH antagonist use per 1,000 IVF cycles. We calculated the percentages of OHSS among fresh autologous and embryo-banking cycles over the entire study period. We used SAS version 9.3 to perform simple linear regression to determine the statistical significance of changes in trends of OHSS diagnosis and GnRH antagonist use over time. To reflect more recent practice patterns, for all subsequent analyses we included aggregated data from fresh autologous and embryo-banking cycles performed from 2010 to 2015. We compared the distribution of patient and treatment characteristics for cycles complicated by moderate or severe OHSS, complicated by any OHSS, and without OHSS. Among fresh autologous and embryo-banking cycles, we compared the distribution of patient age (years) at oocyte retrieval, race/ethnicity, BMI at time of initiation of IVF cycle, number of previous pregnancies conceived by any means, IVF or otherwise, miscarriages, preterm or full-term births, infertility diagnosis, hospitalization for a complication, and cancellation before retrieval for cycles with and without OHSS. Use of GnRH receptor antagonist or agonist for pituitary suppression was compared for fresh autologous cycles only, because this information is not reported for embryo-banking cycles. Comparisons of number of oocytes retrieved were restricted to fresh autologous and embryo-banking cycles in which oocyte retrieval was performed. The proportion of cycles in which a transfer was attempted was calculated only for fresh autologous cycles with an oocyte retrieval. Among fresh autologous cycles with transfer of at least one embryo, we compared distributions of number of supernumerary embryos cryopreserved, maximum number of fetal hearts before reduction (if any), positive hCG (including intrauterine, biochemical, and ectopic pregnancies), live birth, and plurality for cycles with and without OHSS. Pearson chi-square test was performed to assess homogeneity for distributions of each maternal and treatment characteristic for cycles with any type of OHSS versus those without OHSS. We used multivariable predicted marginal proportions from logistic regression models to assess predictive factors for OHSS in two sample groups. One group included data from all fresh autologous IVF cycles that resulted in embryo transfer. In this group, multivariable logistic regression was used to calculate adjusted relative risk (aRR) of developing OHSS for the following factors: number of oocytes retrieved, infertility diagnosis, BMI, age, pregnancy following fresh embryo transfer, use of a GnRH antagonist, and history of one or more previous pregnancies conceived by any means, IVF or otherwise. The second group included all noncancelled fresh autologous and embryo-banking cycles. The same variables were included in this second model, with the exception of GnRH antagonist use, pregnancy, and the number of oocytes retrieved per cycle. These variables were excluded because GnRH antagonist use is not reported in embryo-banking cycles, pregnancy is not possible after an embryo-banking cycle, and the number of oocytes retrieved was not reported for 7% of embryo-banking cycles. To assess pregnancy outcomes associated with OHSS, we restricted the population to fresh autologous IVF cycles resulting in pregnancy and examined outcomes separately for singleton pregnancies and pregnancies with multiple fetuses. The outcomes that we examined were first-trimester pregnancy loss (<14 weeks), second-trimester pregnancy loss (14–20 weeks), stillbirth after 20 weeks of gestation, low birth weight (<2,500 g) for either infant in a multiple pregnancy, and preterm delivery (delivery at <37 completed weeks of gestation) among live births. We calculated the absolute number and percentage of each pregnancy outcome in cycles with OHSS and cycles without OHSS. We used multivariable predicted marginal proportions from logistic regression models to calculate the aRR for each pregnancy outcome in cycles with OHSS compared with cycles without OHSS, adjusting for known confounders. For first- and second-trimester pregnancy losses, we adjusted for maternal age, race/ethnicity, number of previous miscarriages, BMI, and uterine-factor infertility. For low birth weight and preterm delivery, we adjusted for age, race, number of previous preterm deliveries, BMI, and uterine-factor infertility. For stillbirth, we adjusted for maternal age, race, BMI, uterine-factor infertility, and number of previous pregnancies ( 25 – 27 ). We also reported the average birth weight in grams and gestational age in weeks for singleton and multiple gestations in pregnancies in which OHSS occurred. Approximately 19% of all cycles were missing information regarding BMI, and 25% of cycles did not include data on race/ethnicity. For this reason, we used multiple imputation (using the SUDAAN HOTDECK procedure under the assumption of missing at random) to derive missing BMI and race/ethnicity data to complete all regression analyses. The assumption that BMI data are missing at random is reasonable because, although reporting of BMI and race/ethnicity likely varies from clinic to clinic, there is no reason to suspect that missingness of the data varies systematically within a specific clinic. Use of multiple imputation with the use of the HOTDECK procedure for data contained within the NASS dataset has been reported previously, and studies of other large datasets support the validity of our assumption that data are missing at random ( 28 – 30 ). Five multiply imputed datasets were created to compute within- and between-variations of imputed datasets in parameter estimates, and the SUDAAN RLOGISTIC procedure was applied to fit logistic ( 28 – 30 ) regression models with the use of imputed datasets. Variables used for BMI imputation included state where the cycle was performed, patient race, reason for ART, gestational age, parity, patient age, number of previous preterm births (<37 weeks), number of previous spontaneous abortions (<20 weeks), and total number of embryos transferred. To further explore whether embryo banking cycles were associated with lower rates of OHSS, we stratified infertility clinics into quartiles based on the percentage of embryo banking cycles that they performed over the study period. We then calculated the rate of OHSS within each quartile and assessed trends in the rate of OHSS by quartile using simple linear regression. SAS version 9.3 and SUDAAN version 11.0 were used for the analyses. SUDAAN was used for ease of analysis of clustered imputed data. Statistical significance was determined using a two-side alpha of 0.05. This study was approved by the Institutional Review Board of the Centers for Disease Control and Prevention (IRB protocol no. 2238).

The rate of any OHSS occurring in women undergoing fresh autologous and embryo-banking cycles increased significantly from 9.96 to 14.34 cases per 1,000 from 2000 to 2006 ( P trend < .01) and then decreased significantly to 5.27 cases per 1,000 cycles from 2006 to 2015 ( P trend < .0001; Fig. 1 ). Similar trends were seen in the rates of moderate and severe OHSS. The rate of moderate OHSS increased significantly from 8.10 to 10.51 cases per 1,000 cycles from 2000 to 2006 ( P trend < .05) and then decreased significantly to 4.22 cases per 1,000 cycles from 2006 to 2015 ( P trend < .001). The rate of severe OHSS increased significantly from 1.87 to 3.83 cases per 1,000 cycles from 2000 to 2006 ( P trend < .0001) and then decreased significantly to 1.06 cases per 1,000 cycles from 2006 to 2015 ( P trend < .0001). The rate of GnRH antagonist use in IVF cycles increased significantly from ~279 per 1,000 cycles in 2004 to 634 per 1,000 cycles in 2015 ( P < .0001; Fig. 1 ). OHSS was more prevalent, and a higher proportion of IVF cycles that resulted in any type of OHSS occurred among younger women and those who had fewer previous pregnancies, miscarriages, and full-term births than fresh autologous and embryo-banking cycles uncomplicated by OHSS ( Table 1 ; Supplemental Table 1 [available online at www.fertstert.org ]). GnRH agonists and GnRH antagonists were used in a greater proportion of IVF cycles that resulted in OHSS than in fresh autologous and embryo-banking cycles that did not result in OHSS. Cycle cancellation occurred in a lower proportion of cycles that resulted in OHSS than in cycles that did not result in OHSS. OHSS was more prevalent in IVF cycles that had a diagnosis of ovulatory disorder, male factor, and diminished ovarian reserve. Twenty or more oocytes were retrieved in a greater proportion of cycles that resulted in OHSS than in cycles that did not result in OHSS. Among cycles in which OHSS occurred, hospitalization was more common, the number of supernumerary embryos that were cryopreserved was higher, pregnancy was more likely to occur, and multiple gestation was more common than in cycles uncomplicated by OHSS ( Table 1 ; Supplemental Table 1 ). Among all fresh autologous and embryo-banking IVF cycles, the risk of OHSS increased with decreasing age ( Table 2 ), with patients under the age of 30 years having the highest risk of OHSS (aRR 6.47, 95% confidence interval [CI] 5.46–7.66) compared with cycles in which the patient was ≥41 years old. Non-Hispanic black (aRR 1.37, 95% CI 1.25–1.49) and Hispanic (aRR 1.11, 95% CI 1.02–1.21) women were at an increased risk of OHSS compared with non-Hispanic white women. There was a significantly lower risk of OHSS in women of Asian/Pacific Islander race (aRR 0.86, 95% CI 0.79–0.92), of other race (aRR 0.49, 95% CI 0.25–0.94), and with diminished ovarian reserve (aRR 0.27, 95% CI 0.24– 0.31). There was also a significantly lower risk of OHSS in women who were overweight (aRR 0.86, 95% CI 0.80–0.92) or obese (aRR 0.70, 95% CI 0.65–0.76) or had previous pregnancy conceived by any means, IVF or otherwise (aRR 0.84, 95% CI 0.80–0.89). The risk of OHSS was also significantly increased in cycles resulting in pregnancy (aRR 3.12, 95% CI 2.81–3.47) compared with cycles that did not result in a pregnancy. There was a significantly increased risk of OHSS in cycles among women with a diagnosis of ovulatory disorder, a diagnostic category which includes PCOS and any other type of ovulatory disorder (aRR 2.61, 95% CI 2.47–2.76), uterine-factor infertility (aRR 1.17, 95% CI 1.05–1.31), or tubal-factor infertility (aRR 1.14, 95% CI 1.07–1.23) compared with cycles among women without these diagnoses, as well as in embryo-banking cycles (aRR 0.66, 95% CI 0.61–0.72) compared with women without these factors ( Table 2 ). Among women undergoing fresh autologous IVF cycles in which an egg retrieval occurred, the aRR of ovarian hyperstimulation increased as the number of oocytes retrieved increased and was highest for cycles in which more than 30 oocytes were retrieved (aRR 3.85, 95% CI 3.36–4.41) compared with cycles in which 11–15 oocytes were retrieved ( Table 2 ). The risk of OHSS was reduced in cycles in which a GnRH antagonist instead of a GnRH agonist was used for pituitary suppression (aRR 0.79, 95% CI 0.73– 0.85). In fresh autologous IVF cycles resulting in a singleton pregnancy, OHSS was associated with a significantly higher risk of low birth weight (aRR 1.29, 95% CI 1.10–1.51) and preterm delivery (aRR 1.32, 95% CI 1.15–1.51) compared with cycles without OHSS that resulted in pregnancy ( Table 3 ). The unadjusted mean birth weight for singleton pregnancies resulting from cycles in which OHSS occurred was 3,101 g (95% CI 3,065–3,138 g), and the unadjusted mean gestational age at which delivery occurred was 38.4 weeks (95% CI 38.2–38.6 weeks) of gestation. Among singleton pregnancies in which OHSS did not occur, the unadjusted mean birth weight was 3,230 g (95% CI 3,225–3,237 g) and the unadjusted mean gestational age at which delivery occurred was 38.7 weeks (95% CI 38.7–38.8 weeks). In twin pregnancies, OHSS was associated with a significantly increased risk of preterm delivery (aRR 1.16, 95% CI 1.10– 1.22) and low birth weight (aRR 1.06, 95% CI 1.02–1.12). In all multiple gestations, OHSS was associated with an increased risk of second-trimester loss (aRR 1.81, 95% CI 1.17–2.81). The unadjusted mean gestational age for twin pregnancies resulting from cycles in which OHSS occurred was 34.6 weeks (95% CI 34.3–34.9 weeks) compared with 35.6 weeks (95% CI 35.6–35.6 weeks) for twin pregnancies in which OHSS did not occur. The unadjusted mean birth weight for twin pregnancies resulting from cycles in which OHSS occurred was 2,206 g (95% CI 2,171–2,242 g) compared with 2,352 g (95% CI 2,347–2,356 g) for twin pregnancies resulting from cycles in which OHSS did not occur. All differences in gestational age at delivery and birth weight were tested for statistical significance with the use of the Welch-Sattherwaite t test and were found to be significantly different ( P < .0001 in each case). There was no significant difference in the risk of first-trimester loss or stillbirth in IVF cycles affected by OHSS for singleton or multiple gestation. Stratifying IVF clinics into quartiles by percentage of embryo-banking cycles that they perform revealed a trend of decreasing incidence of severe OHSS as the mean percentage of embryo-banking cycles increased ( P=. 04; Supplemental Table 2 [available online at www.fertstert.org ]). Of note, among clinics in the highest quartile of embryo banking cycles, an average of 37.9% of IVF cycles performed were embryo-banking cycles.

The results of this national study indicate that the rate of OHSS increased from 2000 to 2006 and then decreased from 2006 to 2015. The rates of severe OHSS, which is associated with the highest risk of significant morbidity and mortality, were similar at the beginning and the end of the study period. Although the decreasing rate of OHSS from 2006 to 2015 is encouraging, there is considerable room for improvement in preventing OHSS. Determining national trends in the diagnosis of OHSS is a key step in increasing provider awareness of this serious but potentially preventable condition. The NASS dataset provided a rare opportunity to examine pregnancy outcomes associated with OHSS, for which only limited data are available in the published literature. The cause of the increased rate of OHSS from 2000 to 2006 is unclear; it may reflect changes in medication regimens, changes in patient characteristics, and other factors. The decrease in the rate of OHSS from 2006 to 2015 may reflect the increasing use of GnRH antagonists, which is an effective and well-studied means of preventing OHSS. The use of other measures that reduce the risk of OHSS, many of which were first introduced into practice during the study period, have likely had an impact on the rate of OHSS as well, although it is not possible to study these with the use of the NASS dataset. These measures include use of a GnRH agonist trigger ( 31 ), cabergoline administration after COH ( 5 ), treatment with metformin before and during COH in patients with PCOS ( 32 , 33 ), and cryopreserving all embryos to avoid pregnancy immediately after COH ( 6 ). These measures have, to varying degrees, been shown to be effective in preventing OHSS but they could not be assessed in the present study because, except for embryo cryopreservation, this information is not collected in the NASS. The rate of moderate OHSS and any type of OHSS declined steadily from 2006 to 2015, although the rates of severe OHSS at the beginning and at the end of the study period were similar. This suggests that although efforts by practitioners of ART to reduce the risk of OHSS during COH have been effective, there is room for improvement in reducing the risk of the form of OHSS that is responsible for the most severe clinical sequelae ( 33 ). We observed a higher rate of OHSS in cycles performed in the United States with the use of the NASS dataset than the rate of OHSS in cycles performed in several European countries according to European IVF Monitoring (EIM) Consortium data. In the most recent report issued by the EIM, the incidence of moderate and severe OHSS in all cycles included in the EIM dataset was 0.3% ( 34 ). In 2000, 1.1% of all cycles included in the EIM dataset were affected by OHSS ( 35 ). In the most recent data from the International Committee for Monitoring Assisted Reproductive Technology (ICMART), which includes data from 2011, the incidence of OHSS was 0.5% ( 36 ). As with NASS data, data regarding OHSS from EIM and ICMART are not validated. Review of validated incidence data for diagnosis of OHSS from the Australian and New Zealand Assisted Reproduction Database revealed that in 2000 2.1% of women were hospitalized for OHSS, and in 2015 0.6% of cycles were affected by OHSS ( 37 ). Although NASS, EIM, and ICMART data regarding OHSS are not validated, this would be expected to yield an underestimate of the true incidence of OHSS. Accordingly, the differences in the incidence of OHSS from data obtained from NASS compared with international datasets were somewhat surprising. The differences in reported incidence may be due to several factors, including the subjective nature of diagnosing and reporting cases of OHSS in the United States and, possibly, differences in notification requirements between the United States and Europe. For example, with the introduction of the European Union Tissue and Cells Directive, serious adverse events (SAEs) are reportable to the NOTIFY library, which is a global surveillance database for medical products of human origin. OHSS that requires intensive care is considered a reportable SAE in Europe, whereas no such reporting requirements exist in the United States ( 38 ). There may be differences in the practice of ART in the United States and Europe that result in a genuinely reduced incidence of OHSS in the United States. The causes of the observed differences in the reported rates of OHSS between the United States and Europe require further study and reveal a need to validate OHSS data in national ART surveillance systems. In the present study, the predictive factors for OHSS that are potentially modifiable included retrieval of more than 15 oocytes, pregnancy after fresh embryo transfer, and type of pituitary suppression. Pregnancy after an IVF cycle was associated with an increased risk of OHSS, suggesting that not performing an embryo transfer in patients at high risk of OHSS may reduce the risk of OHSS. These findings are similar to the results of previous studies ( 3 , 8 , 9 , 14 ) and are consistent with what is known about the physiologic events that result in OHSS. It is likely the case that as the number of corpus lutea increases, the amount of VEGF that is produced following exposure to hCG increases, therefore increasing the risk of OHSS in women who respond vigorously to COH, such as those who have PCOS. The finding that retrieval of a large number of oocytes increases the risk of OHSS indicates that careful, limited ovarian hyperstimulation in patients who are likely to be high responders may provide another means of preventing OHSS. There are many factors that may motivate IVF providers to attempt to maximize the oocyte yield of each IVF cycle, including a desire to maintain a high live birth rate per IVF cycle. Recent data have suggested, however, that rather than attempting to maximize oocyte yield, COH with the goal of retrieving 11–15 oocytes per cycle may produce an optimal live birth rate and limit the risk of OHSS. This was demonstrated in at least two large studies showing that the relationship between oocyte yield and live birth rate plateaus in cycles in which more than 15 oocytes were retrieved ( 39 , 40 ). In our study and in previous studies, the risk of OHSS increases significantly in cycles in which more than 15 oocytes are retrieved ( 39 ). Considering the maternal and pregnancy-related risks associated with OHSS, it is important for IVF providers to recognize that limiting COH in high responders may prevent OHSS without sacrificing live birth rate. We found that use of a GnRH antagonist and embryo banking were associated with a lower risk of OHSS. The efficacy of embryo banking as a means of preventing OHSS has been uncertain ( 6 , 7 ) owing to conflicting data from previous studies ( 13 , 14 ). In the present study, cycles in which embryo banking was performed were 35% less likely to result in OHSS than cycles in which embryo banking was not performed. Because some clinics may routinely cryopreserve all embryos without transfer to limit the risk of OHSS, we stratified clinics by the proportion of embryo cryopreservation cycles performed annually. There was a significant trend toward a lower incidence of severe OHSS as the proportion of embryo-banking cycles that a clinic performed increased. Therefore, withholding embryo transfer and cryopreserving all embryos after a cycle of IVF may reduce the risk of developing severe OHSS. Our findings regarding the protective effects of GnRH antagonist use are also consistent with the findings of previous studies ( 4 , 41 ). However, we were unable to assess whether using a GnRH agonist trigger also reduced the risk for OHSS, which has been shown in other studies ( 31 ). Predictive factors for OHSS that are not readily modifiable included diagnoses of ovulatory disorder/PCOS, uterine factor, tubal factor, endometriosis, and diminished ovarian reserve, age, and BMI. The increased risk of OHSS in women with uterine- or tubal-factor infertility or endometriosis may reflect a higher ovarian reserve in women with these specific diagnoses compared with women without these diagnoses, who may have had diminished ovarian reserve. Whether these diagnoses independently increase the risk of OHSS requires further study. Similar to previous smaller studies ( 9 , 12 ), we found that the risk of OHSS was increased in younger women and was highest for women under 30 years of age. Low BMI was significantly associated with higher risk of OHSS when embryo-banking cycles were excluded. Over-weight and obese BMI was associated with a decreased risk of OHSS. Given the small variations in relative risk of OHSS, the clinical significance of BMI in determining which patients are at risk of OHSS may be limited. The decreased risk of OHSS associated with diminished ovarian reserve is consistent with the pathophysiology of OHSS, because the amount of VEGF produced during COH varies with number of follicles and their response to exogenous gonadotropin administration ( 42 ). Fewer corpus lutea will likely result in production of less VEGF and a lower risk of developing OHSS. Although we found that OHSS was associated with an increased risk of preterm delivery and low birth weight, the unadjusted differences in average length of pregnancy and birth weight were relatively small and likely not clinically significant. The one notable exception is that for patients with twins, the average gestational age at delivery was less than 37 weeks for all cycles and, on average, nearly 1 week earlier in cycles with OHSS than in cycles without OHSS. Given that a difference of 1 week may significantly alter morbidity and mortality in babies born before 37 weeks ( 42 – 45 ), the effect of OHSS on preterm delivery and birth weight may be more clinically relevant in twin pregnancies. There are several limitations inherent in using data from the NASS dataset. Information regarding race/ethnicity and BMI is inconsistently reported, resulting in missing data in a high percentage of cycles. We used multiple imputation to impute race/ethnicity and BMI data for cycles in which it was missing, allowing us to overcome this limitation. We also lacked specific biochemical and hormonal information about each patient, and therefore we were not able to evaluate certain potential risk factors, such as increased ovarian reserve. Similarly, we were not able to evaluate how frequently many measures intended to prevent OHSS were used, including the use of GnRH agonist triggers, cabergoline, or metformin, because data regarding these characteristics and interventions are not included in the NASS dataset. Although this study uses data provided by ART practitioners, it is not possible to determine what criteria were used in establishing a diagnosis of OHSS and to what extent OHSS was reported. Finally, it is possible that residual confounding biased our effect estimates. In summary, this study demonstrates that OHSS remains a serious complication of ART that affects not only women undergoing ART, but also the outcomes of pregnancies that result from it. Physicians who perform ART have many effective means available to reduce the risk of OHSS in their patients. These include careful controlled ovarian hyperstimulation with a goal of obtaining ~15 mature oocytes in all patients undergoing IVF, use of a GnRH antagonist and, when appropriate, a GnRH agonist trigger in women who are at moderate or high risk of experiencing OHSS, use of metformin and low-dose aspirin in patients who are at high risk of experiencing OHSS, use of cabergoline at the time of trigger, and, when appropriate, embryo cryopreservation and subsequent frozen-thawed embryo transfer in lieu of fresh embryo transfer. Physicians who practice ART should make use of these whenever possible to limit morbidity to patients and the risk of adverse pregnancy outcomes.