Abnormal implantation after fresh and frozen in vitro fertilization cycles

Assisted reproductive technology contributes to 1.5% of all infants born in the United States, resulting in 66,000 infants in 2013 ( 1 ). Studies have demonstrated that singleton infants from in vitro fertilization (IVF) are a higher risk adverse perinatal outcomes ( 2 , 3 ). Recently, attention has turned to the supraphysiologic estradiol levels resulting from ovarian stimulation during IVF, which can affect the peri-implantation maternal environment and possibly placental cell migration and invasion ( 4 ). Endometrial gene expression profiles are significantly altered after ovarian stimulation compared to natural cycles ( 5 – 7 ), specifically for TGFβ signaling, leukocyte transendothelial migration, and the cell cycle ( 6 ). Several studies concluded that infants conceived in fresh IVF cycles have a higher risk of low birth weight and/or preterm delivery than infants conceived after frozen IVF cycles ( 8 – 10 ). On the other hand, there are also some reports of increased risk of preeclampsia and perinatal mortality associated with infants conceived with frozen-thawed embryos ( 9 , 11 , 12 ). Adverse perinatal outcomes associated with IVF may be secondary to abnormal placentation, and as such, it may be of value to investigate abnormal implantation as a much earlier manifestation of the pathophysiology associated with abnormal placentation. There are limited studies evaluating the effect of fresh and frozen IVF cycles on ectopic pregnancy ( 13 ) and first trimester pregnancy loss ( 14 ); however, there are no large studies evaluating abnormal implantation as a composite outcome or studies which account for the interaction between fresh and frozen IVF cycles and the embryo stage at transfer (cleavage-stage vs. blastocyst). To address this gap in knowledge, we conducted a retrospective cohort study with a large national database to determine whether fresh IVF cycles have a higher risk of abnormal implantation compared to frozen IVF cycles while accounting for the embryo stage at transfer.

The baseline characteristics of cycles included in the analyses are shown in Table 1 . Over the course of the study period from 2004–2013, the proportion of fresh and frozen blastocyst transfers steadily increased, while the proportion of fresh and frozen non-blastocyst transfers decreased ( Figure 1 ). Of the 509,938 cycles included in this study, 25.4% were fresh blastocyst, 49.8% fresh non-blastocyst, 13.9% frozen blastocyst, and 10.9% frozen non-blastocyst transfers. Given the large sample size, age, parity, and infertility diagnosis were significantly different across all 4 categories. Women were more likely to be <35 years of age in the fresh blastocyst group (62.4%). The number of oocytes retrieved in the fresh blastocyst group was higher than in the fresh non-blastocyst group (16.3 ± 7.7 vs. 12.4 ± 7.2). The number of embryos transferred was higher in the non-blastocyst transfer groups compared to the blastocyst transfer groups. Across the board, 31.8% of cycles resulted in abnormal implantation (16.9% biochemical, 1.5% ectopic/heterotopic, 13.4% first trimester pregnancy loss). In regression analyses, a fresh blastocyst transfer resulted in the lowest risk of abnormal implantation, with fresh non-blastocyst (OR 1.22, 95% CI 1.20–1.24), frozen blastocyst (OR 1.36, 95% CI 1.34–1.39), and frozen non-blastocyst (OR 1.57, 95% CI 1.54–1.61) having sequentially increased risk ( Table 2 ). Overall, frozen embryo transfers have a higher risk of abnormal implantation than fresh IVF cycles (OR 1.32, 95% CI 1.30–1.34) and non-blastocyst transfers have a higher risk of abnormal implantation than blastocyst transfers (OR 1.19, 95% CI 1.17–1.21). Together, the interaction was negative, suggesting that cryopreservation is a more significant factor than embryo stage at transfer for risk of abnormal implantation in this data analysis. Analyses for live birth demonstrated consistent results with fresh blastocyst transfers having the highest odds of live birth, followed by fresh non-blastocyst transfer (OR 0.82, 95% CI 0.80–0.83), frozen blastocyst transfer (OR 0.73, 95% CI 0.72–0.75), and frozen non-blastocyst transfer (OR 0.64, 95% CI 0.62–0.65) ( Table 2 ). We reanalyzed the data using 1) the subset of fresh embryo transfers with blastocyst transfer defined by both the blastocyst transfer data field and embryo morphology, and 2) the subset of fresh embryo transfers with blastocyst transfer defined by both the blastocyst transfer data field and day of transfer, as described in the methods section, and the results were unchanged for fresh non-blastocyst transfers compared to fresh blastocyst transfers (data not shown). When we investigated the independent outcomes of abnormal implantation, the results for biochemical pregnancy and pregnancy loss were consistent with the composite outcome ( Table 2 ). Fresh blastocyst transfers had the lowest risk of biochemical pregnancy and pregnancy loss. However, when we examined ectopic/heterotopic pregnancy as an independent outcome, fresh blastocyst and fresh non-blastocyst transfers were comparable, but frozen embryo transfers had a lower risk of ectopic/heterotopic pregnancy with frozen blastocyst transfers having the lowest risk ( Table 2 ). In an attempt to capture good prognosis cycles, we repeated the analysis in the subset of women with age <40 without the diagnosis of diminished ovarian reserve. The results were consistent with the main analysis in which a fresh blastocyst transfer resulted in the lowest risk of abnormal implantation, with fresh non-blastocyst (OR 1.18 95% CI 1.16–1.20), frozen blastocyst (OR 1.47, 95% CI 1.44–1.51), and frozen non-blastocyst (OR 1.80, 95% CI 1.64–1.73) having sequentially increased risk.

In this large retrospective cohort study, we concluded that the risk for abnormal implantation is lowest with fresh blastocyst transfers, followed sequentially by fresh non-blastocyst transfers, frozen blastocyst transfers, and frozen non-blastocyst transfers. Cryopreservation is a more significant adverse factor than the embryo stage at transfer. Our results were consistent in the subset of women <40 years of age without a diagnosis of diminished ovarian reserve, which are presumably better prognosis patients. This is the first paper to assess the risk for abnormal implantation based on the interaction of fresh vs. frozen embryo transfer and blastocyst vs. non-blastocyst embryo transfer. Over the last 10 years, we have seen an increase in blastocyst transfers, initially with fresh blastocyst transfers and then subsequently frozen blastocyst transfers ( Figure 1 ). At the end of the study period, fresh and frozen blastocyst transfers accounted for comparable proportions of all transfers taking place in the United States. Conversely, the proportion of fresh non-blastocyst transfers has dropped dramatically and frozen non-blastocyst transfers have also declined. One of the main reasons for cryopreservation or “freeze-all” of embryos has been attributed to a method of decreasing the risk of moderate-severe or delayed ovarian hyperstimulation syndrome ( 16 ). Some clinics are now advocating for a policy for elective cryopreservation of all cycles based on the endometrial peri-implantation environment ( 17 ), reduced pregnancy rates ( 18 , 19 ) and an increase in perinatal morbidity ( 8 – 10 , 20 ) associated with fresh embryo transfers. In terms of pregnancy outcomes, a meta-analysis of three trials of 633 cycles in women 27–33 years of age demonstrated higher clinical pregnancy rates and ongoing pregnancy rates in frozen embryo transfers compared to fresh embryo transfers ( 21 ). In terms of perinatal morbidity, most of the data is on the outcome of low birth weight. In a compelling study, Kalra et. al. utilized SART CORS data to demonstrate that infants conceived in fresh embryo transfer cycles have a higher risk of low birth weight than infants conceived after frozen embryo transfer cycles, potentially due to the supraphysiologic estradiol levels resulting from ovarian stimulation ( 8 ). The study also demonstrated that there was no difference in low birth weight in recipient cycles with fresh or frozen embryo transfers as the recipient is not stimulated in either case ( 8 ). In a meta-analysis of 11 observational studies, singleton pregnancies after the transfer of frozen embryos had a decreased risk of antepartum hemorrhage, preterm birth, small for gestational age, low birth weight, and perinatal mortality ( 10 ). Another recent study concluded that IVF singletons born after fresh embryo transfers had an increased odds of small for gestational age compared to non-IVF singletons, but this risk was reduced in IVF singletons born after frozen embryo transfers ( 22 ). Despite the large body of literature on improved outcomes associated with “freeze-all” cycles as described above ( 23 , 24 ), there remains some controversy regarding this approach and whether we can extrapolate the available data to a uniform policy for all patients In terms of pregnancy rates, Shapiro et al. reported results of a randomized controlled trial concluding that frozen blastocyst transfers have a higher pregnancy rate than fresh blastocyst transfers in normal responders under the age of 41 with 8–15 antral follicles ( 18 ) and high responders with >15 antral follicles ( 19 ). The results may not be applicable to poor responders or older women. Furthermore, for the cryopreservation arm of this trial, cryopreservation occurred at the 2PN stage, which were subsequently thawed and cultured to the blastocyst stage. Another recent study reported an improved live birth rate with cryopreserved embryos among women with polycystic ovarian syndrome ( 11 ); however, over 95% of patients underwent transfers of cleavage-stage embryos, making the data less applicable to the United States as we move toward blastocyst transfers. In terms of adverse perinatal outcomes, there are some reports of an increased risk of preeclampsia associated with infants conceived with frozen embryos ( 11 , 12 ). One observational study based on the Nordic population even reported that singletons born after frozen embryo transfer was associated with increase in perinatal, neonatal, and infant mortality ( 9 ). Adverse perinatal outcomes may be secondary to abnormal placentation, and as such, abnormal implantation may be a much earlier manifestation of the pathophysiology associated with abnormal placentation. We have previously shown that IVF pregnancies are associated with a higher risk of retained placenta ( 25 ) and severe maternal mortality largely due to hemorrhage ( 26 ), compared to spontaneous conceptions. Other studies have also suggested that IVF pregnancies have an increased risk of placenta previa and placenta abruption ( 27 ), as well as a higher incidence of marginal cord insertion and abnormal placental shapes ( 28 ). Small studies on first trimester trophoblast and placental bed vascular volume as an indicator of normal placentation did not show a difference between IVF and spontaneous conceptions ( 29 ). Frozen embryo transfers may create an environment more prone to abnormal placentation, which in turn translates to the higher risk of biochemical pregnancy and pregnancy loss observed in this study. Our data adds to the above-mentioned body of evidence and suggests that use of frozen-thawed embryos may be associated with a higher risk of abnormal implantation. There are a few studies that have examined the association of frozen embryos with ectopic pregnancy and pregnancy loss, but no studies included biochemical pregnancies. Our study confirms the results of a recent paper utilizing the Centers for Disease Control and Prevention’s National ART Surveillance System, which reported an increased risk of first trimester pregnancy loss after the transfer of frozen embryos compared to fresh embryos in women <38 years of age ( 14 ). This difference persisted in women <30 years of age even in a subanalysis to correct for embryo quality of fresh and frozen cycles among first transfer cycles only. However, this study did not account for embryo stage at transfer (cleavage-stage vs. blastocyst) ( 14 ), which has a significant interaction with the outcomes of fresh vs. frozen embryos and may explain the slight differences between our two studies. Our study is consistent with the existing literature which demonstrates a lower risk of ectopic/heterotopic pregnancy in frozen embryo transfer cycles. Another study of SART data including blastocyst transfers only from 2008–2011 concluded that the odds of ectopic pregnancy were significantly lower in women who had a frozen transfer compared to a fresh transfer in autologous cycles ( 13 ). Fan et. al. examined a four-level predictor similar to our study and concluded that frozen blastocyst transfers had the lowest rate of ectopic pregnancies, followed by frozen day 3 transfers, fresh blastocyst transfers, and fresh day 3 transfers ( 30 ). Several mechanisms have been proposed to explain the increase risk of ectopic pregnancy for a fresh embryo transfer compared to a frozen embryo transfer. The supraphysiologic hormone levels associated with ovarian stimulation may affect uterine contractility ( 31 ) or have a negative effect on endometrial receptivity ( 5 – 7 ), both of which could influence ectopic implantation. The actual transfer technique is unlikely to play a role as we would not expect technique to vary in fresh versus frozen embryo transfers. Our study is restricted primarily by the limited details available for frozen embryo transfers, both in terms of methodology and indication, in the SART dataset. We could not account for the method by which embryos were cryopreserved., nor could we account for the type of endometrial preparation protocols used for frozen embryo transfers, in terms of exogenous hormones or the length of time used to prepare the endometrium. We did not address cryopreservation cycles due to preimplantation genetic screening. We were also unable to account for elective “freeze-all” cycles in which the “highest quality” embryos were not transferred in a fresh cycle but cryopreserved instead. There are inherent biases for frozen embryo cycles in both directions. On one hand, patients undergoing frozen embryo cycles may be the better prognosis patients based on the assumption that they had supernumerary embryos to freeze. On the other hand, older patients or poor prognosis patients may opt for multiple fresh IVF cycles to bank embryos and thereby improve their success rates with a frozen embryo cycle. We attempted to correct for embryo quality and good prognosis patients with a secondary analysis of women <40 years of age without a diagnosis of diminished ovarian reserve, which presumably were the better prognosis patients and the results were consistent with the overall cohort. We recognize that clinical and laboratory practices have changed significantly during the 10-year period of this study. All analyses were covaried for calendar year of treatment to avoid confounding due to changes in clinical practice over time. However, a full analysis of these changes is beyond the scope of this study. The strengths of our study include utilization of a large national database incorporating data from multiple clinics, assessing a four-level predictor of accounting for fresh vs. frozen embryos and the embryo stage at transfer, and investigating a composite outcome of abnormal implantation. We believe that the composite outcome allows for clinically relevant information for patients once a positive pregnancy test is noted. In conclusion, our study contributes to the body of literature debating the risks and benefits of fresh versus frozen embryo cycles. Fresh blastocyst transfers had the lowest overall risk of abnormal implantation, but a higher risk of ectopic/heterotopic pregnancy. Although embryo cryopreservation is indicated in certain treatment cycles, elective embryo cryopreservation of all cycles may not be the optimal strategy to adopt.

The data in this study are provided by the Society of Assisted Reproductive Technology Clinical Outcome Reporting System (SART CORS). The database includes de-identified demographic and clinical information, cycle characteristic information, and pregnancy outcomes reported by >90% of ART clinics in the United States. Data were collected and verified by SART and 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 data in the SART CORS are validated annually with some clinics having on-site visits for chart review based on an algorithm for clinic selection. During each visit, data reported by the clinic were compared with information recorded in patients’ charts. Ten out of 11 data fields selected for validation were found to have discrepancy rates of ≤5% ( 15 ). Specifically, the “treatment outcome” field and the “pregnancy outcome” field used in this study had a discrepancy rate of 1.8% and 2.3%, respectively. This study was approved by the Institutional Review Board at Cedars-Sinai Medical Center (Pro00031667). We analyzed data from 509,938 fresh and frozen autologous IVF cycles from 394,247 patients which resulted in a positive pregnancy test between 2003–2014. We excluded cycles in which both fresh and thawed embryos were transferred. Given that the focus of this project was to investigate abnormal implantation and early pregnancy loss, we excluded cases of pregnancy loss after the first trimester, stillbirth, maternal death, and unknown or unentered outcomes. By utilizing a specific data field which was the pregnancy outcome date minus the transfer date plus 14 days, we excluded cycles which resulted in second trimester losses (defined as 14 0/7 to 27 6/7 gestational weeks) and third trimester losses (defined as >28 gestational weeks). The data were parameterized into a four-level predictor of 1) fresh blastocyst transfer, 2) fresh non-blastocyst transfer, 3) frozen blastocyst transfer, and 4) frozen non-blastocyst transfer. There is a specific data field in SART CORS to indicate blastocyst transfer for both fresh and frozen embryo transfers, which we utilized for our main analyses. We cross-validated this data field for fresh transfer cases by both embryo morphology data and day of transfer data. We considered embryo morphology documented as early blastocyst, expanded blastocyst, and hatching blastocyst to be blastocysts, and all other embryo morphology (i.e. 1–8 cells, > 8 cells, and morula) to be non-blastocysts. Embryo morphology data indicated a 90% accuracy rate compared to the blastocyst transfer data field (kappa agreement value of 0.76). The average day of transfer for a blastocyst transfer was 5.04 ± 0.29 days and 3.34 ± 0.89 days for a non-blastocyst transfer. We considered embryos transferred on day 5 or later to be blastocyst transfers. The day of transfer data indicated an 89% accuracy rate compared to the blastocyst transfer field (kappa agreement value of 0.76). The primary outcome of interest was abnormal implantation, defined as a composite outcome to include biochemical pregnancy, ectopic/heterotopic pregnancy, and first trimester pregnancy loss. We also analyzed the individual components of the composite outcome. Live birth, the inverse of abnormal implantation, was a secondary outcome of interest. Baseline and demographic characteristics was analyzed across categories with either ANOVA for continuous data or Chi-square test for categorical data. Body mass index and race were excluded for analysis given a significant proportion of missing data and data errors. Regression modeling to predict factors contributing to abnormal implantation was performed with repeated measures multivariable logistic regression techniques using an exchangeable covariance structure for intra-subject observations, which assumes that the observations within each subject are equally correlated. Regression modeling accounted for possible confounders including patient age (at the time of oocyte retrieval), parity, number of embryos transferred, infertility diagnosis, and calendar year of treatment. We conducted secondary analyses to investigate each outcome individually independent of the composite outcome – i.e. risk of biochemical pregnancy, risk of ectopic/heterotopic pregnancy, and risk of first trimester pregnancy loss. We conducted a secondary analysis for good prognosis patients, defined as women with age <40 without a diagnosis of diminished ovarian reserve. Means are presented with standard deviations (SD) and Odds Ratios (OR) are presented with 95% Confidence Intervals (CI). All data was tested at the two-sided alpha level of 0.05. Analysis performed using SAS v 9.4.