Comparison of obstetric outcomes between single and double embryo transfers among singleton live births from vitrified donor oocyte IVF cycles: a cohort study

We utilized a retrospective cohort study of nonidentified vitrified oocyte donation IVF cycles that occurred between 2008 and 2016 at Reproductive Biology Associates (RBA) in Sandy Springs, Georgia [ 30 ]. Although RBA is part of a national egg bank, for our analysis, we only utilized donors with oocytes retrieved and vitrified at RBA, the original clinic. Similarly, all recipient couples had oocytes thawed and embryos transferred at RBA. During this time frame, there were 856 embryo transfer cycles that resulted in a live birth. From there we excluded non-singleton births (n=149), cycles that had more than two embryos transferred as there were too few cycles to meaningfully analyze (n=2) as well as cycles with missing pregnancy outcome data (n=7). Ultimately, we included 698 singleton births from 273 unique donors and 610 unique recipients. 86 recipients had more than one embryo transfer cycle that resulted in live singleton birth during the timeframe. Briefly, oocyte donors were screened prior to donation according to ASRM recommendations [ 31 ]. Antagonist protocols were used for controlled ovarian stimulation. Once collected, mature oocytes were vitrified 39–40 hours after trigger injection using the ‘minimum volume’ method [ 32 ]. At the time of the first oocyte retrieval, donors provided demographic information and reproductive history which were collected through chart abstraction. For endometrial preparation, recipients received standard protocols including leuprolide acetate, estrogen, and progesterone. Oocytes were warmed 2 to 3 hours prior to fertilization using ICSI with either male partner or donor sperm [ 33 ]. Embryos were transferred during either the cleavage stage (Day 3) or the blastocyst stage (Day 5/6) with high quality embryos being transferred first and the rest being cryopreserved for future use. DET was typically done due to patient request. It was possible for recipients to have multiple frozen embryos transfers from a single cohort of oocytes [ 34 ]. None of the transferred embryos underwent pre-implantation genetic testing. Live births occurring before 37 weeks were considered preterm. Infants weighing less than 2500 grams were considered low birth weight [ 35 , 36 ]. We examined donor and recipient characteristic by the number of embryos transferred for each singleton live birth using simple descriptive statistics. To describe the association between the number of embryos transferred and birth outcomes we used cluster weighted generalized estimating equations (GEE) models. All of these models accounted for the fact that women could contribute more than one singleton live birth to the analysis. To account for the presence of this non-ignorable cluster size, we weighted each recipient inversely according to the number of singleton live births they achieved (i.e. if a woman had 2 singleton live births their weight for each birth was 1 / 2). We used weighting because prior studies have demonstrated that when cluster size is informative using an unweighted approach would provide biased marginal estimates [ 37 , 38 ]. For the binary outcomes (preterm birth and low birth weight), we specified a binomial distribution and a logit link function. For the continuous outcomes (gestational age and birth weight), we specified a normal distribution and an identify link function. Confounders were assessed using a priori knowledge and a directed acyclic graph (DAG). Confounders included: year of transfer (continuous), embryo stage at transfer (cleavage or blastocyst stage), number of prior donor embryo transfers (continuous), recipient age (continuous), recipient body mass index (BMI, kg/m 2 ; continuous), female recipient infertility factor (present or absent), donor age (continuous), and donor BMI (continuous). Female infertility factor included diminished ovarian reserve, uterine factor, endometriosis, repeated pregnancy loss, tubal factor, or ovulation factor. When data were missing for potential confounders (<6%), we utilized single imputation with the mean value for the continuous variables. We ran several sensitivity analyses to test the robustness of our results. We examined the unadjusted and adjusted estimates restricting the sample to individuals with complete confounder information (n=659). To further account for any differences in embryo quality between SET and DET cycles, we restricted our analysis to singleton live births that had high quality embryos transferred (n=630) and blastocyst stage embryos transferred (n=662). Third, because there is some evidence that suggests vanishing twin syndrome (VTS) can affect birth outcomes, we restricted the sample to pregnancies with only one gestational sac reported on ultrasound (n=653). Finally, because recent evidence suggested a difference between fresh and frozen embryo transfers [ 27 ], we restricted our analysis to only fresh embryo transfers (n=500). All statistical tests were two sided with a significance level of p<0.05. All statistical analysis took place in SAS 9.4 (SAS Institute Inc., Cary, NC, USA). The Emory University Institutional review board (IRB) approved this study (Approval number: IRB00080463). We did not obtain informed consent. The Emory University IRB granted us a HIPPA waiver to pull information from their medical records (without explicit consent) given the low risk. REDCap at Emory University was used for data collection as previously described elsewhere [ 39 ].

From 2008 to 2016, 698 singleton infants were born to 610 women undergoing donor oocyte IVF. The majority (n=500, 71.6%) of transfers were vitrified oocyte thaw with a subsequent fresh embryo transfer (within 7 days), while the rest (n=198, 28.4%) were frozen embryos that were thawed and transferred. The donor average age and BMI were 25.8 years and 22.3 kg/m 2 , respectively ( Table 1 ). The average age and BMI of the recipients was 41.3 years and 24.5 kg/m 2 . A majority of recipients were white (71.6%, n=500) and were diagnosed with female infertility factor (77.8%, n=543). Over 90% of the singleton infants had a single gestational sac detected at the first ultrasound (93.7%, n=653). Approximately 15% of live births were considered preterm (n=105) and 10% were considered low birth weight (n=69). Among the preterm births, 60% were late preterm birth (35 weeks - <37 weeks; n=63) and 40% were early preterm birth (<35 weeks; n=42). Among the low birth weight births, 80% were between 1,500 and 2,500 grams (n=55) while 20% were less than 1,500 grams (n=14). Among the 698 singleton live births, 482 were the result of a SET while 216 were the result of a DET ( Table 1 ). When comparing the sample characteristics by number of embryos transferred, some differences were observed. More Day 3 transfers occurred in the DET group (15.3%, n=33) compared to the SET group (0.6%, n=6). More DETs occurred from 2008 to 2010 (39.5%, n=86) as compared to 2014 to 2016 (13.0%, n=28). Additionally, the DET group had more instances where 2 or more gestational sacs were initially detected on ultrasound (13.4%, n=29) compared to the SET group (0.4%, n=2). Finally, more recipients in the DET group (n=50; 23.2%) had tubal infertility compared to recipients in the SET group (n=69; 14.3%). We observed no association with preterm birth when comparing DET versus SET (aOR: 1.00; 95% CI: 0.61, 1.65) ( Table 2 ). When gestational age was modeled as a continuous variable, on average births from a DET were 0.23 weeks earlier than births from a SET after adjustment for potential confounding variables (Mean: −0.23; 95% CI: −0.63, 0.17). However, this association was not statistically significant. The adjusted odds of having a low birth weight infant were slightly higher, but not significantly so, in singleton births from DET compared to SET (aOR: 1.45; 95% CI: 0.80, 2.60) ( Table 2 ). Similarly, when birth weight was modeled as a continuous variable, the adjusted average birth weight of singleton live births resulting from DET was 27.9 grams lighter than those from SET (Mean: −27.9; 95% CI: −137.7, 81.9) but this difference was not statistically significant. We observed similar results when utilizing a complete case analysis, when restricting the sample to only high-quality and blastocyst embryo transfers, when restricting the sample to single gestational sac pregnancies, and when restricting to fresh embryo transfers ( Supplemental Tables 1 , 2 , 3 , 4 , 5 ).

Since the first United States (US) birth was achieved by assisted reproductive technology (ART) in 1981, the utilization of ART has increased dramatically in the US [ 1 ]. As the usage of ART has increased, several studies have investigated perinatal and infant health when compared to spontaneously conceived births [ 2 – 4 ]. Multiple gestation is a significant risk factor for pregnancy complications, including gestational diabetes, gestational hypertension, preeclampsia, preterm birth, low birth weight, and birth defects [ 5 – 11 ]. For many years, ART procedures were a large contributor to multiple births in the US [ 12 , 13 ], however the percent of ART-conceived live births resulting in a multiple-birth delivery decreased from 29.4% in 2010 to 7.5% in 2019 [ 14 ]. This was largely due to guidance from several organizations, including American Society for Reproductive Medicine [ASRM] and the Society for Assisted Reproductive Technology [SART], to reduce the number of embryos transferred [ 15 ]. Some countries have even strongly suggested or mandated that single embryo transfers (SET) be used whenever clinically possible [ 16 , 17 ]. Despite the decrease in iatrogenic multiple birth, IVF remains a significant risk factor for obstetric complications. Studies in autologous oocyte IVF have noted similar or better perinatal outcomes when compared to live births from eSET as compared to double embryo transfers (DET) or multiple embryo transfers (MET) [ 18 – 23 ]. Although the main driver of improved outcomes in eSET is decrease in multiple births, several studies have also investigated perinatal outcomes in singleton births following eSET compared to DET or MET [ 24 – 27 ]. While findings concerning gestational age were inconsistent, all four studies observed a decline in mean birth weight (or higher odds of low birthweight) as the number of transferred embryos increased [ 24 – 27 ]. There may be other risks of DET/MET beyond multiple births that contribute to adverse obstetric outcomes. The biological mechanism for this association is not fully understood but may be due to early competition between transferred embryos or changes in the placental environment [ 28 , 29 ]. While these studies provide compelling evidence, all were autologous oocyte cycles, where the number of embryos transferred strongly correlates with the patient’s fertility potential. Using a donor oocyte model can decrease this bias since the number of embryos being transferred does not as strongly depend on the recipient’s infertility diagnosis. To this end, we examined the association between number of embryos transferred and perinatal outcomes (gestational age and birth weight) exclusively among singleton births resulting from nonidentified vitrified donor oocyte IVF.

We observed no evidence for an association between DET and adverse perinatal outcomes when the result is a singleton birth from vitrified donor oocyte IVF, in this single clinic study. While DET did not significantly increase risk of preterm birth or low birthweight in singletons, clinicians should continue to transfer the minimal number of embryos possible, especially in donor oocyte cycles, to prevent adverse maternal and perinatal outcomes associated with multiple births, which can result from DET.

Among 698 singleton births resulting from SET or DET among 610 vitrified donor oocyte recipients, we observed no association between the number of embryos transferred with gestational age or birth weight. Moreover, the results were robust to several sensitivity analysis. Our findings do not support the hypothesis that early competition between transferred embryos or changes in the placental environment as a result of DET may adversely impact perinatal outcomes of singleton pregnancies. Much of the previous literature in autologous oocyte IVF cycles has found an association between greater number of embryos transferred and lower singleton birth weights. Among a sample of 1647 singleton pregnancies resulting from autologous ART in China, Wu et al observed that mean birth weight was about 163 grams higher in the SET group compared to the DET group [ 24 ]. Similarly, two other studies utilizing autologous IVF cohorts, one from Belgium and one from the US, found that the mean birth weight in the SET group was about 120 grams and 181 grams higher, respectively, as compared to the DET group [ 25 , 26 ]. In contrast, a study of 35,836 singleton ART pregnancies from Sweden, observed no overall association between DET and risk of low birthweight [ 27 ]. However, after stratifying based on fresh versus frozen cycles, there was a higher risk of low birth weight after DET in frozen cycles but lower risk of low birth weight after DET in fresh cycles [ 27 ]. Among our cohort of vitrified donor oocyte cycles we observed limited evidence for a difference in the mean birth weight in singletons between SET and DET. Although our point estimates were in a similar direction to the majority of previous research, the magnitude was much smaller (~28 gram difference) and effect estimates were not statistically significant. Several reasons could account for these differences. The most likely reason is that all four of these cited papers’ analyses utilized fresh [ 24 , 25 ] or frozen [ 26 ] autologous IVF cycles [ 29 ], which has a greater potential for residual confounding by recipient infertility diagnosis. When embryos are transferred in autologous cycles, the oocytes come directly from the individual who is experiencing infertility, so we observe a combined effect of the number of embryos transferred and the recipient’s infertility diagnosis, which does not occur in donor oocyte cycles. It is also possible that autologous and donor oocyte cycles may have different endometrium receptivity, which could account for some of the differences in birth weight we observed. Donor oocyte cycles may also have worse maternal and perinatal outcomes, including higher rates of preterm birth and low birthweight than autologous IVF cycles [ 40 , 41 ], which could contribute to differences we observed. Next, each of the studies utilized slightly different methodologies to analyze their data. Wu et al, due to sample size constraints, utilized unadjusted bivariate analysis [ 24 ] while De Sutter et al utilized adjusted linear regression [ 25 ] and Rodriguez-Wallberg et al utilized adjusted logistic regression [ 27 ]. In all three of these methods, the authors did not control for potential correlation between births from the same woman. Our methodology is much closer to Devine et al, who utilized GEE models to account for repeated subjects [ 26 ]. The GEE models used in the present study however, also included a weighting factor which has previously been shown to be less biased when the cluster size is informative [ 37 , 38 ]. Next, it is possible that the number of initial gestational sacs and vanishing twin syndrome could have played a role in the slightly lower birth weight we observed among singletons resulting from DET; however, this is unlikely. For example, when the studies by Devine et al and Wu et al restricted their samples to only pregnancies with a single gestational sac, they still observed similar associations [ 24 , 26 ]. Our results for birth weight were also very similar after excluding pregnancies with vanishing twin syndrome. Finally, our sample was predominately included fresh embryo transfers which could have played a role in our null findings. A recent study observed an adverse effect of MET on birthweight and gestational only among frozen embryo transfers [ 27 ]. Therefore, it is possible that we did not observe an adverse effect because our sample included mostly fresh transfers rather than frozen transfers. Our results finding no association between the number of embryos transferred and gestational age is consistent with most of the previous research. Among the four studies investigating singleton births resulting from autologous IVF, only two found an association between the number of embryos transferred and higher odds of preterm birth. De Sutter et al, utilized data collected from 404 SET and 431 DET live births in Belgium from 2000 to 2004 [ 25 ]. They observed that the odds of preterm birth in the DET group was higher than the odds in the SET group (OR: 1.77, 95% CI: 1.06, 2.94) [ 25 ]. Rodriguez-Wallberg et al, observed among 11,854 frozen autologous cycles, the odds of very preterm birth was higher after DET than SET (OR: 1.90; 95% CI: 1.11, 3.27) [ 27 ]. They did not observe a significant association for fresh cycles or for moderately preterm among frozen cycles [ 27 ]. While Wu et al did observe an increased incidence of preterm birth with an increasing number of embryos transferred, this was not statistically significant [ 24 ]. Devine did not observe an association between the number of embryos transferred and gestational age [ 26 ]. The discordance in results could be due to timing of the studies which likely correlates with improvements in ART. For example, earlier studies included ART cycles that could have selected embryos for transfer based on now outdated criteria. Another potential reason for the discordance could be simply sample size or chance. While our results do not suggest any deviations in current clinical practices, it was important to undertake this study given that DET/MET is still relatively common among donor cycles [ 42 , 43 ], even though it is not recommended by current clinical guidelines in many countries, and much of the previous literature on this topic is based on autologous cycles. Additionally, donor cycles are an excellent model for examining additional perinatal and obstetric risks that are more prevalent in IVF pregnancies compared to spontaneous conceived pregnancies. SET has been shown to reduce the number of multiple births which, in turn, can reduce maternal and perinatal morbidity [ 16 , 17 , 19 , 21 , 22 , 26 ]. Our results, along with others, indicate that eSET is the standard of care for patients as the goal of treatment is a live singleton birth with normal birth weight (≥2500 grams) and normal gestation (≥37 weeks). Current guidelines of transferring the least number of embryos as clinically indicated will likely result in reduced maternal and perinatal morbidity and should be continued, especially as use of DET and MET remain high among donor oocyte cycles [ 42 , 43 ]. Our study does have several limitations. The sample size may have limited our ability to detect statistically significant associations, particularly for the outcomes of preterm birth and low birthweight; however, it is unclear if it would be clinically significant. Additionally, while guidance around and practice of DET have changed since our study was conducted, DET was still quite prevalent in donor cycles during the years of study, which provided a more powerful model for studying this pathophysiology [ 42 , 43 ]. Second, because this is an observational study and women were not randomly assigned to having one or two embryos transferred, there still may be factors that differ between these two groups, which could result in residual confounding. However, as noted above, by focusing on a cohort of women using vitrified donor oocytes, we accounted for the most critical confounder (underlying fertility potential) by design and we also adjusted for several other donor and recipient characteristics to further account for confounding. Third, it is possible that the quality of the embryos transferred could increase the odds of preterm birth or low birth weight and, since more low-quality embryos were transferred in DETs, our results could have been biased. However, when we restricted our analysis to transfers with only high-quality embryos, we observed similar estimates which may indicate a low chance of bias from this source. It is also possible that transferring more than one embryo could have initially resulted in a multiple pregnancy (e.g. more than one gestational sac visualized on ultrasound) but only one infant survived until birth. The increased number of gestation sacs could have had a negative impact on the placental environment [ 29 ] and/or the surviving fetus [ 28 ]. However, when we restricted our analysis to cycles without evidence of vanishing twin syndrome, we observed similar results. Finally, because the women in our study are all undergoing donor oocyte IVF, they tended to be older and more likely to be diagnosed with diminished ovarian reserve. Therefore, these results may not be generalizable to a wider IVF population. However, our sample does appear to be similar to the wider donor oocyte population [ 42 ]. Our analysis does have several strengths. First, our study population and design helped us eliminate many of the biases and confounding factors that were present in prior studies. By collecting detailed information on not only the recipients but also the donors and oocyte retrievals, we were able to account for multiple factors that may impact birthweight and gestational age. Because of this, we were able to consider multiple potential confounding variables in our analysis that previous studies could not. All of our cycles also came from a single vitrified donor oocyte bank, which allowed us to control for donor and recipient clinical protocols. Since all of the transferred embryos were coming from nonidentified donors with generally high-quality oocytes, regardless of the recipient’s infertility diagnosis, there should not be a relationship between embryo quality and prognosis of the recipient. Also, in donor cycles, we are able to eliminate the effect of varying stimulation protocols on endometrial receptivity that is inherent to autologous cycles. Finally, we conducted several sensitivity analyses to check the robustness of our findings and observed similar associations.