Maternal Vascular Health in Pregnancy and Postpartum After Assisted Reproduction

All data that support the findings of this study are available within the article and its online data supplement . Patients with viable, sonographically confirmed, singleton intrauterine pregnancies from autologous oocytes were recruited at 6-8 weeks’ gestation at the Stanford University School of Medicine, Division of Reproductive Endocrinology and Infertility from June 2015 until March 2018 after providing written informed consent for the collection of demographic and clinical data, as well as participation in physiological assessments and blood draw. The longitudinal cohort study approved by the Institutional Review Board (IRB) at Stanford University, adhered to the principles of the Declaration of Helsinki and Title 45, U.S. Code of Federal Regulations and follows procedures in accordance with institutional guidelines. Subjects were studied 2 times during pregnancy (11-14 weeks and 35-37 weeks’ gestation) and one time postpartum (6-12 weeks). Exclusion criteria were age >50 years, BMI >30 kg/m 2 , latex allergy, known vascular disease (i.e., chronic hypertension, lupus erythematosus, rheumatoid disease, etc.), steroid or heparin intake, donor egg transfer, or gestational carrier pregnancies. Three groups with different CL status at conception were included in this study: 1) 0 CL from FET in a programmed cycle; (2) 1 CL (FET in a natural cycle or from spontaneous conception after infertility); or (3) > 3 CL associated with IVF and fresh embryo transfer. An overview about enrollment numbers and study participation at each time point can be found in Figure S1 . Following an overnight fast and avoidance of caffeinated products, alcohol and pain medications during the preceding 24 hours, the participant’s weight and height were measured on testing day (Scale-Tronix scale, White Plains, NY; Seca stadiometer, Columbia, MD). 15 After at least five minutes of rest four sets of resting blood pressures were taken on the testing arm in a sitting position using the oscillometric method (Connex Vital Signs Monitor, Welch Allyn, Beaverton, OR). The average of the last three blood pressure measurements was used as per laboratory protocol. Mean arterial pressure (MAP) was calculated using the following formula: (MAP= [systolic blood pressure + 2x diastolic blood pressure] / 3). The preconception blood pressures and body-mass index (BMI) were measured before the start of fertility treatment at a clinical visit and obtained from the medical record. Vascular endothelial function was evaluated non-invasively using the EndoPAT 2000 device (Itamar Medical, Caesarea, Israel) in accordance with the manufacturer’s recommendations and in line with the published literature. 15 , 23 - 26 Outcome variables included 1) peripheral endothelial function using the reactive hyperemia index (RHI, arbitrary units) to assess endothelial function by reactive hyperemia, 2) augmentation index (AI) normalized to heart rate of 75 (AI75) to reflect arterial stiffness and 3) baseline pulse wave amplitude (BPWA) to reflect arterial tone. The delta of outcome variables between time points were calculated as percent change ([time point A value– time point B value])/time point A value x100). Plasma fasting levels of total cholesterol, triglycerides, LDL, HDL, and non-HDL cholesterol were measured using automated methods in the clinical laboratory at Stanford Medical Center at time of blood draw. The primary outcome of interest was impact of CL on maternal peripheral endothelial function measured as RHI over time and involved the comparison by CL status regardless of method of conception. Secondary outcomes of interested included change in MAP, AI75 and BPWA. The 1 CL group served as the control and 0 CL and > 3 CL as exposure groups. To disregard the impact of embryo treatment in culture and cryopreservation, secondary analyses included the comparisons of conception methods, e.g. natural cycle FET (CL=1) vs. programmed cycle FET (CL=0) or spontaneous conceptions after a period of infertility vs. natural cycle FET (both CL=1). A sample size calculation using published data on differences in RHI (α=0.05, Power=0.8, effect size 10%) revealed a goal of 12-15 participants to finish all 3 visits per group and an initial enrollment target was set for a total of 50 participants assuming a modest withdrawal/drop-out rate. 25 , 27 Given a higher than expected withdrawal and drop-out rate over time (60%) we increased enrollment to greater than 80 subjects. Study data were managed using REDCap (Research Electronic Data Capture). 28 Normality of the data distribution was evaluated by the Shapiro–Wilk test with p<0.05 indicating no normal distribution. Continuous data at each visit are summarized by mean and standard deviation (normality distribution) or median and interquartile range (non-normality distribution). Categorical variables are expressed as frequencies and percentages. Group comparisons were performed using t-test to compare means or the Mann-Whitney test to compare ranks. The reference group of the pairwise comparisons was either the 1 CL or the natural cycle FET group. For the major endpoints, SBP, DBP, MAP, RHI, AI75 and BPWA, linear mixed models were used to model the trajectories of the respective data at pre-conception (SBP, DBP, MAP only) in first trimester, third trimester and postpartum. Major endpoints were analyzed using a restricted maximum likelihood (REML)-based repeated measures approach in combination with the Newton-Raphson algorithm. Analyses included the fixed categorical effects of CL group/method of conception, time and CL group/method of conception-by-time interaction. An unstructured (co)variance) structure was used to model the within-subject errors and implicitly handle missing data. Significance tests were based on Type III F-test using a two-sided α =0.05. For all analyses, the statistical software JMP Pro 14 software (SAS Institute Inc., Cary, NC, USA), Prism 7 (GraphPad Software Inc., San Diego, CA, USA) or SAS (SAS Institute Inc., Cary, NC, USA) were used.

Participant age, ethnicity, parity, gravidity and maternal smoking were comparable among the three cohorts who conceived with 0, 1 or > 3 CL ( Table 1 ). There were more Asian women in the 0 CL and more White women in the > 3CL group when compared to each other. There was a higher rate of aspirin intake until 10 weeks’ gestation in the 0 and > 3 CL groups. Diminished ovarian reserve was more frequent in the > 3 CL group while male factor infertility was more often an infertility diagnosis in the 0 CL and > 3 CL cohorts compared to the 1 CL group. When comparing the demographic data by method of conception, more Asian women received a programmed FET and more White women conceived spontaneously. Aspirin use and male factor infertility were more frequent in both FET protocols compared to spontaneous conceptions. Diminished ovarian reserve and male factor infertility were more often reported as reason for infertility in women who received a programmed FET compared to women with a natural cycle FET. Maternal BMI was comparable among the 0 CL, 1 CL and > 3 CL cohorts at first trimester and postpartum. Maternal BMI was 8.3% lower in third trimester of pregnancy in 0 CL conceptions compared to 1 CL. ( Table 2 ). Blood pressure parameters were analyzed across the cohorts at the 3 different times points. Using type III F-Test for the fixed effects in the mixed model, the CL group by time interaction for diastolic blood pressure (DBP), systolic blood pressure (SBP) and MAP was not significant, but there was a significant effect of time on DBP (p=0.03) ( Table S1 ). Pre-conception blood pressure levels were not different among CL groups. 15 While we have shown before that women without a CL lacked the normal physiologic drop in blood pressure in the first trimester 15 compared to 1 CL and > 3 CL conceptions, blood pressure levels in the 0 CL group remained highest throughout pregnancy and postpartum ( Figure 1A , Table 2 ). There was a significant effect of method of conception and time for SBP (p=0.01), DBP (p=0.045) and MAP (p=0.02) when comparing results of type III F-Test ( Table S1 ). While we have reported differences in pre-conception and first trimester blood pressure before 15 diastolic blood pressure of women with conceptions in a natural cycle FET rose more between first and third trimester of pregnancy (7.5% vs. 0.8%; p=0.03), but there was a greater decline of diastolic blood pressure from third trimester to postpartum compared to spontaneous conceptions (−6.3% vs. 2.9%; p=0.005), ( Figure 2A , Table 2 ). Endothelial function parameters are presented in Figures 1 and 2 , Table 3 and Table S1 . Using type III F-test, the group by time interaction for RHI, a measure of endothelial function, was not significant (p=0.09) for CL number comparisons. However, there was a significant effect of time (RHI; p=0.0002), ( Table S1 ). As reported before RHI was different during the first trimester 15 and dropped by 22.2% between first and third trimester in 1 CL conceptions (p=0.04). RHI was not different between CL groups in late pregnancy and postpartum. In general, in the 1 CL and > 3 CL groups RHI declined during pregnancy (−22.2% and −10.5%) and there was an overall rise in RHI postpartum (13.6% and 27.9%) compared to late pregnancy while RHI has hardly changed in the 0 CL group in late pregnancy (−5.5%) and postpartum (−1.9%) ( Figure 1B , Table 3 ). AI75 was 4.7-fold higher in the third trimester of pregnancy in 0 CL conceptions compared to 1 CL conceptions (p=0.03, Figure 1C ) but not different in the mixed model. There was no significant group by time interaction for BPWA in the mixed model (p=0.91). However, there was an effect of time (p<0.0001) and CL number (p=0.002) ( Table S1 ). BPWA rose from first to third trimester and fell after pregnancy within all groups (p 3CL compared to 1 CL was 209.5% higher in first trimester (p=0.009) 15 , 151.2% higher in third trimester (1 CL vs. > 3CL: p=0.006) and 152.0% higher postpartum (p=0.03) ( Figure 2D , Table 3 ). For the comparisons of conception method there was a significant effect of time in the mixed model for RHI (p=0.0003) and BPWA (p<0.0001) ( Table S1 ). While differences in RHI in first trimester have been shown before 15 , RHI dropped significantly between first and third trimester especially in natural cycle FET conceptions (−22.7%; p=0.006). Postpartum, RHI in women with spontaneous conceptions reached 20.7% higher levels compared to FETs in a programmed cycle (p=0.05) ( Figure 2B ). AI75 was 6.4-fold higher in late pregnancy (p=0.02) in programmed cycle FETs compared to natural cycle FETs ( Figure 2C ). BPWA rose between first and third trimester and dropped postpartum and was not different at all time points between the method of conception groups ( Figure 2D , Table 3 ). As expected, total cholesterol and triglyceride concentrations rose during pregnancy and dropped after delivery in all groups ( Table S2 ). HDL concentrations were lowest in conceptions with > 3 CL at all time points. Aside from higher total cholesterol levels in first trimester in programmed FETs compared to FETs in a natural cycle or spontaneous conceptions, 15 lipid concentrations were comparable among conception groups and time points (data not shown). Although the study was not powered for adverse maternal and neonatal outcomes, preeclampsia incidence was significantly higher in conceptions in the absence of a CL compared to the presence of 1 or >3 CL. Further, preeclampsia with severe features only occurred in 0 CL conceptions ( Table 4 ). The birth weight of singleton infant was significantly less and the number of infants with low birth weight ( 3 CL compared to the groups with 0 CL and 1 CL. There were no differences in the number of very low birth weight infants (<1500g), incidence of preterm birth, admission rate to neonatal intensive care unit (NICU), APGAR at 1 and 5 minutes, or infant sex. When comparing FET there was a higher rate for preeclampsia in conceptions occurring after a programmed cycle FET compared to a FET in a natural cycle.

In this study, we found impaired vascular health parameters in conceptions with non-physiologic CL numbers. These were most dramatic in the first trimester which is the time of greatest CL function, however, later in pregnancy, patients developed stiffer arteries. Despite the lack of persistent differences in endothelial function in later pregnancy, preeclampsia incidence was significantly higher in the group with absence of a CL during early pregnancy, while pregnancies with multiple CL had infants with lower birth weights, suggesting that placental function might be altered in both these types of pregnancies.

In this longitudinal follow-on study, impaired maternal vascular health predominated in the first trimester in women who conceived in the absence of a CL. This is reflected by an elevated MAP and impaired endothelial responsiveness while in late pregnancy the augmentation index was higher compared to conceptions with 1 CL or > 3 CL. BPWA was higher at all time points in conceptions with supraphysiologic CL numbers. We previously reported the impaired blood pressure adaptation and reduced endothelial function (RHI) during early pregnancy in our 0 CL cohort. 15 As expected, this finding was corroborated herein, but importantly, these deficiencies recovered by the third trimester and postpartum. However, in the third trimester AI75, a measure of arterial stiffness, was higher in 0 CL compared to 1 CL conception as well as for the comparison of FET in a programmed (CL=0) and natural cycle (CL=1). An association between increased arterial stiffness (reduced arterial compliance) and preeclampsia was previously reported 29 - 31 and might contribute to the observed increased rate of preeclampsia in our cohort. Interestingly, BPWA, a measure of vascular tone, was significantly higher from early pregnancy to several weeks postpartum in conceptions from fresh IVF. These conceptions start in the presence of multiple CL. Supraphysiologic levels of vasoactive CL hormones, e.g. the vasodilator relaxin, might contribute to the enhanced vasodilation that wasn’t associated with adverse maternal outcomes in our cohort although we were not powered to look at adverse outcomes. 15 In normal human pregnancy maternal systemic vascular resistance decreases and arterial compliance increases reaching a nadir and peak, respectively, by the end of the first or beginning of the second trimester. 32 Other vascular measures, e.g. a decrease in augmentation index, correspond with these adaptations. 33 These major physiological adaptations in normal human pregnancy are preceded by a peak of CL product concentrations in the maternal circulation. 17 In first trimester, the CL is a major source of vasoactive hormones such as relaxin, vascular endothelial growth factor (VEGF), and other angiogenic products. 20 , 34 , 35 Deficient circulating relaxin is one biologically plausible mediator for any effect of absent CL 36 as relaxin emanates solely from the CL during human pregnancy 18 , 19 and is a potent vasodilator 37 which mediates circulatory changes and increases arterial compliance in the gravid rat model. 38 , 39 Other vasoactive products of the CL, such as VEGF, are also not replaced during a programmed cycle FET, and thus, absent CL VEGF could theoretically mediate the impaired circulatory adaptations in women conceiving without a CL. An insufficient vascular adaptation during early pregnancy in 0 CL conceptions was evident from an attenuated decline in central (aortic) pulse-wave velocity and rise in pulse-wave transit time compared to pregnancies occurring in the presence of a CL or multiple CL. 14 Consistent with our findings of impaired blood pressure adaptation and endothelial function, these vascular deficits in the 0 CL cohort also recovered after the first trimester. 14 In conceptions with multiple CL, e.g. in fresh IVF transfers after controlled ovarian hyperstimulation, supraphysiologic concentrations of vasoactive CL products could have contributed to the higher BPWA. One might speculate that the absence of CL products rather than an excess contributes to higher incidence of preeclampsia an preeclampsia with severe features as described herein, and as we previously reported. 14 Although our study was not powered to detect differences in maternal and neonatal outcomes, some observations deserve further discussion. Specifically, the higher incidence of preeclampsia in conceptions in the absence of a CL compared to the presence of 1 CL or > 3 CL or in programmed compared to natural cycle FETs deserves mentioning, particularly given the CL=0 group had lower BMI, which is otherwise associated with a lower preeclampsia risk. 40 An increased risk for the development of preeclampsia is well-known for many years in pregnancies after ART compared to spontaneous conceptions of the general population. 4 , 41 , 42 More recent data from observational studies report a higher preeclampsia incidence for FET compared to fresh IVF transfers. 4 , 6 - 13 A recent population based study from Sweden 13 and our data from a prospective cohort of 683 deliveries after infertility 14 suggest that absence of CL in programmed FET cycles may be the major contributor to the increased preeclampsia risk reported in FET. Aspirin is recommended for women at high risk for preeclampsia starting at 12 gestation to reduce the incidence of the disease. 43 The medication is also used during fertility treatment until 10 weeks’ gestation to improve implantation and pregnancy rate. Aspirin was discontinued per protocol in our cohort at 10 weeks which is prior to the first study visit. Although women with aberrant CL number were more likely to take aspirin, which would potential decrease preeclampsia risk, the preeclampsia rate was higher in this group lacking a CL. Therefore, it seems unlikely that short term aspirin use in the first weeks of gestation impacts preeclampsia incidence in the ART population more than CL status. In this study infants from pregnancies with > 3 CL after fresh embryo transfer had on average lower birth weights and a higher rate of low birth weight infants (< 2.500g) when compared to conceptions in the absence of a CL or in the presence of 1 CL. Data from our small cohort, therefore, are consistent with the increased risk for low birth weight in fresh IVF transfer as abundantly documented in the literature. 2 , 44 - 46 The number of retrieved oocytes appears to be negatively correlated with birth weight. 47 Several recent studies reported increased birth weights and a higher risk for large for gestational age infants after FET compared to fresh embryo transfers. 48 - 50 Fetal growth might already be impacted early in pregnancy as a study suggests smaller crown-rump-length at 6 and 8 weeks’ gestation after fresh IVF transfers compared to FETs. 51 If these growth differences are already present as early as 6 to 8 weeks as it has been shown before it is likely that peri-conceptional effects can account for the differences in birth weight. 51 In our cohort, total cholesterol and triglyceride levels rose as expected in all groups from first trimester to late pregnancy and dropped postpartum. The finding of higher total cholesterol concentrations in first trimester in women lacking a CL compared to 1 CL or > 3 CL deserves further discussion. During the first weeks of pregnancy when most organs are formed, the fetus depends largely on maternal cholesterol. 52 , 53 Conditions affecting maternal cholesterol levels may have an adverse effect on the development and growth of the fetus. Data on the effect of CL or method of conception on lipid parameters are lacking. One recent observational study reported different metabolomic profiles between spontaneous and IVF conceptions, including higher levels of acylcholine derivates, phosphatidylinositols, phosphatidylethanolamines and diacylglyerols in late first trimester. 54 The authors didn’t discriminate in their results between fresh embryo transfers and FETs, nor FET protocol used. Thus, we can only speculate at this point that in the published cohorts the higher birth weights and a higher rate of large for gestational age infants after FET might be attributable to higher nutrient availability in early gestation in programmed FET cycles with absence of a CL. Maternal nutrient concentrations and placental nutrient production and transport impact fetal weight and infant outcomes. Pregnancy leads to substantial changes in maternal lipid profiles compared to the non-pregnant state. Serum lipid and lipoprotein levels gradually increase from 12 weeks’ gestation, especially in the second and third trimester with greatest elevation in triglycerides. 55 , 56 These changes reflect a physiologic adaptation of the mother including hepatic enzyme activities and placental steroid synthesis to satisfy the energy demand of mothers and the fetus. While it is known from several observational studies that elevated lipid levels during late pregnancy are associated with complications and adverse outcome for both mother 57 and newborn 58 , a disturbed lipid profile during early pregnancy has been shown to have similar negative associations. Specifically, an increase in triglycerides was linearly associated with an increased risk of preeclampsia (OR 1.69, P= 0.018) 59 and a higher rate of large for gestational age infants. 60 How fertility treatment and method of conception affect maternal lipid profiles remains largely unknown. There are several strength and limitations to our study that deserve mention. This is the first study to report longitudinal RHI data from conceptions after infertility in pregnancy and postpartum. An advantage of the study is the inclusion of women with different CL numbers and methods of conception along with broad information about demographic, clinical and outcome data as well as the follow-up throughout pregnancy and postpartum. Although every effort was made to achieve a reasonable sample size at all time points, the nature of the study and the study population left us with decreasing participant numbers over time. Due to the small sample size true effects might have been masked or overvalued and a correction of data for potential confounders was not possible. Also, pre-conception data were only available for BMI and blood pressure, but not endothelial function or lipid concentrations. Therefore, some women might have already been different before conception.

With the increasing utilization of FET and a reported higher risk for the development of preeclampsia, further studies exploring the impact of different FET protocols on pregnancy physiology as well as randomized controlled trials for FET protocols to investigate maternal and offspring outcomes are needed to provide clinical guidance to maximize maternal and child health following IVF conception.

With the increasing utilization of assisted reproductive technology (ART), which has resulted in the birth of over 8 million babies worldwide, our understanding of the potential adverse risks is steadily expanding. 1 Many of the reported adverse maternal and perinatal outcomes are associated with placental defects, including small for gestational age infants (SGA), preeclampsia, placental abruption, placenta accreta and previa which likely have their origin in early gestation. 2 Recent data suggest an increased rate of hypertensive disorders of pregnancy especially in donor egg recipients 3 , 4 , 5 and after frozen-thawed embryo transfer (FET) cycles. 4 , 6 - 13 The underlying mechanisms are not yet fully understood but recent studies suggest a contribution of the treatment protocol to maternal adaptation to pregnancy and preeclampsia risk. 13 - 16 Pregnancies after spontaneous conceptions and fertility treatments performed in a women’s natural cycle start in a physiologic endocrine environment as there is one corpus luteum (CL). The CL is the major source of reproductive hormones, e.g. estrogen, progesterone, relaxin, and others in early pregnancy. 17 - 22 Certain protocols utilized in ART affect the CL number and disrupt this maternal hormonal milieu. On the one hand, FET cycles performed in a programmed setting lack a CL due to suppression of the hypothalamic-pituitary axis by estradiol intake. On the other hand, controlled ovarian stimulation in in vitro fertilization (IVF) cycles with fresh embryo transfer leads to the development of multiple CL. These two scenarios are associated with extremes of CL hormone concentrations in early pregnancy – absent or very low versus supraphysiologic levels (the latter in many, but not all women). While estrogen and progesterone are usually replaced as part of the treatment and are ultimately secreted by the developing placenta, other potentially important CL hormones are not. In a prospective cohort we recently showed an increased risk for the development of preeclampsia and preeclampsia with severe features in women who conceived in the absence of a CL in programmed FET cycles compared to women with one or multiple CL. 14 In an ancillary study, women with no CL lacked the physiologic drop in mean arterial pressure (MAP) in first trimester and also showed impaired vascular function, including reduced endothelial responsiveness and increased arterial stiffness. 15 Angiogenic and non-angiogenic circulating endothelial progenitor cells were significantly reduced, suggesting an impact on endothelial homeostasis and repair capacity. 15 In this study we present longitudinal follow-on data to clarify whether CL number and method of conception impact various aspects of maternal vascular health throughout pregnancy and postpartum.