Increased Preeclampsia Risk and Reduced Aortic Compliance With In Vitro Fertilization Cycles in the Absence of a Corpus Luteum

Abstract

In vitro fertilization involving frozen embryo transfer (FET) and donor oocytes increases preeclampsia risk. These IVF protocols typically yield pregnancies without a corpus luteum (CL), which secretes vasoactive hormones. We investigated whether IVF pregnancies without a CL disrupt maternal circulatory adaptations and increase preeclampsia risk. Women with 0 (n=26), 1 (n=23), or >1 (n=22) CL were serially evaluated before, during and after pregnancy. Because increasing arterial compliance is a major physiological adaptation in pregnancy, we assessed carotid-femoral pulse wave velocity (cfPWV) and transit time (cfPWTT). In a parallel, prospective cohort study, obstetric outcomes for singleton livebirths achieved with autologous oocytes were compared between groups by CL number (n=683). The expected decline in cfPWV and rise in cfPWTT during the first trimester were attenuated in the 0 CL compared to combined single/multiple CL cohorts, which were similar (group-time interaction: p=0.06 and 0.03, respectively). The blunted changes of cfPWV and cfPWTT from pre-pregnancy in the 0 CL cohort were most striking at 10–12 weeks gestation (p=0.01 and 0.006, respectively, versus 1 and >1 CL). 0 CL was predictive of preeclampsia (adjusted odds ratio 2.73; 95%CI 1.14–6.49) and preeclampsia with severe features (6.45; 95%CI 1.94–25.09) compared to 1 CL. Programmed FET cycles (0 CL) were associated with higher rates of preeclampsia (12.8% vs 3.9%, p=0.02) and preeclampsia with severe features (9.6% vs 0.8%, p=0.002) compared with modified natural FET cycles (1 CL). In common IVF protocols, absence of the CL perturbed the maternal circulation in early pregnancy, and increased the incidence of preeclampsia.

INTRODUCTION

Recent studies demonstrated increased risk for hypertensive disorders of pregnancy following in vitro fertilization (IVF), particularly for pregnancies occurring in donor egg recipient^1,2,3 and frozen-thawed embryo transfer (FET) cycles^{2, 4–10}. Mechanisms underlying this increased risk are not clearly understood¹¹. Their elucidation could inform specific modifications of IVF protocols to reduce preeclampsia risk in these patients.

We hypothesized that the increased preeclampsia risk in IVF-conceived pregnancies may be partly attributable to the degree by which the IVF protocol impacts the maternal hormonal environment in the first trimester when the corpus luteum (CL) is a major source of reproductive hormones¹². While unassisted (“spontaneous”) pregnancies usually develop in the presence of one CL, IVF involves two extremes—either formation of supra-physiologic numbers of CLs associated with ovarian stimulation in fresh IVF cycles or hypothalamic-pituitary suppression and absence of the CL in artificial, programmed cycles routinely used for donor egg recipients and FET. Although estradiol and progesterone are replaced during a programmed FET or donor egg cycle in the first trimester, other vasoactive products of the CL which may be important for maternal cardiovascular adaptation to pregnancy, such as relaxin^13–16, are not administered. Deficient circulatory adaptations during early gestation have been linked to adverse pregnancy outcomes including preeclampsia^{12, 17, 18}.

Given the increased risk of preeclampsia in IVF, particularly with donor egg and FET cycles, we hypothesized that absence of a CL would be associated with impairment of maternal circulatory adaptations to pregnancy, particularly in the first trimester, and with a higher risk of hypertensive disorders of pregnancy^{12, 13}. Accordingly, our first objective was to investigate whether abnormal CL number adversely affects the maternal circulation, specifically central and peripheral arterial compliances, thus providing a potential link between aberrant CL number and pregnancy outcome. Typically, arterial compliance increases during normal pregnancy as reflected by decreases in pulse wave velocity (see Table S1 in the online-only Data Supplement for literature review). Moreover, an association between perturbed arterial compliance and preeclampsia was previously reported ^{19, 20}. Our second objective was to test whether abnormal CL number is associated with increased preeclampsia risk.

METHODS

The authors declare that all supporting data are available within the article and its online-only Data Supplement (please see http://hyper.ahajournals.org).

Longitudinal Study of Pulse Wave Velocities

After written informed consent, subjects were recruited to participate in this study approved by the IRB at the University of Florida. Patients planning to conceive with IVF were identified by Reproductive Endocrinology and Infertility physicians at the University of Florida and referred to the Study Coordinator. Subjects planning to become pregnant without assisted reproduction were recruited through advertisement. The Study Coordinator administered informed consent and enrolled those who fulfilled the inclusion and exclusion criteria. The 3 cohorts consisted of pregnant women conceiving (i) “spontaneously” without IVF (singleton pregnancies/single CL); (ii) by transfer of a frozen embryo(s) produced using donor or autologous eggs and in vitro fertilization (IVF), or fresh donor embryo transfer (absent CL); or (iii) after ovarian stimulation, IVF and fresh embryo transfer (multiple CL). Of the 17 women conceiving with embryos derived from donor eggs, 14 were transferred as fresh embryos.

Baseline measurements before pregnancy were made in the absence of circulating CL factors (follicular phase or last 4 days of leuprolide suppression). All subjects were then studied 6 times during pregnancy: 5.8±0.7, 8.3±0.8, 11.6±0.9, 15.2±0.9, 24.0±0.8, and 33.6±1.0 weeks of gestation, and finally 46.0±12.5 weeks after delivery (mean ± standard deviation (SD)). Following an overnight fast and avoidance of caffeinated products, alcohol and pain medications during the preceding 24 hours, subjects reported to the Clinical Research Center at approximately 8:00 AM where they underwent a number of physiological assessments performed over a period of ~5 hr including pulse wave velocity (PWV) and transit time (PWTT).

Standard IVF protocols were used for the fresh IVF cycles. Identical programmed cycle protocols were used for donor oocyte recipient fresh transfer, donor oocyte recipient FET, and autologous oocyte FET. The number of CL in the fresh IVF cycles was estimated by the number of eggs retrieved. In the programmed cycles, absent CL was verified in each woman by undetectable concentrations of circulating relaxin at all time points during pregnancy (data not shown). For IVF protocol details, see Methods in the online-only Data Supplement.

Applanation tonometry (SphygmoCor CvMS; AtCor) was utilized to assess carotid, femoral and radial pressure waveforms non-invasively. Carotid-femoral pulse wave velocity (cfPWV), carotid-radial PWV (crPVW) and corresponding pulse wave transit times (cf- and crPWTT) were calculated using standard methods after at least 10 min of supine rest and within 5 min of peripheral blood pressure and heart rate determination (Datascope). PWTT was also analyzed since unlike PWV, the PWTT does not depend on the distance measurements, which can be somewhat variable. For procedure details, see Methods in the online-only Data Supplement.

Data at each visit were summarized by mean, standard deviation and range, or 1.96 X standard error (95% confidence interval). For the major endpoints, PWV and PWTT, linear mixed models were used to model the trajectories of the PWV and PWTT before and during pregnancy with an account of the correlation between observations over time and an accommodation for missing data. The nonparametric cubic spline was used to characterize and summarize the trajectories. The effects of group (without CL, one CL, or multiple CL), time and group-time interaction were evaluated. A p-value of <0.05 was considered to be significant. For all analyses, the statistical software SAS 9.4 and R 3.4.3 were used. Further details, as well as the statistical analyses for categorical and continuous covariates, and for data presented in Table 1 are available in Methods in the online-only Data Supplement.

Table 1.

Baseline Demographic and Clinical Characteristics^*

	0 CL	1 CL	> 1 CL (Fresh IVF)		p-value
Longitudinal study of pulse wave velocities	(n = 26)	(n = 23)	(n = 22)
Maternal Age (y) at 5–6 weeks	37 ± 7	32 ± 5	33 ± 4		0.009
BMI (kg/m2) pre-pregnancy	28 ± 5	25 ± 6	24 ± 6		0.04
Height (cm)	166 ± 8	164 ± 6	166 ± 7		0.92
Nulliparity	11 (42)	12 (52)	14 (64)		0.35
History of Medical Comorbidities	9 (35)	5 (25)	4 (18)		0.43
Participant Race n (%) Asian Black White Other	2 (7.7) 2 (7.7) 22 (84.6) 0 (0)	1 (4.3) 1 (4.3) 21 (91.3) 0 (0)	0 (0) 1 (4.5) 19 (86.4) 2 (9.1)
Participant Ethnicity Hispanic/Latino Non-Hispanic/Non-Latino	2 (7.7) 24 (92.3)	2 (8.7) 21 (91.3)	1 (4.5) 21 (95.5)
	0 CL	1 CL	> 1 CL	Fresh IVF	p-value
Prospective cohort study of obstetric outcomes†	(n = 94)	(n = 290)	(n = 153)	(n=146)
Age - yr	35.4 ± 4.2	35.5 ± 4.0	35.0 ± 3.9	36.3 ± 3.9	0.048
Age > 40 yr	16 (17.2)	41 (14.3)	15 (10.0)	24 (17.4)	0.27
BMI - kg/m2	23.6 ± 4.1	23.8 ± 4.4	24.3 ± 4.3	23.8 ± 4.5	0.37
BMI > 30	10 (10.6)	22 (7.6)	21 (13.7)	14 (9.7)	0.23
Nulliparity	34 (36.2)	71 (24.5)	51 (33.3)	48 (32.9)	0.08
Maternal smoking in pregnancy	1 (1.1)	2 (0.7)	2 (1.3)	2 (1.4)	0.82
History of hypertensive disease in previous pregnancy	1 (1.1)	5 (1.7)	1 (0.7)	2 (1.4)	0.93
History of chronic hypertension	2 (2.1)	9 (3.1)	3 (2.0)	9 (6.1)	0.23
PCOS	21 (22.3)	28 (9.7)	41 (26.8)	11 (7.5)	<0.001
Number of other medical problemsǂ	0.02 ± 0.1	0.1 ± 0.3	0.1 ± 0.2	0.1 ± 0.2	0.67
Partner Age - yr	38.1 ± 5.7	37.7 ± 5.4	37.3 ± 5.4	38.8 ± 6.1	0.12
Participant Race Asian Black White Other Unknown	48 (51) 1 (1) 38 (40) 4 (4) 4 (4)	133 (46) 1 (0.3) 131 (45) 25 (9) 2 (0.7)	70 (46) 1 (0.6) 71 (46) 14 (9) 0 (0)	65 (44) 0 (0) 70 (48) 9 (6) 2 (1.4)
Participant Ethnicity Hispanic/Latino Non-Hispanic/Non-Latino Unknown	3 (3) 88 (94) 3 (3)	17 (6) 273 (94) 0 (0)	11 (7) 139 (91) 3 (2)	8 (5) 137(94) 1 (~1)

^*Data are presented as mean ± SD or as number (% of total). A few participants listed more than one race.

^†Singleton live births from autologous oocytes.

^ǂNumber of other medical problems as risk factors for preeclampsia includes at least one of the following (ACOG recommendations 2013): cardiovascular disease, systemic lupus erythematosus, chronic renal disease, history of inherited thrombophilia, antiphospholipid antibody syndrome, type I or II diabetes mellitus, insulin resistance.

BMI: body mass index. CL: corpus luteum. Fresh IVF: in vitro fertilization with fresh embryo transfer.

Prospective cohort study of obstetric outcomes

This prospective cohort study was approved by Stanford’s IRB. Women receiving fertility care at Stanford were enrolled following confirmation of a viable pregnancy at ~8 weeks gestation. All women with a viable pregnancy at ~8 weeks gestation were eligible to participate with recruitment beginning in October 2011 and ending with deliveries projected to occur by the end of December 2017. All participants provided informed consent for collection of delivery records and also completed a research questionnaire. Prenatal care and delivery records were collected from whichever obstetrician and hospital provided care for each patient; data collection was not limited only to participants receiving prenatal care and delivery at Stanford. Analysis was restricted to singleton pregnancies conceived with autologous oocytes (Figure 1, Figure S1).

Figure 1.

Prospective cohort study of obstetric outcomes: enrollment, inclusions, and exclusions.

Participants were grouped according to categories of CL number: 1) 0 CL (FETs in a programmed cycle); 2) 1 CL (spontaneous conceptions, intrauterine insemination or FETs in a modified natural ovulatory cycle); 3) >1 CL (ovulation induction with timed intercourse, intrauterine insemination or FET); 4) IVF with fresh embryo transfer. Participants were also grouped by method of conception. For treatment protocols, see Methods in the online-only Data Supplement.

The primary outcome was the incidence of preeclampsia using the current American College of Obstetricians and Gynecologists definitions²¹ (also see Methods in the online-only Data Supplement). Preeclampsia was defined as the development of hypertension (systolic blood pressure [BP] 140 mmHg or higher and/or diastolic BP 90 mmHg or higher measured at least two times at least 4 hours apart) after 20 week’s gestation in a previously normotensive woman and proteinuria (≥ 300 mg protein in a 24-hour urine collection, protein/creatinine ratio > 0.3 or dipstick reading of at least 1+). In the absence of proteinuria but new onset hypertension, at least one of the following additional symptoms qualified for the diagnosis of preeclampsia: thrombocytopenia (< 100,000/μl), renal insufficiency (creatinine > 1.1 mg/dl), impaired liver function (liver transaminases levels twice normal), pulmonary edema, and cerebral or visual symptoms. These symptoms also qualified for the diagnosis of preeclampsia with severe features. Additional criteria defining preeclampsia with severe features were a systolic BP of ≥ 160 mmHg and/or a diastolic BP of ≥ 110 mmHg, HELLP syndrome (hemolysis [LDH increase], elevated liver enzymes [liver transaminases levels twice normal] and low platelet count [(< 100,000/μl]). Hypertensive disorders of pregnancy were adjudicated by review of the prenatal care and delivery records by the first author, a trained obstetrician-gynecologist without knowledge of CL number or method of treatment. Data collection instruments used for entry of pregnancy and neonatal outcomes were separate from records detailing patient demographics and treatment history.

The primary analysis focused on the comparison of hypertensive outcomes by CL status. Sub-analysis focused principally on comparison of hypertensive outcomes for two FET protocols—programmed versus modified natural cycle. Characteristics of the study population were described using frequencies and proportions for categorical variables and mean ± SD for continuous variables. Student’s t-test or Mann-Whitney test was used to ascertain the significance of differences between mean values of two continuous variables, ANOVA or Kruskal-Wallis test for more than two variables. Chi-square (χ²) or Fisher’s exact test was performed to test for differences in proportions of categorical variables between groups.

Multivariate logistic regression models were used with preeclampsia and preeclampsia with severe features (sPE) as the outcome and CL number group, type of FET protocol or retrieved oocyte number ≥ 20 as the predictor variables. We chose this oocyte number cut-off as it would be considered a high response based on US practice²² and included nearly 15% of our population. We adjusted for the effect of the following covariates: nulliparity, age, prior history of hypertension, BMI, diabetes (pre-gestational and gestational), and polycystic ovary syndrome (PCOS). Significance level was set at the <0.05 level. Statistical analyses were performed using statistical software R 3.4.3.²³.

RESULTS

Longitudinal Study of Pulse Wave Velocities

Participant Characteristics, Infertility Diagnoses, and Obstetrical and Neonatal Outcomes

Participant race, ethnicity, parity, maternal smoking, and history of hypertensive disease of pregnancy were not different among the cohorts. Average maternal age and BMI were 4–5 yr and 3–4 kg/m² greater, respectively, in the 0 CL compared to the single and multiple CL cohorts (both p< 0.05, Table 1). Infertility diagnoses were similar between women conceiving with IVF involving 0 or multiple CL with the exception of diminished ovarian reserve and male factor, which were more frequent in the 0 CL and multiple CL cohorts, respectively (Table S2). Although the study was not specifically powered for adverse obstetric outcomes., numerically more occurred in women conceiving by IVF (Table S3). The cesarean section rate and number of twin pregnancies were significantly greater, and the gestational age at delivery and newborn weight significantly less in the pregnancies conceived with IVF compared to those spontaneously conceived pregnancies (Table S4).

PWV and PWTT

In the women conceiving without IVF (singleton pregnancy/single CL), cfPWV declined by ~1 m/s during pregnancy, reached a nadir at 22–25 weeks of gestation, and returned to non-pregnant levels thereafter (Figure 2A and Table S5). These findings were consistent with the literature (Table S1). Because we hypothesized that the potential impact of the CL would be greatest in the first trimester, we analyzed cfPWV across the 3 cohorts for the first 4 study visits. In the full mixed model, the group-time interaction was not significant (p=0.14). In the reduced mixed model, there were significant effects of time (p<0.001) and group (p=0.02). Because cfPWV in the single and multiple CL groups were similar (p=0.95), we combined cfPWV for these 2 groups, and compared the aggregate with the cohort lacking a CL. This analysis revealed a borderline significant group-time interaction (p=0.06) and a non-significant group effect (p=0.14). Modeling the 0 CL group alone revealed that change in cfPWV over time was not significant in the first trimester (p=0.47) consistent with an attenuation of the expected decline. In contrast, the decline in cfPWV was significant for both the 1 CL and >1 CL cohorts (p=0.007 and 0.010, respectively). Moreover, the change in cfPWV from the pre-pregnant baseline among the 3 cohorts was most strikingly different at the end of the first trimester (10–12 weeks): 0 CL −0.18 m/s, >1 CL −0.42 m/s and 1 CL −0.66 m/s; p=0.01 (significant after Bonferroni correction for multiple testing).

Figure 2.

Changes in carotid-femoral (A) pulse wave velocity (cfPWV) and (B) pulse wave transit time (PWTT) during pregnancy in women conceiving with and without IVF. Mean ± 1.96SE (95% CI). BP, before pregnancy; PP, postpartum; m/s, meters/second; ms, milliseconds. Wilcoxon signed-rank test with Bonferroni correction was used to compare the three timepoints in the first trimester for each of the three cohorts, separately, with the pre-pregnancy baseline (BP). *p<0.017 for >1 CL, ⁺p<0.017 for 1 CL. 0 CL was not significantly different from BP at any timepoint in the first trimester. For further statistical analysis, see Results.

When cfPWTT was analyzed across the 3 cohorts for the first 4 study visits (Figure 2B), there was a significant effect of time (p<0.001), but in this case, there was also a significant group-time interaction (p=0.049) and a non-significant group effect (p=0.84). With combination of the single and multiple CL cohorts, which were similar (p=0.58), and comparing the aggregate with the group lacking a CL, there was again a significant group-time interaction (p=0.03) and non-significant group effect (p=0.56). Modeling the 0 CL group alone revealed that change in cfPWTT over time was not significant (p=0.32), again consistent with an attenuation of the expected rise. In contrast, the increase in cfPWTT was significant for both the 1 CL and >1 CL cohorts (p=0.003 and 0.024, respectively). Moreover, the change in cfPWTT from the pre-pregnant baseline among the 3 cohorts was most strikingly different at the end of the first trimester (10–12 weeks): 0 CL 2.56 ms, >1 CL 5.47 ms and 1 CL 10.52 ms; p=0.006 (significant after Bonferroni correction for multiple testing). Taken together, the data suggested a subdued decline of cfPWV and rise in cfPWTT during the first trimester in women conceiving without a CL. For analysis of cfPWV and cfPWTT across all study visits, see Results in the online-only Data Supplement. Relative to cfPWV, crPWV was less demonstrably affected during pregnancy in women conceiving without a CL (Figure S2, Table S6 and Results in the online-only Data Supplement).

To determine whether the changes in PWV and PWTT in the first trimester were restricted to only those who eventually developed adverse outcomes (Table S3), we removed these subjects and repeated the analyses across all 3 cohorts for the first trimester and for all visits except postpartum. On balance, after removing adverse obstetrical outcomes, findings comparable to the analysis of all subjects combined were observed for cfPWV and cfPWTT. In contrast, the generally smaller differences observed between the absent CL and other cohorts in the analysis of all subjects for crPWV were further diminished after removing adverse obstetrical outcomes (Results in the online-only Data Supplement).

Covariate and Other Analyses

Maternal age more powerfully affected cf- than crPWV, but the positive relationship between maternal age and PWV at each pregnancy time point was comparable among the 3 cohorts. Pre-pregnancy BMI also affected cf- more than crPWV, but again, the positive relationship between maternal age and PWV at each pregnancy timepoint was similar among the 3 groups. For further details, as well as additional analyses of PWV in relation to autologous and donor oocytes within the 0 CL cohort, singleton pregnancies, parity, fresh vs frozen embryos, number of oocytes retrieved, blood pressure and heart rate, see Results in the online-only Data Supplement and Figures S3 and S4.

Prospective Cohort Study for Obstetrical Outcomes

Participant characteristics

Of the 838 women with a known pregnancy outcome, 787 delivered by the time of closure of the database for analysis on December 31, 2017, with none of the non-viable pregnancies associated with development of preeclampsia (Figure 1). Analysis for this report was restricted to 683 singleton live births achieved using autologous oocytes.

Women who conceived via fresh IVF were slightly older compared to women in the other CL groups (Table 1). Our population was 50% Caucasian, 49% Asian, and 1% African American. Regarding ethnicity, our population was 6% Hispanic and 94% non-Hispanic. Pre-conception body mass index (BMI), race, ethnicity, parity, smoking status, medical history (e.g., chronic hypertension), hypertensive disorder in previous pregnancy, and other medical problems were not different among groups. There were some differences in participant characteristics between methods of conception, as expected based on the nature of each treatment (Tables S7 and S8). Although not the focus of this report, neonatal and maternal outcomes other than hypertensive disorders of pregnancy are reported in Tables S9 and S10. There were no differences in newborn weight, gestational age at delivery, or placental complications such a placenta previa, accreta, or abruption (Table S9). However, our study was not specifically powered to find a difference in these outcomes.

Incidence of hypertensive disorders of pregnancy

The incidence of preclampsia with severe features (sPE) was higher after conception in the absence of a CL compared to 1 CL, > 1 CL or fresh IVF (P=0.004, Table 2). In analysis restricted to FETs, conceptions with programmed FETs had higher frequencies of preeclampsia (12.8% vs 3.9%, P=0.02) and sPE (9.6% vs 0.8%, P<0.001), compared to modified natural cycle FETs (Table 2). FETs in a programmed cycle were also more likely to develop preeclampsia or sPE compared to fresh embryo transfers in IVF cycles (12.8% vs 4.7%, P=0.047 and 9.6% vs 2.7%, P=0.04 in pairwise comparisons). There was no difference in the rate of preeclampsia between conceptions occurring spontaneously with those occurring via modified natural cycle FET.

Table 2.

Hypertensive Disorders of Pregnancy as a Function of Corpus Luteum (CL) Category and Method of Conception for the Prospective Cohort Study of Obstetric Outcomes. Data are presented for singleton live births from autologous oocytes.

Corpus luteum category
Hypertensive disorder of pregnancy	0 CL (n=94)	1 CL (n=290)	> 1 CL (n=153)	Fresh IVF (n=146)	P-value
Gestational hypertension	3 (3.2)	8 (2.8)	3 (2.0)	0	0.14
Preeclampsia	12 (12.8)	14 (4.8)	9 (5.9)	7 (4.7)	0.06
Preeclampsia with severe features	9 (9.6)	4 (1.4)	6 (3.9)	4 (2.7)	0.004
Early-onset preeclampsia	0	1 (0.3)	1 (0.7)	1 (0.7)	1
HELLP-syndrome	0	0	0	2 (1.4)	0.06
Chronic hypertension witd superimposed preeclampsia	1 (1.1)	5 (1.7)	0	1 (0.7)	0.41
Eclampsia	0	0	0	0
Metdod of conception
Hypertensive disorder of pregnancy	Prog FET (n=94)	Nat FET (n=127)	Fresh IVF (n=146)	Spont. (n=143)	P-value
Gestational hypertension	3 (3.2)	4 (3.2)	0	3 (2.1)	0.11
Preeclampsia	12 (12.8)	5 (3.9)	7 (4.7)	7 (4.9)	0.05
Preeclampsia witd severe features	9 (9.6)	1 (0.8)	4 (2.7)	3 (2.1)	0.005
Early-onset preeclampsia	0	1 (0.8)	1 (0.7)	0	0.84
HELLP-syndrome	0	0	2 (1.4)	0	0.26
Chronic hypertension witd superimposed preeclampsia	1 (1.1)	2 (1.6)	1 (0.7)	3 (2.1)	0.78
Eclampsia	0	0	0	0

CL: Corpus luteum; Prog FET: programmed cycle frozen embryo transfer; Nat FET: modified natural cycle frozen embryo transfer; fresh IVF: in vitro fertilization with fresh embryo transfer; spont.: spontaneous conception. Number (% of total) are presented.

The multivariate adjusted odds ratios (AOR) in the absence of a CL compared to 1 CL for preeclampsia and sPE were 2.73 (95% confidence interval (CI) 1.14–6.49) and 6.45 (95% CI 1.94–25.09), respectively (Table 3). Embryo transfer in a programmed cycle compared to a modified natural cycle predicted preeclampsia (AOR 3.55; 95% CI 1.20–11.94; P=0.03) and sPE (AOR 15.05; 95% CI 2.59–286.27; P=0.01). A number of ≥ 20 retrieved oocytes was not associated with a statistically significant difference in the odds ratio for development of preeclampsia (OR 1.95; 95% CI 0.09–15.66; P=0.57).

Table 3.

Multivariable Model Results for the Prospective Cohort Study Of Obstetric Outcomes. Data are presented for singleton live births from autologous oocytes.

Variable	Level	AOR	95% CI	P value
Preeclampsia
CL number	0 CL vs. 1 CL	2.73	1.14–6.49	0.02
Frozen embryo transfer	programmed FET vs. modified natural cycle FET	3.55	1.20–11.94	0.03
Preeclampsia with severe features
CL number	0 CL vs. 1 CL	6.45	1.94–25.09	0.003
Frozen embryo transfer	programmed FET vs. modified natural cycle FET	15.05	2.59–286.27	0.01

CI: confidence interval; CL: corpus luteum; FET: frozen embryo transfer; AOR: adjusted odds ratio OR adjusted for the effect of age, nulliparity, prior history of hypertension, BMI, PCOS, diabetes (pre-gestational and gestational).

DISCUSSION

To our knowledge, this is the first study to demonstrate that women who conceived without a CL experienced higher rates of preeclampsia and preeclampsia with severe features compared with women who conceived with 1 CL or multiple CLs. A sub-analysis of FETs demonstrated that programmed cycles were associated with higher preeclampsia rates compared with modified natural ovulatory cycles. In a parallel clinical physiology study, absence of the CL was associated with impairment of the expected pregnancy-associated increase in central arterial (predominantly elastic) compliance as reflected by attenuated decline in cfPWV and increase of cfPWTT.

The detrimental impact of absent CL on PWV and PWTT predominated in the first trimester, consistent with the absence of circulating CL factor(s), which normally peak at that time of gestation. (Later in pregnancy, the CL regresses, albeit not completely). Thereafter, hormones secreted by the developing placenta^{24, 25} likely mediated the fall of cfPWV and rise in cfPWTT observed after 14–16 weeks in gravid women without a CL, thus allowing at least partial recovery from the earlier deficit. Both age ²⁶ and BMI ²⁷ positively correlate with cfPWV, but due to the exponential shape of the curve, there may be little age-related change up to 40 years. Nevertheless, these factors may have contributed to the modestly elevated pre-pregnant cfPWV in the 0 CL participants who, on average, were 4–5 years older and 3–4 kg/m² heavier. To our knowledge, however, there is no evidence that these modest differences in age or BMI would necessarily preclude the normal gestational decrease in cfPWV and increase in cfPWTT. Indeed, the change in cfPWV and cfPWTT actually did more or less parallel the other two cohorts after 14–16 gestational weeks (vide supra). Thus, the vasculature seems to be inherently capable of increasing compliance during pregnancy in the women without a CL. Moreover, the positive correlation of maternal age and pre-pregnancy BMI with cfPWV was not different among the 3 cohorts at any of the study visits; thus, the modestly greater maternal age and pre-pregnancy BMI in the cohort without a CL is less likely to account for the attenuated decline of cfPWV and rise in cfPWTT during the first trimester. Rather, the lack of a stimulus, i.e., absence of circulating CL factor(s) is a more likely explanation, although we cannot definitively exclude a role for the modestly older age, higher BMI or sub-fertility of this cohort.

In our study, the detriment in central arterial compliance among pregnant women without a CL persisted with sub-group analyses of the uncomplicated pregnancies, suggesting absence of the CL may be a factor predisposing women to develop preeclampsia, but a healthy placenta or maternal compensatory responses can overcome this predisposition. Alternatively, a “second hit” is required to elicit disease. In other studies powered for prediction of obstetrical outcomes, deficient circulatory adaptations during early gestation were shown to be associated with adverse pregnancy outcomes including preeclampsia^{12, 17, 18}.

Although we did observe a higher rate of preeclampsia in the absence of the CL, we did not observe a higher rate of preeclampsia in superovulatory cycles with fresh embryo transfer. In line with this observation, both PWV and PWTT were comparably changed during pregnancy in the single and multiple CL cohorts. Furthermore, a high number of retrieved oocytes (≥ 20) was not predictive for the development of preeclampsia in our cohort, and the number of oocytes retrieved did not show a significant relationship with the degree of gestational rise in arterial compliance as reflected by declining PWV or rising PWTT at any of the time points evaluated during pregnancy. In the medical literature, IVF involving superovulatory cycles has been associated with increased risk of preeclampsia in some^28–32, but not all^{33, 34} reports, particularly when adjusted for plurality³⁵. Closer examination of this literature reveals that studies reporting increased risk of preeclampsia with IVF have often grouped both fresh (multiple CL) and frozen (frequently absent CL) transfers together^{28, 32, 36, 37}; therefore, one cannot discern whether the increased risk was seen with both fresh and frozen transfer. Most studies which distinguished fresh from FET generally reported higher risk with FET^5–10, and similar rates of preeclampsia in the fresh transfer group compared with the general population^{6, 7,}. Thus, our finding of no increased risk of preeclampsia with fresh IVF is essentially consistent with prior work.

FET performed in a modified natural cycle in our study was not associated with increased preeclampsia risk. Although multiple studies reported increased risk of hypertensive disorders of pregnancy with FET compared with fresh embryo transfer, whether this increase is attributable to the absence of the CL is not discernable from the literature, because protocols utilized in the FET cycles have generally not been well-described^{5, 6, 8–10, 38}. In one study, which did not report an increased risk of preeclampsia with FET, 75% of FET cycles were conducted in a natural ovulatory cycle³⁹. Our findings are timely given the increasing utilization of FETs^{40, 41}. Although there are reported benefits^{5, 8, 42,43,44}, if FET increases the risk of preeclampsia^5–10 or other adverse outcomes such as macrosomia^{8, 45}, then its increasing utilization may lead to serious short- and long-term consequences for maternal and child health, and efforts need to be made to reduce risk.

Theoretically, circulating relaxin is one biologically plausible mediator, as relaxin emanates solely from the CL during human pregnancy^46–48, and is a potent vasodilator^49–51, which mediates circulatory changes and increases of arterial compliance in the gravid rat model^{14, 15}. That is, the lack of circulating relaxin (below the limit of assay detection for all subjects throughout pregnancy without CL in our study, data not shown) may contribute to the attenuated increase in central arterial (aortic) compliance and higher preeclampsia risk in pregnant women lacking a CL. However, there were no significant correlations between the change in circulating relaxin concentration and cfPWV or cfPWTT relative to pre-pregnancy baseline for the 1 CL cohort during the first trimester (data not shown), perhaps suggesting that relaxin is not involved at all, does not act alone, or possibly exerts an indirect, permissive action to effect increases in central arterial (aortic) compliance. On the other hand, the limited number of subjects may have precluded significant correlations. Conceivably, the absence of other, as of yet unidentified CL factors(s) may contribute (except estradiol or progesterone which are administered for luteal support). However, proof of this concept requires further study, and ultimately perhaps, demonstration that replacement of the missing CL factor(s) restores the circulatory deficit and reduces the rate of preeclampsia.

This study of maternal physiology has some significant strengths and provides a novel contribution to the medical literature. Serial investigation of circulatory function throughout pregnancy in women conceiving by IVF in the context of CL status is, to our knowledge, unprecedented. There were multiple serial assessments of PWV during pregnancy, as well as a pre-pregnant control performed in the absence of the CL. We analyzed both PWV and PWTT and observed comparable, but reciprocal changes in PWTT relative to PWV.

Our parallel prospective study of obstetric outcomes is also the first to carefully examine the hypothesis that absence of the CL is associated with increased risk of preeclampsia. Women were enrolled prospectively, before the outcome of interest was known. The inclusion of comparator groups of sub-fertile women (spontaneous conceptions after sub-fertility and conceptions with fertility treatments other than IVF) reduces the chance of confounding factors associated with infertility. In contrast to most previous studies, we reported outcomes according to the specific FET protocol used. Analysis was restricted to singleton pregnancies conceived with autologous oocytes. Multiple potentially relevant covariates were included in the regression analysis. Deliveries occurring in both academic and community hospitals were included. Another key strength of our study is careful diagnosis of preeclampsia based on adjudication directly from medical records by an obstetrician-gynecologist blinded to CL group and according to current ACOG guidelines, not from patient report or from a surveillance database. In contrast to many prior studies, distinctions regarding severity of hypertensive disorders were clearly reported in our study. It is also noteworthy that absence of the CL was associated with abnormal maternal cardiovascular adaptation to pregnancy manifested as perturbations of aortic compliance and increased incidence of preeclampsia in the Florida and Stanford populations, respectively, while superovulation did not demonstrate detectable effects in either population.

Our study has several limitations. On the one hand, observations were restricted to conceptions occurring at only two fertility centers, but on the other, it could also be viewed as a strength that an effect of absent CL was noted in two different populations (Florida and Stanford). Although it seems likely that the deficiency noted in maternal physiology and the increased incidence of preeclampsia would be observed in the same population of pregnant women with absent CL, our study design obviously precluded a definitive conclusion in this regard, and about whether the detriment in maternal physiology is causally related to increased preeclampsia risk. The longitudinal study of PWV was underpowered to discern significant or borderline significant group-time interactions in many of the covariate and other analyses, e.g., after removing the IVF participants with adverse obstetrical outcomes from the cfPWV analysis, though the pattern of change in cfPWV over the first trimester for the three cohorts was comparable to the analysis including all participants. The prospective study of obstetric outcomes was underpowered to detect potentially clinically important differences in perinatal outcomes for the infant and rare outcomes for the mother. Although regression analysis adjusted for the 6 most important potential confounders, the number of cases of preeclampsia (n=41) may have limited our ability to definitively determine whether or not a confounding variable contributed to the association between absent CL and incidence of preeclampsia.The studies were observational, and thus it is possible that some confounding factor which was not included in the analysis could also at least partly explain the changes in maternal circulatory function noted with absence of the CL. For, example, sub-fertile subjects were not included in the PWV study as an additional control group for the 0 CL cohort, which was disproportionately affected by reduced ovarian reserve. Of note, however, the 0 CL cohort did manifest a decrease in cfPWV after the first trimester paralleling the other two cohorts. This finding is inconsistent with a confounding influence of sub-fertility or reduced ovarian reserve in some cases of sub-fertility, which should pertain throughout gestation, but rather implicates factor(s) missing in the first trimester. One possibility is absent CL factor(s) that normally predominate in the first trimester before the “corpus luteal-placental shift”, after which placental factor(s) circulate.

Despite these limitations, this study provides important data to support the development of a randomized controlled trial of FET protocols, powered to be able to detect a difference in the rate of preeclampsia, as well as to comprehensively investigate maternal physiology in ancillary studies of subsets of these cohorts in a serial fashion before, during and after pregnancy. Given the potential for impact of protocol choice on the obstetrical outcome of FET, we also recommend that registries and investigators reporting on FET outcomes include details regarding specific protocols used.

In conclusion, we observed an increased rate of preeclampsia with FET in the absence of a CL, but no significant difference in the frequency of preeclampsia between modified natural cycle FET and spontaneous conception. In parallel, we found impaired gestational increases of central arterial compliance in the absence of a CL. If our findings are confirmed in further studies, it is plausible that the increased risk of preeclampsia associated with FET could be mitigated by utilizing physiologic FET protocols which lead to development of a CL, such as a modified natural cycle or ovulation induction for oligoovulatory women. Replacement of missing circulating CL product(s) could eventually be recommended. Although still to be proven, the beneficial impact on pregnancy outcome by more closely mimicking physiological conditions including the presence of one CL in IVF protocols could be mediated at least partly through improved maternal circulatory function.

PERSPECTIVES

Multiple studies reported an increased risk of preeclampsia in women conceiving by in vitro fertilization (IVF), especially those undergoing IVF with frozen embryo transfer (FET), a method that is being utilized more and more. However, the reason for this increased risk is not understood. We found a significantly increased risk for preeclampsia and preeclampsia with severe features in FET cycles performed without the corpus luteum (CL), but not in IVF cycles with FET performed in a natural ovulatory cycle with a CL. In a parallel study of comprehensive and detailed assessment of clinical physiology, the expected pregnancy-associated increase in central arterial (aortic) compliance was attenuated, particularly during early gestation in the absence of the CL. Although the ovary, and hence the CL, can be removed in experimental animal models of pregnancy, IVF is the only way to examine human pregnancy physiology in the absence of a CL. Our finding of deficient elevations in central (aortic) arterial compliance is biologically plausible, insofar as the CL secretes vasoactive hormones like relaxin not replaced in FET cycles, which in the gravid rat model was shown to be critical for the maternal hemodynamic changes of pregnancy at least during early to mid-gestation. Given the increasing utilization of FET in IVF, our finding of perturbed maternal cardiovascular function and increased preeclampsia risk in this setting makes this report particularly relevant and timely. Further studies are warranted to determine if these novel findings hold true in other populations.

Novelty and Significance.

1. What Is New?

   • Gestational increases in central (aortic) arterial compliance were attenuated during pregnancy in women conceiving with IVF in the absence of a CL.
   • Women conceiving by IVF without a CL demonstrated higher incidence of preeclampsia and preeclampsia with severe features.
2. What Is Relevant?

   • Recent studies demonstrated increased incidence of preeclampsia after conceiving with IVF using frozen embryo transfer. Our work offers a potential explanation, i.e., FET is typically performed using a programmed cycle that precludes CL formation. In gcontrast, when FET was performed in a modified natural cycle with a CL present, the risk of preeclampsia and preeclampsia with severe features was not elevated.
   • IVF without a CL was also associated with impaired maternal circulatory adaptation to pregnancy.
3. Summary

   • These insights support the need for further study of maternal physiology and preeclampsia risk in FET and other IVF protocols conducted with and without a CL, particularly given the increasing utilization of FET in clinical practice.
   • If the finding of increased preeclampsia risk after FET in programmed cycles is confirmed, FET performed in a natural cycle might alleviate this increased risk. Ultimately, the crucial, missing circulating CL factor(s) might be replaced.
   • Absence of circulating CL factor(s), perturbed maternal circulatory adaptation to pregnancy and increased preeclampsia risk may be linked, but additional investigation is also needed to establish causality.