Cryopreserved embryo transfer

is an independent risk factor

for placenta accreta
Daniel J. Kaser, M.D.,a Alexander Melamed, M.D., M.P.H.,a Charles L. Bormann, Ph.D.,a Dale E. Myers, Sc.M.,a
Stacey A. Missmer, Sc.D.,a,b,c Brian W. Walsh, M.D.,a Catherine Racowsky, Ph.D.,a
and Daniela A. Carusi, M.D., M.Sc.a
a
Department of Obstetrics, Gynecology and Reproductive Biology, and b Department of Medicine at Channing Laboratory,
Brigham and Women's Hospital, Harvard Medical School; and c Department of Epidemiology, Harvard School of Public
Health, Boston, Massachusetts

Objective: To explore the association between cryopreserved embryo transfer (CET) and risk of placenta accreta among patients uti-
lizing in vitro fertilization (IVF) and/or intracytoplasmic sperm injection (ICSI).
Design: Case-control study.
Setting: Academic medical center.
Patient(s): All patients using IVF and/or ICSI, with autologous or donor oocytes, undergoing fresh or cryopreserved transfer, who
delivered a live-born fetus at R24 weeks of gestation at our center, from 2005 to 2011 (n ¼ 1,571), were reviewed for placenta
accreta at delivery.
Intervention(s): Cases of accreta (n ¼ 50) were matched by age and prior cesarean section to controls (1:3) without accreta. The asso-
ciation between CET and accreta was modeled using conditional logistic regression, controlling a priori for age and placenta previa.
Receiver operating characteristic curves were used to determine thresholds of endometrial thickness and peak serum E2 levels related
to accreta.
Main Outcome Measure(s): Placenta accreta.
Result(s): Univariate predictors of accreta were non-Caucasian race (odds ratio [OR] 2.85, 95% confidence interval [CI] 1.25–6.47);
uterine factor infertility (OR 5.80, 95% CI 2.49–13.50); prior abdominal or laparoscopic myomectomy (OR 7.24, 95% CI 1.92–27.28);
and persistent or resolved placenta previa (OR 4.25, 95% CI 1.94–9.33). In multivariate analysis, we observed a significant
association between CET and accreta (adjusted OR 3.20, 95% CI 1.14–9.02), which remained when analyses were restricted to cases
of accreta with morbid complications (adjusted OR 3.87, 95% CI 1.08–13.81). Endometrial thickness and peak serum E2 level were
each significantly lower in CET cycles and those with accreta.
Conclusion(s): Cryopreserved ET is a strong independent risk factor for accreta among patients
using IVF and/or ICSI. A threshold endometrial thickness and a ‘‘safety window’’ of optimal peak
E2 level are proposed for external validation. (Fertil SterilÒ 2015;103:1176–84. Ó2015 by Amer- Use your smartphone
ican Society for Reproductive Medicine.) to scan this QR code
Key Words: Adherent placenta, estradiol safety window, frozen embryo, IVF, trophoblast and connect to the
discussion forum for
this article now.*
Discuss: You can discuss this article with its authors and with other ASRM members at http://
fertstertforum.com/kaserd-cryopreserved-embryo-transfer-placenta-accreta/ * Download a free QR code scanner by searching for “QR
scanner” in your smartphone’s app store or app marketplace.

P
lacenta accreta is defined by the that the anchoring villi of the placenta sents as an adherent placenta and is
absence of the decidua basalis are directly apposed to the myome- often accompanied by substantial
layer of the endometrium, such trium. Clinically, this condition pre- morbidity at the time of delivery. The
Received November 19, 2014; revised December 24, 2014; accepted January 15, 2015; published online pathogenesis of this condition is not
March 4, 2015. well defined, but is thought to be related
D.J.K. has nothing to disclose. A.M. has nothing to disclose. C.L.B. has nothing to disclose. D.E.M. has
nothing to disclose. S.A.M. has nothing to disclose. B.W.W. has nothing to disclose. C.R. is on the
primarily to a deficiency in endometrial
Board of the American Society for Reproductive Medicine (ASRM); is a consultant for Auxogyn, decidualization, and secondarily to
Inc.; has received honoraria, travel expense reimbursement, and payment for developing educa- aberrant maternal vascular remodeling
tional materials from ASRM; and receives royalties from UpToDate, and Cambridge University
Press. D.A.C. has nothing to disclose. and/or excessive invasion of the extra-
Reprint requests: Daniel J. Kaser, M.D., Division of Reproductive Medicine, Department of Obstetrics, villous trophoblast (1).
Gynecology and Reproductive Biology, Brigham and Women's Hospital, 75 Francis St, ASB1-3,
Boston, Massachusetts 02115 (E-mail: dkaser@partners.org). Established risk factors for placenta
accreta include prior cesarean section,
Fertility and Sterility® Vol. 103, No. 5, May 2015 0015-0282/$36.00 placenta previa, and advanced maternal
Copyright ©2015 American Society for Reproductive Medicine, Published by Elsevier Inc.
http://dx.doi.org/10.1016/j.fertnstert.2015.01.021 age (2). Recently, in vitro fertilization

1176 VOL. 103 NO. 5 / MAY 2015


(IVF) (3–5), and specifically cryopreserved embryo transfer (CET)
(6), has been proposed to be an additional risk factor for placenta
accreta. The association between CET and accreta was reported
by Ishishara et al. (6) in a registry study after transfer of
previously vitrified day-3 or day-5 embryos. Their analysis,
however, did not control for known accreta risk factors;
measuring the true contribution of CET per se was therefore
not possible.
The purpose of the current study was to investigate
whether CET remained a significant predictor of placenta
accreta, in those using IVF and/or intracytoplasmic sperm
injection (ICSI), when the relationship was controlled for
factors known to increase the risk of accreta. In addition,
the study was designed to evaluate whether certain cycle-
specific parameters are associated with the development of
accreta.
MATERIALS AND METHODS
Study Population
Our center maintains a prospectively collected infertility
database that systematically records all IVF cycle-specific
variables, along with a prospectively collected obstetric
database that captures salient information regarding the
antepartum, intrapartum, and postpartum courses of pa-
tients delivering at Brigham and Women's Hospital. This
study, approved by the Partners' Healthcare Institutional
Review Board, reconciled these 2 databases, identifying
all patients who underwent a fresh or cryopreserved day-
3 or day-5 ET at our center and subsequently delivered at
our hospital between 2005 and 2011. Patients utilizing
IVF and/or ICSI who delivered at or beyond 24 weeks of
gestation were selected for our initial study population (n
¼ 1,571). Both autologous and donor oocyte cycles were
included, whereas cycles involving a gestational carrier,
and those with embryos imported from another facility,
were excluded.
From this cohort, cases of accreta were identified by
automated electronic medical record review, using the Bool-
ean phrase ‘‘*creta’’ OR ‘‘adhere*’’ as a search query. Any de-
gree of placental invasion, including placenta increta or
percreta, was classified as ‘‘accreta’’ for the purpose of this
analysis. All initial records identified in this manner were
subjected to a secondary screen by comprehensive chart re-
view for the presence of accreta at delivery or on placental
pathology. Only the first case of accreta contributed by
each patient was included. Further demographic characteris-
tics, such as surgical history, presence of intrauterine adhe-
sions, and placenta previa, were collected by retrospective
review of our electronic medical record, with specific atten-
tion to consultation notes, prenatal visits, operative reports,
baseline and obstetric ultrasounds, and hospital discharge
summaries.
Cases with accreta were matched to controls without
accreta, according to patient age at delivery and history
of prior cesarean section, in a 1:3 allocation ratio. If <3
control cycles with the appropriate cesarean status were
available, then the age range considered to be acceptable
for matching was expanded in half-year increments, from
VOL. 103 NO. 5 / MAY 2015
Fertility and Sterility®
1 year up to a maximum of 5 years. If <3 controls were
available for a given case at this point, no further matching
was performed. If, on the other hand, multiple control cy-
cles fulfilled the matching criteria for a particular case, the
set of controls was chosen at random.
Clinical and Laboratory Protocols
For fresh autologous cycles, standard controlled ovarian
stimulation and monitoring protocols were used. When R2
lead follicles reached a mean diameter of R18 mm, ovulation
was triggered with 10,000 international units (IU) of human
chorionic gonadotropin (hCG) (Profasi; EMD Serono), by
intramuscular (IM) injection. Approximately 36 hours post-
trigger, ultrasound-guided oocyte retrieval was performed
under intravenous general anesthesia. Oocytes were insemi-
nated, via IVF or ICSI, 4–6 hours after retrieval in pre-
equilibrated media with 5% CO2 content at 37 C.
Luteal phase support was started 1 day after retrieval
with IM progesterone in oil, or 2 days after retrieval with
vaginal supplementation, and was continued until
10 weeks of gestation. Embryo transfer occurred on day
3 or day 5 of culture; standard clinical practice in our pro-
gram during the study period was for cleavage-stage ET.
Good-quality supernumerary embryos were cryopreserved
using Liebo's 1-step slow-freeze protocol, as previously
described (7, 8).
For fresh donor cycles, recipients were prescribed contin-
uous oral contraceptive pills (OCPs), with a 7-day overlap be-
tween the start of leuprolide acetate (Lupron; AbbVie) and
stopping the OCPs. On cycle day 2, recipients who were
adequately suppressed reduced their doses of leuprolide ace-
tate from 20 IU to 10 IU daily and began exogenous estradiol
(E2) preparations, either oral (3 mg of Estrace, twice daily;
Warner Chilcott), vaginal (1 mg of Estrace, twice daily), or
transdermal (0.1 mg of Vivelle [Novartis], 3 patches every
other day).
Nonsuppressed patients continued the same dose of
leuprolide acetate and repeated baseline testing 1 week
later. Serum E2 was checked after 4 days of supplementa-
tion, and dose adjustments were made as necessary to
achieve a serum level of R180 pg/mL. Because of clinical
protocol, the endometrial thickness was not measured for
recipients of fresh embryos derived from donor oocytes.
Progesterone (25 mg IM) was initiated the night of donor
oocyte retrieval and was increased to 50 mg per day the
following day. Leuprolide acetate was discontinued the
day after retrieval. Patients with a serum progesterone level
of <20 ng/mL on the day of transfer increased their daily
dose by 50%–100%.
For CET cycles, leuprolide acetate pretreatment was
standard practice during the study period; however, 6 pa-
tients had cycles programmed with exogenous E2 and pro-
gesterone only. Patients pretreated with leuprolide acetate
underwent the same baseline and monitoring criteria as
described earlier, with the addition of a transvaginal ultra-
sound, between days 14 and 18 of exogenous E2, to ensure
that the endometrial thickness was R7 mm. If this target
was not achieved, either the dose of E2 was increased or
1177
ORIGINAL ARTICLE: ASSISTED REPRODUCTION
the mode of delivery was changed. Progesterone replace-
ment was initiated 3 days before transfer for thawed
cleavage-stage embryos, and 5 days before transfer for
thawed blastocysts.
For cycles without leuprolide acetate pretreatment, serum
progesterone was checked before initiating exogenous pro-
gesterone, to confirm that the recipient had not ovulated; cy-
cles were cancelled if progesterone was >3 ng/mL. For
natural-cycle CETs, once the endometrial thickness measured
R7 mm and the lead follicle was R15 mm, urinary or blood
luteinizing hormone (LH) was monitored daily. Day-3 em-
bryos were transferred 4 days after urinary LH surge, or
5 days after blood LH surge; Day-5 embryos were transferred
6 days after urinary LH surge, or 7 days after blood LH surge.
Exogenous progesterone was not routinely prescribed during
natural-cycle CETs.
Outcome Variables
Our primary outcome was placenta accreta, defined by histo-
logic confirmation of chorionic villi directly attached to or
invading the myometrium, or clinically by the description
of an adherent placenta by the attending obstetrician at the
time of delivery. Patients diagnosed with accreta based solely
on radiographic findings (i.e., without clinical suspicion at
delivery or pathologic confirmation) were excluded. Our sec-
ondary outcome was morbid accreta, which was identified in
a subset of patients with accreta.
Those who had any of the following, in addition to ac-
creta, were defined as having morbid accreta: hemorrhage
(defined as >1,000 mL of blood loss, regardless of mode of de-
livery, or any blood transfusion); nonroutine procedures to
stop excess blood loss (placement of an intrauterine tampo-
nade balloon; oversewing of the placental bed; ligation of
the uterine, utero-ovarian, or hypogastric arteries; percuta-
neous transcatheter embolization; or gravid hysterectomy);
or additional procedures required to remove the placenta
(manual extraction after vaginal delivery, sharp curettage
or wedge resection for adherent placenta at cesarean section,
suction dilation and/or sharp curettage immediately after de-
livery or up to 6 weeks postpartum, or hysteroscopic resection
of retained products of conception).
Statistical Analyses and Model Building
Statistical analyses were performed with statistical analysis
system (SAS), version 9.3 (SAS Institute, Inc.). The association
between CET and accreta was modeled using conditional lo-
gistic regression. Adjusted odds ratios (aORs), 95% confidence
intervals (CIs), and 2-sided Wald P values were calculated.
Differences were considered statistically significant if the
effect estimate excluded 1.0 from the 95% CI, and the Wald
2-sided P value was < .05.
Models were adjusted a priori for patient age at delivery (to
account for residual confounding by age after matching) and
placenta previa. Any history of confirmed previa, either
resolved or persistent, was used in this definition, given that
both were significantly associated with accreta in univariate
analyses. Additional variables tested as potential confounders
included the following: gravidity; parity; race; body mass
1178
index (BMI); uterine factor infertility (intrauterine synechiae,
adenomyosis, uncorrected uterine septum, bicornuate uterus,
or submucosal fibroids); use of donor oocytes; history of curet-
tage; operative hysteroscopy; or myomectomy (abdominal or
laparoscopic); number of prior failed transfers; any fibroid at
baseline ultrasound; oral contraceptive lead-in; leuprolide ac-
etate pretreatment; micromanipulation (ICSI, assisted hatch-
ing, or embryo biopsy); type of luteal phase support (IM vs.
vaginal supplementation); any fetal reduction at <20 weeks
(spontaneous or selective); and multifetal delivery.
These potential confounders were tested individually in
the relationship between CET and placenta accreta. An indi-
vidual factor was retained in the final model if it altered the
point estimate for CET by >10% (9). The final model for CET
therefore was adjusted for patient age, gravidity, race, uterine
factor infertility, and placenta previa. After these steps, 2 sub-
analyses were performed: the first was restricted to morbid ac-
cretas; the second was restricted to autologous cycles.
Endometrial thickness and serum peak E2 levels were
considered to potentially reside along the causal pathway
from CET to accreta; accordingly, they were not tested as con-
founders. Instead, receiver operating characteristic (ROC)
curves were drawn, and thresholds that optimized Youden's
J-statistic (sensitivity þ specificity  1) were chosen as cut-
offs. Data were dichotomized about these thresholds, and
the proportions of cycles with values above and below each
were calculated. P values were determined by c2 analysis.
As dictated by clinical practice, endometrial thickness
was not routinely measured for fresh donor cycles, and 24 re-
cipients had missing values. For ROC analysis of variables
with missing data (e.g., endometrial thickness), only the
known values were used to generate the thresholds. Univari-
ate analyses were used to determine if these cutoffs were
meaningfully associated with CET. Performance tests
including sensitivity, specificity, and positive and negative
predictive values were calculated for each threshold.
RESULTS
Study Population and Case Definition
Among consecutive patients delivering R1 viable infant
R24 weeks of gestational age, at our hospital from 2005 to
2011 (n ¼ 54,947), 498 cases of accreta were identified
(0.91% overall incidence for IVF and non-IVF deliveries).
All patients who conceived from a day-3 or day-5 transfer
at our center were then identified (n ¼ 1,571), from which
51 cases of placenta accreta were confirmed. The incidence
of accreta after IVF and/or ICSI at our center was thus 3.3%
(51 of 1,571), a significantly higher percentage than that
among the remainder of the population delivering at Brigham
and Women's Hospital during the same time period (0.84%;
447 of 53,376; P< .0001).
Among IVF patients, the incidence of accreta differed ac-
cording to cycle type (fresh: 2.5%; 34 of 1,351 vs. CET: 7.7%;
17 of 220; P< .001). One instance of accreta after CET was
excluded from further analysis, as the patient had already
contributed an accreta to our dataset in a prior pregnancy.
The remaining 50 cases of accreta were subsequently matched
1:3 to controls without accreta from our cohort of patients
VOL. 103 NO. 5 / MAY 2015
 Fertility and Sterility®

who conceived following IVF and/or ICSI, with the exception
of 1 case that had only 2 suitable age-matched controls with
the appropriate prior cesarean status. Thus, for the 50 cases of
accreta, there were 149 controls.
Cases of accreta were classified into 3 groups: pathology
proven; clinically adherent placenta; or both; each group was
further classified as either morbid or nonmorbid (Fig. 1). The
demographic and cycle characteristics for cases and controls
are shown in Table 1. No statistically significant differences
were found in age, parity, race, median BMI, number of prior
failed transfers, surgical history, uterine factor infertility, or
placenta previa between patients who had fresh vs. cryopre-
served transfer (Supplemental Table 1, available online).
Clinical Outcomes
Compared with controls without accreta, cases were more
likely to experience hemorrhagic complications, along with
major and minor surgical morbidities (Supplemental
Table 2, available online). The median estimated blood loss
was higher among accretas, compared with controls:
1,000 mL (interquartile range [IQR] 750–2,000 mL) vs.
700 mL (IQR 500–800 mL) (P< .0001). Patients with accreta
were more likely to deliver by cesarean section (78% vs.
63%, unadjusted OR 2.23, 95% CI 1.01–4.9; P¼ .05). The like-
lihood of experiencing any morbidity remained significant
after controlling for mode of delivery. In addition, no differ-
ences were found in the rates of placenta previa (persistent or
resolved) among patients following fresh vs. cryopreserved
transfer (19.7%; 31 of 157 vs. 16.6%; 7 of 42; P¼ .65).
Predictors of Accreta
Our primary predictor of interest, CET, was the only cycle-
specific variable that was significantly associated with
placenta accreta in univariate analyses (OR 2.58, 95% CI
1.12–6.00; P¼ .03). Other patient-specific predictors of
placenta accreta included non-Caucasian race (OR 2.85,
95% CI 1.25–6.47; P¼ .01); uterine factor infertility (OR
5.80, 95% CI 2.49–13.50; P< .0001), in particular, intrauter-
ine synechiae (OR 4.10, 95% CI 1.14–14.70; P¼ .03), adeno-
myosis (OR 10.18, 95% CI 1.11–93.02; P¼ .04), or a
bicornuate uterus (8%; 4 of 50 vs. 0%; P< .01 by Fisher's
exact test); prior abdominal or laparoscopic myomectomy
(OR 7.24, 95% CI 1.92–27.28; P< .01); and persistent or
resolved placenta previa (OR 4.25, 95% CI 1.94–9.33;
P< .001).
In multivariate analysis, gravidity, race, and uterine fac-
tor infertility individually confounded the relationship be-
tween CET and accreta and thus were retained in the final
model. After controlling for these factors, along with age,
cesarean section, and placenta previa a priori, CET remained
a strong independent predictor of placenta accreta (aOR
3.20, 95% CI 1.14–9.02; P¼ .03). When the analysis was
restricted to only cases of morbid accreta, the observed asso-
ciation was still significant (aOR 3.87, 95% CI 1.08–13.81;
P¼ .04). Finally, when the analysis excluded the 37 donor
oocyte cycles, the association between CET and accreta
among autologous cycles was strengthened (aOR 4.54,
95% CI 1.65–12.45; P< .01), suggesting effect modification
based on the embryo source.

FIGURE 1

Flow diagram showing method of case definition for 51 cases of placenta accreta culled from a cohort who used IVF and/or ICSI (n ¼ 1,571) and
delivered R1 live-born infant at R24 weeks of gestational age at our hospital, from 2005 to 2011. Note that 1 delivery was excluded, as the patient
had already contributed an accreta to this analysis in a prior pregnancy. The proportion of cases that were classified as clinically morbid is likewise
shown. Here, morbidity was defined as: hemorrhage (>1,000 mL of blood loss, regardless of mode of delivery, or any blood transfusion);
nonroutine procedures to stop excess blood loss (gravid hysterectomy or uterine artery embolization); or additional procedures to remove the
placenta (see Materials and Methods section for details).
Kaser. CET and risk of placenta accreta. Fertil Steril 2015.

VOL. 103 NO. 5 / MAY 2015 1179

ORIGINAL ARTICLE: ASSISTED REPRODUCTION

pathway between CET and accreta—endometrial thickness

TABLE 1
and peak serum E2. Univariate and ROC analyses were used
Demographic and cycle characteristics of cases with placenta
to identify cutoffs that would discriminate accreta from
accreta and controls without placenta accreta in a population nonaccreta patients, and CET from fresh cycles.
using IVF and/or ICSI. Both the group of patients with placenta accreta and those
Placenta No placenta using CET had significantly thinner endometrial thickness
accreta accreta values than their counterparts. The median endometrial thick-
Factor (n [ 50) (n [ 149)
ness for patients with accreta was 8.8 mm (IQR 7.3–11.2),
Median age (y) 39.3 (33.9–41.7) 39.1 (34.3–41.3) compared with 10.4 mm for those without accreta (IQR 8.4–
Gravidity 12.1; P< .01). Subanalysis of morbid accretas likewise re-
1 19 (38.0) 41 (27.5)
R2 31 (62.0) 108 (72.5) vealed significantly lower values (morbid accreta: 8.8 mm,
Parity IQR 7.5–10.7; no accreta: 10.3 mm, IQR 8.4–12.0). The median
0 29 (58.0) 88 (59.1) endometrial thickness for CET cycles was also thinner than
R1 21 (42.0) 61 (40.9)
Race
that for fresh cycles: 8.4 mm vs. 10.6 mm (P¼ .001).
Caucasian 36 (72.0) 130 (87.3) In ROC analysis (Fig. 2A), thresholds that optimized You-
Non-Caucasian 14 (28.0) 19 (12.7) den's J-statistic (sensitivity þ specificity  1; Fig. 2B) were
Median BMI (kg/m2) 23.8 (21.7–26.6) 23.7 (21.1–26.8) selected to dichotomize the data. Endometrial thickness of
Any prior failed ET 24 (48.0) 91 (61.1)
Uterine factor infertility 18 (36.0) 13 (8.7) %9.7 mm was chosen as the optimal threshold and was found
Fibroids at baseline 18 (36.0) 46 (30.9) to be predictive of accreta with an area under the curve (AUC) of
Prior surgical history 0.65. Among patients with known values of endometrial thick-
Cesarean section 14 (28.0) 41 (27.5) ness, 74% of patients with accreta had an endometrial thickness
>1 prior cesarean 1 of 14 (7.1) 4 of 41 (9.8)
section of %9.7 mm, compared with 45.6% of patients without accreta
Myomectomy 9 (18.0) 6 (4.0) (OR 4.44, 95% CI 1.85–10.64; P< .001). The sensitivity and
(abdominal or specificity were 70% and 61.4%, respectively. The positive pre-
laparoscopic)
Dilation and 23 (46.0) 73 (49.0)
dictive value (PPV) of endometrial thickness of %9.7 mm for
curettage accreta was 37%, and the negative predictive value (NPV)
Any operative 9 (18.0) 25 (16.8) was 86.2%. The distribution of patients with and without ac-
hysteroscopy creta, according to endometrial thickness, is plotted as a histo-
Any placenta previa 19 (38.0) 19 (12.8)
Resolved previa 9 (22.5) 14 (9.7) gram in Figure 2C. For all patients with an endometrial
Persistent previa 10 (20.0) 5 (3.4) thickness < 9 mm, there was an excess risk of accreta (Fig. 2D).
Any fetal reduction 10 (20.0) 21 (14.1) The median peak serum E2 was lowest for patients with
(<20 wk gestation)
Selective reduction 3 (6) 5 (3.4)
morbid accreta (732 pg/mL, IQR 367–1,939), followed by all
Spontaneous 7 (14.0) 17 (11.4) accreta (1,143 pg/mL, IQR 404–1,939), and those without ac-
reduction creta (1,702 pg/mL, IQR 639–2,558; P¼ .02 for morbid accreta
Multifetal delivery 8 (16.0) 39 (26.2) vs. no accreta, and P¼ .03 for all accreta vs. no accreta). In
Cycle Treatment
CET 16 (32) 26 (17.5) addition, the median peak E2 level was significantly lower
Donor oocyte cycle 10 (20) 27 (18.1) with CET than with fresh cycles (563 vs. 1,883 pg/mL,
ICSI 24 (48.0) 59 (39.6) P¼ .001). A cutoff point for peak serum E2 that was predictive
Assisted hatching 20 (40.0) 73 (49.0) of accreta was determined by ROC analysis (Fig. 3A). The
OCP lead-in 25 (50.0) 73 (49.0)
Leuprolide acetate 33 (66.0) 113 (75.8) threshold value that optimized Youden's J-statistic (Fig. 3B)
pretreatment was 732 pg/mL (AUC 0.60; sensitivity 46.9%; specificity
Mode of E2 for uterine 73.8%; PPV 37.1%; NPV 80.9%), which happened to be the
preparation
Oral 21 (95.5) 33 (75.0)
median peak E2 for morbid accreta. Applying this threshold
Vaginal or 1 (4.6) 11 (25.0) to our population, 48% of accreta patients had an E2
transdermal %732 pg/mL, compared with 26.2% of patients without ac-
Mode of luteal creta (P¼ .004). The distribution of patients with and without
progesterone
IM 33 (66.0) 101 (67.8) accreta, according to serum E2 levels, is shown in Figure 3C.
Vaginal 17 (34.0) 48 (32.2) For all patients with a peak serum E2 < 1,000 pg/mL, there
Note: Values represent n (%) or median (interquartile range). Uterine factor infertility was was an excess risk of accreta (Fig. 3D).
defined as intrauterine synechiae, adenomyosis, uncorrected septum, bicornuate uterus,
or submucosal fibroids. For resolved previa, the denominator excludes those with persistent We found that the route of E2 administration played a sig-
previa. For mode of E2 for uterine preparation, the denominator excludes fresh autologous nificant role in the relationship between peak E2 level and
cycles.
Kaser. CET and risk of placenta accreta. Fertil Steril 2015.
placenta accreta. Median serum E2 levels varied according
to route of exogenous E2 administration (oral: 360 pg/mL
[IQR 280–438]; vaginal: 1,388 pg/mL [IQR 639–1,940]; trans-
Endometrial Thickness, Peak Estradiol, and Risk of dermal: 1,033 pg/mL [IQR 398–1,508]; P< .0001). Further-
Accreta more, the route of E2 supplementation in programmed CET
To clarify the relationship between CET and invasive placen- cycles was significantly associated with accreta: of the 15
tation, we explored the interplay between CET, accreta, and 2 patients with accreta who had CET, 14 of 15 (93.3%) were
cycle-specific factors thought to reside along the causal prescribed oral E2, compared with 16 of 26 (61.5%) without

1180 VOL. 103 NO. 5 / MAY 2015

 Fertility and Sterility®

FIGURE 2

Endometrial thickness and risk of placenta accreta among patients using IVF and/or ICSI (n ¼ 1,571). (A) Receiver operating characteristic curve for
endometrial thickness threshold as a diagnostic test for accreta. Area under the curve was 0.649. (B) Distribution of Youden's J-statistic (sensitivity þ
specificity  1) for endometrial thickness. An optimal threshold of endometrial thickness was chosen as 9.7 mm. (C) Histogram of patients with and
without accreta according to endometrial thickness. (D) Scatterplot of risk difference for accreta according to endometrial thickness. A positive
value indicates an excess risk of accreta.
Kaser. CET and risk of placenta accreta. Fertil Steril 2015.

accreta who had CET (P¼ .03, by Fisher's exact test). All pa-
tients having CET who received oral supplementation had a
serum E2 value that was lower than our proposed threshold
of 732 pg/mL, whereas 81.8% of the patients having CET
who had non-oral (i.e., vaginal or transdermal) routes had a
serum E2 value above this threshold.
After deriving the endometrial thickness and peak serum
E2 values that optimally predicted accreta, we determined
whether these cutoffs significantly differentiated patients
having CET from those having a fresh transfer. We found
that patients who had CET were significantly more likely to
have an endometrial thickness of %9.7 mm (74% vs. 38%,

VOL. 103 NO. 5 / MAY 2015 1181

ORIGINAL ARTICLE: ASSISTED REPRODUCTION

FIGURE 3

Peak serum E2 and risk of placenta accreta among patients using IVF and/or ICSI (n ¼ 1,571). (A) Receiver operating characteristic curve for peak
serum E2 threshold. Area under the curve was 0.603. (B) Distribution of Youden's J-statistic for peak E2. One proposed threshold of peak serum E2
level was chosen as 732 pg/mL, as it optimized the absolute difference between true and false positives and additionally happened to be the median
E2 for morbid accretas. Given the multimodal nature of this curve, however, other relevant threshold values may exist. (C) Histogram of patients with
and without accreta, according to peak serum E2 level. (D) Scatterplot of risk difference for accreta, according to peak serum E2 level. A positive
value indicates an excess risk of accreta.
Kaser. CET and risk of placenta accreta. Fertil Steril 2015.

respectively, P< .0001), as well as a peak serum E2 level of DISCUSSION

%732 pg/mL (83% vs. 18%, respectively, P< .0001). These In this study, we investigated whether CET was a significant
findings support the possibility that a thin endometrium predictor of placenta accreta in patients who undergo IVF
and low serum E2 levels may mediate the relationship and/or ICSI, after controlling for other patient and cycle
between CET and invasive placentation.

1182 VOL. 103 NO. 5 / MAY 2015


characteristics. We found that CET seems to be a strong inde-
pendent risk factor for placenta accreta, even after controlling
for patient age, prior cesarean section, placenta previa, and
uterine factor infertility. Restricting our analysis to only clin-
ically morbid accretas upheld the observed effect, which led
us to conclude that the observed increased risk of accreta
may be directly related to factors associated with CET, rather
than to underlying patient characteristics. The CET factors
may be related to the freeze and/or thaw process itself, the
mode of uterine preparation, or an interaction of the 2.
Here, we determined that cycles with accreta had thinner
endometrial linings and lower peak serum E2 levels than did
cycles without accreta, allowing us to propose threshold
values at which the risk of accreta was significantly higher.
In addition, we observed that the relationship between CET
and accreta may be mediated by the route of E2 administra-
tion. Patients for whom oral E2 was prescribed had lower
serum E2 levels than did those for whom vaginal or trans-
dermal E2 was prescribed. The proportion of patients who
developed accreta after CET was significantly higher among
those who had been prescribed oral as opposed to transdermal
or vaginal E2. This finding suggests that route of E2 adminis-
tration in uterine-preparation cycles may be a modifiable risk
factor for accreta.
One possible mechanism to explain the association be-
tween CET and accreta is that the degree of trophoblastic in-
vasion and extent of vascular remodeling at the time of
implantation may be modulated by the serum E2 level (10).
In both human (11) and nonhuman primates (12, 13),
aberrant placentation and its sequelae, such as pre-
eclampsia and fetal growth restriction, may be related to
supraphysiologic E2 concentrations. In CET cycles, the peak
E2 level is typically much lower than in a fresh stimulation,
as demonstrated both in the current study and in the literature
(14, 15). As a result, trophoblastic invasion may go
unchecked. In support of this possibility is the finding that
repeated low doses of E2 in a murine model have been
shown to maintain the uterus in an artificially prolonged
receptive state that permits trophoblast ingrowth; when
exposed to high doses of E2, the uterus becomes refractory
to implantation within 24 hours (16). Accordingly, a lower
serum E2 level, coupled with the second ‘‘hit’’ of a thinner
decidualized endometrium, as shown in the current study
with CET cycles, may result in exuberant trophoblastic
growth during a protracted window of implantation.
Thus, we propose a model in which serum E2 levels, at
either extreme, affect maternal and neonatal outcomes: Low
levels may be associated with placental overgrowth (e.g.,
placenta accreta, as in the current study, and large-for-
gestational-age infants [17, 18]), whereas high levels may
be associated with placental undergrowth (e.g., pre-
eclampsia and growth-restricted infants) (11). Our ROC anal-
ysis suggests that the lower limit of optimal serum E2 may be
approximately 700 pg/mL (Fig. 3B); however, given the multi-
modal nature of this curve, other relevant cutoffs may exist.
An upper limit, generated from the 90% percentile of fresh cy-
cles in another program, above which the odds were higher of
both pre-eclampsia and fetal-growth restriction, was
3,450 pg/mL (11). External validation of these thresholds,
VOL. 103 NO. 5 / MAY 2015
Fertility and Sterility®
with consideration of the route of E2 supplementation, should
be performed before they are applied clinically.
Our findings are consistent with the prior study by Ishish-
ara et al. (6), in which the authors report higher odds (aOR
3.16) of placenta accreta among recipients of thawed vs. fresh
embryos. In contrast to that Japanese registry study, which
assessed the risk of placenta accreta after transfer of vitrified
cleavage-stage embryos or blastocysts, our analysis was
limited to slow-frozen embryos, and all but 1 of our patients
(198 of 199) had a day-3 ET. Furthermore, the study by Ishish-
ara et al. (6) did not specify how placenta accreta was defined,
and adjusted for only patient age, stage of embryo develop-
ment, fresh or cryopreserved transfer, and infant gender.
This group acknowledged the limitations of such a registry
study and called for future analyses to control for known
risk factors of accreta.
Our study design allowed us to do just that—we controlled
or adjusted a priori for known risk factors of accreta,
including patient age, prior cesarean delivery, and placenta
previa. In addition, we tested 16 other covariates, including
uterine factor infertility, that plausibly could confound the
relationship between predictor and outcome. Through use of
conditional logistic regression, we confirmed that CET was a
strong independent predictor of placenta accreta.
The current study has several limitations. During the study
period, our center performed only slow-freezing, and trans-
ferred almost exclusively cleavage-stage embryos; thus, our
data cannot be extrapolated to transfer of warmed embryos
that were previously vitrified or that were at the blastocyst
stage. Our study, though, complements the one by Ishishara
et al. (6), in which vitrification was used exclusively, and
included both day-3 and day-5 transfers. Given the remark-
ably concordant results between the 2 studies, the method of
freeze-thaw and the duration of in vitro culture likely did
not contribute to accreta risk; rather, factors associated with
endometrial preparation, the general process of freezing and
thawing, or other factors not explored in this analysis, may
be related to development of accreta. Additionally, our dataset
included only 1 natural-cycle CET, and our database excluded
gestational carrier cycles, so we likewise cannot comment on
accreta risk in these types of uterine-preparation cycles.
To study cycle- and embryo-specific factors that may be
associated with development of accreta, only patients who
both conceived and subsequently delivered in our program
were included. This factor may have inflated the observed rates
of accreta, as patients diagnosed prenatally with a complicated
pregnancy may have been more likely to deliver at a tertiary
care center, whereas those with uncomplicated pregnancies
may have chosen to deliver elsewhere. However, this effect
would be true for both CET and fresh cycles, and would not
be expected to influence our results; furthermore, identifying
cases and randomly selecting controls from a single patient
population should support the internal validity of our
findings.
With regard to our outcome definition, we included cases
of accreta that were defined clinically and/or histologically.
As the presence of a focally adherent placenta and lower de-
grees of hemorrhage may allow for uterine conservation, the
published literature includes clinically identified accretas in
1183
ORIGINAL ARTICLE: ASSISTED REPRODUCTION
the absence of pathologic confirmation. These definitions
have ranged from any adherent placenta (19, 20), to
adherence that requires ‘‘active management’’ (4), to
excessive bleeding from the placental bed (3). We included
all of these clinical definitions, and analyzed our data with
both a broad case definition (any clinical or pathologic
identification) and a narrower case definition (only those
that involved excessive bleeding or additional procedures,
i.e., ‘‘morbid’’ accretas). The increased risk of accreta after
CET was observed using both categorizations. Owing to the
retrospective nature of our study, we relied on the
description of an adherent placenta given by the delivering
obstetrician and were not able to corroborate this finding
objectively.
Finally, routes of administration for supplemental E2 were
not standardized in this study, and patients may have been
prescribed oral, vaginal, or transdermal E2 for uterine prepara-
tion. As serum and uterine E2 levels are known to vary accord-
ing to route of administration (because of gastrointestinal
conversion to estrone and the first uterine pass effect), these
serum values may underestimate the true estrogenic activity
at the level of the uterus. Thus, the observed low serum E2
levels among patients undergoing CET, who largely received
oral E2, may not correlate directly with end-organ effect. How-
ever, the significant association observed between the oral
route of administration and development of accreta suggests
that delivery route may be important.
Full clarification of any contribution that CET or the type,
extent, or duration of endometrial preparation used in these cy-
cles may make to the incidence of placenta accreta requires
further analysis. A critical unresolved question is whether
altering the specifics of the uterine preparation, such as pre-
treatment with or without leuprolide acetate, and mode of E2
supplementation (oral, vaginal, or transdermal), will modify
the risk of accreta in CET cycles. Follow-up studies should
explore these variables, with consideration given to the derived
cutoff values for serum peak E2 and endometrial thickness.
Additionally, an important issue to tease apart is whether
embryo cryopreservation specifically, and not just, more
generally, the transfer of an embryo to a prepared uterus, is
related to development of accreta. The effect modification
that we observed between CET and the use of a donor vs.
autologous embryo suggests that more than just uterine prep-
aration is involved. Future research should address both of
these factors to elucidate whether one or both are principal
drivers of the observed effect.
In conclusion, to our knowledge, this is the first multivar-
iate analysis reporting that CET is a strong independent pre-
dictor of accreta, after controlling exhaustively for known
accreta risk factors and other possible confounders unique
to assisted reproductive technology and the infertile popula-
tion. Possible explanatory mechanisms include a thinner
endometrial lining and lower serum E2 level in CET cycles
that result in unchecked growth of the extravillous tropho-
blast into the myometrium. Taken together with literature
showing morbidity at the opposite end of the E2 spectrum,
including an increased risk of pre-eclampsia and fetal-
growth restriction, a ‘‘safety window’’ of serum E2 (700–
3,450 pg/mL) for assisted reproductive technology cycles is
1184
thus proposed. External validation of a safe range of serum
E2 that minimizes maternal and neonatal morbidities should
be a research priority.
REFERENCES
1. Tantbirojn P, Crum CP, Parast MM. Pathophysiology of placenta creta: the
role of decidua and extravillous trophoblast. Placenta 2008;29:639–45.
2. Miller DA, Chollet JA, Goodwin TM. Clinical risk factors for placenta previa-
placenta accreta. Am J Obstet Gynecol 1997;177:210–4.
3. Esh-Broder E, Ariel I, Abas-Bashir N, Bdolah Y, Celnikier DH. Placenta accreta
is associated with IVF pregnancies: a retrospective chart review. BJOG 2011;
118:1084–9.
4. Fitzpatrick KE, Sellers S, Spark P, Kurinczuk JJ, Brocklehurst P, Knight M. Inci-
dence and risk factors for placenta accreta/increta/percreta in the UK: a na-
tional case-control study. PLoS One 2012;7:e52893.
5. Hayashi M, Nakai A, Satoh S, Matsuda Y. Adverse obstetric and perinatal
outcomes of singleton pregnancies may be related to maternal factors asso-
ciated with infertility rather than the type of assisted reproductive technol-
ogy procedure used. Fertil Steril 2012;98:922–8.
6. Ishishara O, Araki R, Kuwahara A, Itakura A, Saito H, Adamson G. Impact of
frozen-thawed single-blastocyst transfer on maternal and neonatal
outcome: an analysis of 277,042 single-embryo transfer cycles from 2008
to 2010 in Japan. Fertil Steril 2014;101:128–33.
7. Leibo SP. A one step method for direct nonsurgical transfer of frozen-
thawed bovine embryos. Theriogenology 1984;21:767–90.
8. Kaser DJ, Ginsburg ES, Missmer SA, Correia KF, Racowsky C. Intramuscular
progesterone versus Crinone 8% vaginal gel for luteal phase support for day
3 cryopreserved embryo transfer. Fertil Steril 2012;98:1464–9.
9. Greenland S. Modeling and variable selection in epidemiologic analysis. Am
J Public Health 1989;79:340–9.
10. Simon C, Dominguez F, Valbuena D, Pellicer A. The role of estrogen in uter-
ine receptivity and blastocyst implantation. Trends Endocrinol Metab 2003;
14:197–9.
11. Imudia AN, Awonuga AO, Doyle JO, Kaimal AJ, Wright DL, Toth TL, et al. Peak
serum estradiol level during controlled ovarian hyperstimulation is associated
with increased risk of small for gestational age and preeclampsia in singleton
pregnancies after in vitro fertilization. Fertil Steril 2012;97:1374–9.
12. Albrecht ED, Bonagura TW, Burleigh DW, Enders AC, Aberdeen GW, Pepe GJ.
Suppression of extravillous trophoblast invasion of uterine spiral arteries by
estrogen during early baboon pregnancy. Placenta 2006;27:483–90.
13. Bonagura TW, Pepe GJ, Enders AC, Albrecht ED. Suppression of extravillous
trophoblast vascular endothelial growth factor expression and uterine spiral
artery invasion by estrogen during early baboon pregnancy. Endocrinology
2008;149:5078–87.
14. Hancke K, More S, Kreienberg R, Weiss JM. Patients undergoing frozen-
thawed embryo transfer have similar live birth rates in spontaneous and arti-
ficial cycles. J Assist Reprod Genet 2012;29:403–7.
15. Niu Z, Feng Y, Sun Y, Zhang A, Zhang H. Estrogen level monitoring in arti-
ficial frozen-thawed embryo transfer cycles using step-up regime without pi-
tuitary suppression: Is it necessary? J Exp Clin Assist Reprod 2008;5:4.
16. Ma WG, Song H, Das SK, Paria BC, Dey SK. Estrogen is a critical determinant
that specifies the duration of the window of uterine receptivity for implan-
tation. Proc Natl Acad Sci 2003;100:2963–8.
17. Pinborg A, Henningsen AA, Loft A, Malchau SS, Forman J, Andersen AN.
Large baby syndrome in singletons born after frozen embryo transfer
(FET): Is it due to maternal factors or the cryotechnique? Hum Reprod
2014;29:618–27.
18. Wennerholm UB, Henningsen AK, Romundstad LB, Bergh C, Pinborg A,
Skjaerven R, et al. Perinatal outcomes of children born after frozen-
thawed embryo transfer: a Nordic cohort study from the CoNARTaS group.
Hum Reprod 2013;28:2545–53.
19. Eshkoli T, Weintraub AY, Sergienko R, Sheiner E. Placenta accreta: risk fac-
tors, perinatal outcomes, and consequences for subsequent births. Am J Ob-
stet Gynecol 2013;208:219.e1–7.
20. Mercer B, Berghella V, Foley M, Kilpatrick S, Saade G, Grobman W, et al.
Placenta accreta. Am J Obstet Gyencol 2010;203:430–9.
VOL. 103 NO. 5 / MAY 2015
 Fertility and Sterility®

SUPPLEMENTAL TABLE 1

Demographic and cycle characteristics of patients having fresh vs. cryopreserved ET.
Variable Fresh ET (n [ 157) CET (n [ 42) P value
Median age (y) 39.1 (34.9–41.2) 38.1 (33.3–45.5) .75
Gravidity
1 107 (68.2) 32 (76.2) .31
R2 50 (31.8) 10 (23.8)
Parity
R1 63 (40.1) 19 (45.2) .55
Race
Caucasian 129 (82.2) 37 (88.1) .36
Non-Caucasian 28 (17.8) 5 (11.9)
Median BMI (kg/m2) 23.7 (21.5–26.5) 23.9 (20.8–26.8) .73
Any prior failed ET 86 (54.8) 29 (69.0) .11
Uterine factor infertility 23 (14.6) 8 (19.0) .48
Fibroids at baseline 51 (32.5) 13 (31.0) .85
Prior surgical history
Cesarean section 43 (27.4) 12 (28.6) .85
>1 prior cesarean section 4 (9.3) 1 (8.3) 1.0
Myomectomy (abdominal or laparoscopic) 10 (6.4) 5 (11.9) .32
Dilation and curettage 78 (49.7) 18 (42.2) .49
Any operative hysteroscopy 25 (15.9) 9 (21.4) .49
Any placenta previa 31 (19.8) 7 (16.6) .82
Resolved previa 19 (13.1) 4 (10.3) .79
Persistent previa 12 (7.6) 3 (7.1) 1.0
Note: Values represent n (%) or median (interquartile range). Uterine factor infertility was defined as intrauterine synechiae, adenomyosis, uncorrected septum, bicornuate uterus, or submucosal
fibroids. Denominator excludes those with persistent previa.
Kaser. CET and risk of placenta accreta. Fertil Steril 2015.

VOL. 103 NO. 5 / MAY 2015 1184.e1

ORIGINAL ARTICLE: ASSISTED REPRODUCTION

SUPPLEMENTAL TABLE 2

Comparison of hemorrhagic and surgical morbidities among cases with placenta accreta and controls without placenta accreta in a population
using IVF and/or ICSI.
Placenta No placenta Unadjusted
Variable accreta (n [ 50) accreta (n [ 149) OR (95% CI) P value
Any morbidity 34 (68) 16 (11) 20.1 (7.1–57.1) < .0001
Hemorrhage 21 (42) 13 (8.7) 7.5 (3.1–17.8) < .001
Median blood loss 1,000 (750–2,000) 700 (500–800) – < .0001
Blood loss >1,000 mL 20 (40) 12 (8) 7.3 (3.1–17.5) < .001
Transfusion 15 (30) 5 (3.4) 9.0 (3.3–24.8) < .001
Major surgical morbidity 7 (14) 1 (0.7) 21.0 (2.6–170.7) .004
Hysterectomy 4 (8) 1 (0.7) 12.0 (1.3–107.4) .03
Uterine artery embolization 4 (8) 0 (0) – .004
Minor surgical morbidity 24 (48) 5 (3) 17.5 (6.1–50.6) < .001
Intrauterine balloon catheter placement 2 (4) 2 (1.3) 1.6 (.4–7.4) .50
Ascending uterine artery ligation 4 (8) 0 (0) – .004
Placental bed oversewing 7 (14) 0 (0) – < .001
Uterine curettage 12 (24) 2 (1.3) 18.0 (4.0–80.4) < .001
Manual extraction of placenta 7 (14) 3 (2) 7.9 (2.0–31.9) < .01
Hysteroscopic resection of retained 2 (4) 0 (0) – .06
products of conception
Note: Values represent n (%) or median (interquartile range). ‘‘Any morbidity’’ was defined as postpartum hemorrhage, or major or minor surgical morbidity. ‘‘Hemorrhage’’ was defined as blood
loss >1,000 mL or receipt of any blood transfusion Uterine curettage category includes those done in the time period from immediately after delivery up to 6-weeks postpartum. Manual extraction
of placenta was for vaginal deliveries only.
Kaser. CET and risk of placenta accreta. Fertil Steril 2015.

1184.e2 VOL. 103 NO. 5 / MAY 2015