Cryopreserved embryo transfer is an independent risk factor for placenta accreta

ORIGINAL ARTICLE: ASSISTED REPRODUCTIVE TECHNOLOGY

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 utilizing 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 association 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 Use your smartphone E2 level are proposed for external validation. (Fertil SterilÒ 2015;-:-–-. Ó2015 by American to scan this QR code Society for Reproductive Medicine.) and connect to the Key Words: Adherent placenta, estradiol safety window, frozen embryo, IVF, trophoblast Discuss: You can discuss this article with its authors and with other ASRM members at http:// fertstertforum.com/kaserd-cryopreserved-embryo-transfer-placenta-accreta/

P

lacenta accreta is defined by the absence of the decidua basalis layer of the endometrium, such

that the anchoring villi of the placenta are directly apposed to the myometrium. Clinically, this condition pre-

Received November 19, 2014; revised December 24, 2014; accepted January 15, 2015. 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 Board of the American Society for Reproductive Medicine (ASRM); is a consultant for Auxogyn, Inc.; has received honoraria, travel expense reimbursement, and payment for developing educational materials from ASRM; and receives royalties from UpToDate, and Cambridge University Press. D.A.C. has nothing to disclose. Reprint requests: Daniel J. Kaser, M.D., Division of Reproductive Medicine, Department of Obstetrics, Gynecology and Reproductive Biology, Brigham and Women's Hospital, 75 Francis St, ASB1-3, Boston, Massachusetts 02115 (E-mail: [email protected]). Fertility and Sterility® Vol. -, No. -, - 2015 0015-0282/$36.00 Copyright ©2015 American Society for Reproductive Medicine, Published by Elsevier Inc. http://dx.doi.org/10.1016/j.fertnstert.2015.01.021 VOL. - NO. - / - 2015

discussion forum for this article now.*

* Download a free QR code scanner by searching for “QR scanner” in your smartphone’s app store or app marketplace.

sents as an adherent placenta and is often accompanied by substantial morbidity at the time of delivery. The pathogenesis of this condition is not well defined, but is thought to be related primarily to a deficiency in endometrial decidualization, and secondarily to aberrant maternal vascular remodeling and/or excessive invasion of the extravillous trophoblast (1). Established risk factors for placenta accreta include prior cesarean section, placenta previa, and advanced maternal age (2). Recently, in vitro fertilization 1

ORIGINAL ARTICLE: ASSISTED REPRODUCTIVE TECHNOLOGY (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 cyclespecific 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 patients 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 day3 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 Boolean phrase ‘‘*creta’’ OR ‘‘adhere*’’ as a search query. Any degree 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 review 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 characteristics, such as surgical history, presence of intrauterine adhesions, and placenta previa, were collected by retrospective review of our electronic medical record, with specific attention 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 2

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 cycles 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 posttrigger, ultrasound-guided oocyte retrieval was performed under intravenous general anesthesia. Oocytes were inseminated, via IVF or ICSI, 4–6 hours after retrieval in preequilibrated 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 program 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 continuous oral contraceptive pills (OCPs), with a 7-day overlap between 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 acetate 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 supplementation, 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 patients had cycles programmed with exogenous E2 and progesterone only. Patients pretreated with leuprolide acetate underwent the same baseline and monitoring criteria as described earlier, with the addition of a transvaginal ultrasound, 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 VOL. - NO. - / - 2015

Fertility and Sterility® the mode of delivery was changed. Progesterone replacement 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 progesterone, to confirm that the recipient had not ovulated; cycles 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 embryos 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 histologic 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 secondary outcome was morbid accreta, which was identified in a subset of patients with accreta. Those who had any of the following, in addition to accreta, were defined as having morbid accreta: hemorrhage (defined as >1,000 mL of blood loss, regardless of mode of delivery, or any blood transfusion); nonroutine procedures to stop excess blood loss (placement of an intrauterine tamponade balloon; oversewing of the placental bed; ligation of the uterine, utero-ovarian, or hypogastric arteries; percutaneous 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 delivery 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 logistic 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 VOL. - NO. - / - 2015

index (BMI); uterine factor infertility (intrauterine synechiae, adenomyosis, uncorrected uterine septum, bicornuate uterus, or submucosal fibroids); use of donor oocytes; history of curettage; operative hysteroscopy; or myomectomy (abdominal or laparoscopic); number of prior failed transfers; any fibroid at baseline ultrasound; oral contraceptive lead-in; leuprolide acetate pretreatment; micromanipulation (ICSI, assisted hatching, 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 individual 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 subanalyses were performed: the first was restricted to morbid accretas; 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 confounders. Instead, receiver operating characteristic (ROC) curves were drawn, and thresholds that optimized Youden's J-statistic (sensitivity þ specificity  1) were chosen as cutoffs. 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 recipients 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. Univariate 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 (0.8%; 447 of 52,929; P< .0001). Among IVF patients, the incidence of accreta differed according 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 3

ORIGINAL ARTICLE: ASSISTED REPRODUCTIVE TECHNOLOGY 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. cryopreserved 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 likelihood of experiencing any morbidity remained significant after controlling for mode of delivery. In addition, no differences 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 cyclespecific 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, intrauterine synechiae (OR 4.10, 95% CI 1.14–14.70; P¼ .03), adenomyosis (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 factor infertility individually confounded the relationship between 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 association 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.

4

VOL. - NO. - / - 2015

Fertility and Sterility®

TABLE 1 Demographic and cycle characteristics of cases with placenta accreta and controls without placenta accreta in a population using IVF and/or ICSI.

Factor

Placenta accreta (n [ 50)

No placenta accreta (n [ 149)

Median age (y) 39.3 (33.9–41.7) 39.1 (34.3–41.3) Gravidity 1 19 (38.0) 41 (27.5) R2 31 (62.0) 108 (72.5) Parity 0 29 (58.0) 88 (59.1) R1 21 (42.0) 61 (40.9) Race Caucasian 36 (72.0) 130 (87.3) Non-Caucasian 14 (28.0) 19 (12.7) 23.8 (21.7–26.6) 23.7 (21.1–26.8) Median BMI (kg/m2) Any prior failed ET 24 (48.0) 91 (61.1) Uterine factor infertility 18 (36.0) 13 (8.7) Fibroids at baseline 18 (36.0) 46 (30.9) Prior surgical history Cesarean section 14 (28.0) 41 (27.5) >1 prior cesarean 1 of 14 (7.1) 4 of 41 (9.8) section Myomectomy 9 (18.0) 6 (4.0) (abdominal or laparoscopic) Dilation and 23 (46.0) 73 (49.0) curettage Any operative 9 (18.0) 25 (16.8) hysteroscopy Any placenta previa 19 (38.0) 19 (12.8) Resolved previa 9 (22.5) 14 (9.7) Persistent previa 10 (20.0) 5 (3.4) Any fetal reduction 10 (20.0) 21 (14.1) (<20 wk gestation) Selective reduction 3 (6) 5 (3.4) Spontaneous 7 (14.0) 17 (11.4) reduction Multifetal delivery 8 (16.0) 39 (26.2) Cycle Treatment CET 16 (32) 26 (17.5) Donor oocyte cycle 10 (20) 27 (18.1) ICSI 24 (48.0) 59 (39.6) Assisted hatching 20 (40.0) 73 (49.0) OCP lead-in 25 (50.0) 73 (49.0) Leuprolide acetate 33 (66.0) 113 (75.8) pretreatment Mode of E2 for uterine preparation Oral 21 (95.5) 33 (75.0) Vaginal or 1 (4.6) 11 (25.0) transdermal Mode of luteal progesterone IM 33 (66.0) 101 (67.8) Vaginal 17 (34.0) 48 (32.2) Note: Values represent n (%) or median (interquartile range). Uterine factor infertility was defined as intrauterine synechiae, adenomyosis, uncorrected septum, bicornuate uterus, or submucosal fibroids. For resolved previa, the denominator excludes those with persistent previa. For mode of E2 for uterine preparation, the denominator excludes fresh autologous cycles. Kaser. CET and risk of placenta accreta. Fertil Steril 2015.

Endometrial Thickness, Peak Estradiol, and Risk of Accreta To clarify the relationship between CET and invasive placentation, we explored the interplay between CET, accreta, and 2 cycle-specific factors thought to reside along the causal VOL. - NO. - / - 2015

pathway between CET and accreta—endometrial thickness and peak serum E2. Univariate and ROC analyses were used to identify cutoffs that would discriminate accreta from nonaccreta patients, and CET from fresh cycles. Both the group of patients with placenta accreta and those using CET had significantly thinner endometrial thickness values than their counterparts. The median endometrial thickness for patients with accreta was 8.8 mm (IQR 7.3–11.2), compared with 10.4 mm for those without accreta (IQR 8.4– 12.1; P< .01). Subanalysis of morbid accretas likewise revealed significantly lower values (morbid accreta: 8.8 mm, IQR 7.5–10.7; no accreta: 10.3 mm, IQR 8.4–12.0). The median endometrial thickness for CET cycles was also thinner than that for fresh cycles: 8.4 mm vs. 10.6 mm (P¼ .001). In ROC analysis (Fig. 2A), thresholds that optimized Youden's J-statistic (sensitivity þ specificity  1; Fig. 2B) were selected to dichotomize the data. Endometrial thickness of %9.7 mm was chosen as the optimal threshold and was found to be predictive of accreta with an area under the curve (AUC) of 0.65. Among patients with known values of endometrial thickness, 74% of patients with accreta had an endometrial thickness of %9.7 mm, compared with 45.6% of patients without accreta (OR 4.44, 95% CI 1.85–10.64; P< .001). The sensitivity and specificity were 70% and 61.4%, respectively. The positive predictive value (PPV) of endometrial thickness of %9.7 mm for accreta was 37%, and the negative predictive value (NPV) was 86.2%. The distribution of patients with and without accreta, according to endometrial thickness, is plotted as a histogram in Figure 2C. For all patients with an endometrial thickness < 9 mm, there was an excess risk of accreta (Fig. 2D). The median peak serum E2 was lowest for patients with morbid accreta (732 pg/mL, IQR 367–1,939), followed by all accreta (1,143 pg/mL, IQR 404–1,939), and those without accreta (1,702 pg/mL, IQR 639–2,558; P¼ .02 for morbid accreta vs. no accreta, and P¼ .03 for all accreta vs. no accreta). In addition, the median peak E2 level was significantly lower with CET than with fresh cycles (563 vs. 1,883 pg/mL, P¼ .001). A cutoff point for peak serum E2 that was predictive of accreta was determined by ROC analysis (Fig. 3A). The threshold value that optimized Youden's J-statistic (Fig. 3B) was 732 pg/mL (AUC 0.60; sensitivity 46.9%; specificity 73.8%; PPV 37.1%; NPV 80.9%), which happened to be the median peak E2 for morbid accreta. Applying this threshold to our population, 48% of accreta patients had an E2 %732 pg/mL, compared with 26.2% of patients without accreta (P¼ .004). The distribution of patients with and without accreta, according to serum E2 levels, is shown in Figure 3C. For all patients with a peak serum E2 < 1,000 pg/mL, there was an excess risk of accreta (Fig. 3D). We found that the route of E2 administration played a significant role in the relationship between peak E2 level and 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]; transdermal: 1,033 pg/mL [IQR 398–1,508]; P< .0001). Furthermore, the route of E2 supplementation in programmed CET cycles was significantly associated with accreta: of the 15 patients with accreta who had CET, 14 of 15 (93.3%) were prescribed oral E2, compared with 16 of 26 (61.5%) without 5

ORIGINAL ARTICLE: ASSISTED REPRODUCTIVE TECHNOLOGY

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 patients 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. 6

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. - NO. - / - 2015

Fertility and Sterility®

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 %732 pg/mL (83% vs. 18%, respectively, P< .0001). These findings support the possibility that a thin endometrium and low serum E2 levels may mediate the relationship between CET and invasive placentation. VOL. - NO. - / - 2015

DISCUSSION In this study, we investigated whether CET was a significant predictor of placenta accreta in patients who undergo IVF and/or ICSI, after controlling for other patient and cycle 7

ORIGINAL ARTICLE: ASSISTED REPRODUCTIVE TECHNOLOGY characteristics. We found that CET seems to be a strong independent 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 clinically 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 administration. Patients for whom oral E2 was prescribed had lower serum E2 levels than did those for whom vaginal or transdermal 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 administration in uterine-preparation cycles may be a modifiable risk factor for accreta. One possible mechanism to explain the association between CET and accreta is that the degree of trophoblastic invasion 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 preeclampsia 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-forgestational-age infants [17, 18]), whereas high levels may be associated with placental undergrowth (e.g., preeclampsia and growth-restricted infants) (11). Our ROC analysis suggests that the lower limit of optimal serum E2 may be approximately 700 pg/mL (Fig. 3B); however, given the multimodal nature of this curve, other relevant cutoffs may exist. An upper limit, generated from the 90% percentile of fresh cycles 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, 8

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 Ishishara 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 Ishishara et al. (6) did not specify how placenta accreta was defined, and adjusted for only patient age, stage of embryo development, 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 transferred 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 remarkably 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 degrees of hemorrhage may allow for uterine conservation, the published literature includes clinically identified accretas in VOL. - NO. - / - 2015

Fertility and Sterility® 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 preparation. As serum and uterine E2 levels are known to vary according 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. However, 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 cycles 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 pretreatment 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 preparation 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 multivariate analysis reporting that CET is a strong independent predictor of accreta, after controlling exhaustively for known accreta risk factors and other possible confounders unique to assisted reproductive technology and the infertile population. Possible explanatory mechanisms include a thinner endometrial lining and lower serum E2 level in CET cycles that result in unchecked growth of the extravillous trophoblast 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 fetalgrowth restriction, a ‘‘safety window’’ of serum E2 (700– 3,450 pg/mL) for assisted reproductive technology cycles is VOL. - NO. - / - 2015

thus proposed. External validation of a safe range of serum E2 that minimizes maternal and neonatal morbidities should be a research priority.

REFERENCES 1. 2. 3.

4.

5.

6.

7. 8.

9. 10.

11.

12.

13.

14.

15.

16.

17.

18.

19.

20.

Tantbirojn P, Crum CP, Parast MM. Pathophysiology of placenta creta: the role of decidua and extravillous trophoblast. Placenta 2008;29:639–45. Miller DA, Chollet JA, Goodwin TM. Clinical risk factors for placenta previaplacenta accreta. Am J Obstet Gynecol 1997;177:210–4. 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. Fitzpatrick KE, Sellers S, Spark P, Kurinczuk JJ, Brocklehurst P, Knight M. Incidence and risk factors for placenta accreta/increta/percreta in the UK: a national case-control study. PLoS One 2012;7:e52893. Hayashi M, Nakai A, Satoh S, Matsuda Y. Adverse obstetric and perinatal outcomes of singleton pregnancies may be related to maternal factors associated with infertility rather than the type of assisted reproductive technology procedure used. Fertil Steril 2012;98:922–8. 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. Leibo SP. A one step method for direct nonsurgical transfer of frozenthawed bovine embryos. Theriogenology 1984;21:767–90. 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. Greenland S. Modeling and variable selection in epidemiologic analysis. Am J Public Health 1989;79:340–9. Simon C, Dominguez F, Valbuena D, Pellicer A. The role of estrogen in uterine receptivity and blastocyst implantation. Trends Endocrinol Metab 2003; 14:197–9. 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. 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. 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. Hancke K, More S, Kreienberg R, Weiss JM. Patients undergoing frozenthawed embryo transfer have similar live birth rates in spontaneous and artificial cycles. J Assist Reprod Genet 2012;29:403–7. Niu Z, Feng Y, Sun Y, Zhang A, Zhang H. Estrogen level monitoring in artificial frozen-thawed embryo transfer cycles using step-up regime without pituitary suppression: Is it necessary? J Exp Clin Assist Reprod 2008;5:4. 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 implantation. Proc Natl Acad Sci 2003;100:2963–8. 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. Wennerholm UB, Henningsen AK, Romundstad LB, Bergh C, Pinborg A, Skjaerven R, et al. Perinatal outcomes of children born after frozenthawed embryo transfer: a Nordic cohort study from the CoNARTaS group. Hum Reprod 2013;28:2545–53. Eshkoli T, Weintraub AY, Sergienko R, Sheiner E. Placenta accreta: risk factors, perinatal outcomes, and consequences for subsequent births. Am J Obstet Gynecol 2013;208:219.e1–7. Mercer B, Berghella V, Foley M, Kilpatrick S, Saade G, Grobman W, et al. Placenta accreta. Am J Obstet Gyencol 2010;203:430–9. 9

ORIGINAL ARTICLE: ASSISTED REPRODUCTIVE TECHNOLOGY

SUPPLEMENTAL TABLE 1 Demographic and cycle characteristics of patients having fresh vs. cryopreserved ET. Variable Median age (y) Gravidity 1 R2 Parity R1 Race Caucasian Non-Caucasian Median BMI (kg/m2) Any prior failed ET Uterine factor infertility Fibroids at baseline Prior surgical history Cesarean section >1 prior cesarean section Myomectomy (abdominal or laparoscopic) Dilation and curettage Any operative hysteroscopy Any placenta previa Resolved previa Persistent previa

Fresh ET (n [ 157)

CET (n [ 42)

P value

39.1 (34.9–41.2)

38.1 (33.3–45.5)

.75

107 (68.2) 50 (31.8)

32 (76.2) 10 (23.8)

.31

63 (40.1)

19 (45.2)

.55

129 (82.2) 28 (17.8) 23.7 (21.5–26.5) 86 (54.8) 23 (14.6) 51 (32.5) 43 (27.4) 4 (9.3) 10 (6.4) 78 (49.7) 25 (15.9) 31 (19.8) 19 (13.1) 12 (7.6)

37 (88.1) 5 (11.9) 23.9 (20.8–26.8) 29 (69.0) 8 (19.0) 13 (31.0) 12 (28.6) 1 (8.3) 5 (11.9) 18 (42.2) 9 (21.4) 7 (16.6) 4 (10.3) 3 (7.1)

.36 .73 .11 .48 .85 .85 1.0 .32 .49 .49 .82 .79 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.

9.e1

VOL. - NO. - / - 2015

Fertility and Sterility®

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. Variable Any morbidity Hemorrhage Median blood loss Blood loss >1,000 mL Transfusion Major surgical morbidity Hysterectomy Uterine artery embolization Minor surgical morbidity Intrauterine balloon catheter placement Ascending uterine artery ligation Placental bed oversewing Uterine curettage Manual extraction of placenta Hysteroscopic resection of retained products of conception

Placenta accreta (n [ 50)

No placenta accreta (n [ 149)

Unadjusted OR (95% CI)

P value

34 (68) 21 (42) 1,000 (750–2,000) 20 (40) 15 (30) 7 (14) 4 (8) 4 (8) 24 (48) 2 (4) 4 (8) 7 (14) 12 (24) 7 (14) 2 (4)

16 (11) 13 (8.7) 700 (500–800) 12 (8) 5 (3.4) 1 (0.7) 1 (0.7) 0 (0) 5 (3) 2 (1.3) 0 (0) 0 (0) 2 (1.3) 3 (2) 0 (0)

20.1 (7.1–57.1) 7.5 (3.1–17.8) – 7.3 (3.1–17.5) 9.0 (3.3–24.8) 21.0 (2.6–170.7) 12.0 (1.3–107.4) – 17.5 (6.1–50.6) 1.6 (.4–7.4) – – 18.0 (4.0–80.4) 7.9 (2.0–31.9) –

< .0001 < .001 < .0001 < .001 < .001 .004 .03 .004 < .001 .50 .004 < .001 < .001 < .01 .06

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.

VOL. - NO. - / - 2015

9.e2