Original Research Three-Dimensional Power for Diagnosing Abnormally Invasive and Quantifying the Risk

Sally L. Collins, BMBCh, DPhil, Gordon N. Stevenson, DPhil, Abdulla Al-Khan, MD, Nicholas P. Illsley, DPhil, Lawrence Impey, MBBS, Leigh Pappas, RDMS, RDCS, and Stacy Zamudio, PhD

OBJECTIVE: To test an objective marker for and compared in women subsequently diagnosed with diagnosing the presence and severity of abnormally abnormally invasive placenta and women in a control invasive placenta. group who did not have abnormally invasive placenta. METHODS: Women at risk of abnormally invasive pla- Receiver operating characteristic curves were plotted centa underwent a three-dimensional power Doppler for prediction of abnormally invasive placenta and ultrasound scan. The volumes were examined offline by abnormally invasive placenta requiring cesarean hyster- a blinded observer. The largest area of confluent three- ectomy. dimensional power Doppler signal (Area of Confluence RESULTS: Ninety-three women were recruited. Results 2 [Acon], cm ) at the uteroplacental interface was measured were available for 89. Abnormally invasive placenta was clinically diagnosed in 42 women; 36 required hyster- From the Nuffield Department of & Gynaecology, University of ectomy and had abnormally invasive placenta con- Oxford, and the Fetal Medicine Unit, John Radcliffe Hospital, Oxford, and firmed histopathologically. Median and interquartile the Evelyn Perinatal Imaging Centre, Rosie Hospital, Cambridge University range for A was greater for abnormally invasive pla- Hospitals NHS Foundation Trust, Cambridge, United Kingdom; and the Center con 2 for Abnormal Placentation, Division of Maternal Fetal Medicine and Surgery, centa (44.2 [31.4–61.7] cm ) compared with women in Department of Obstetrics and Gynecology, Hackensack University Medical Cen- the control group (4.5 cm2 [2.9–6.6], P,.001) and even ter, Hackensack, New Jersey. greater in the 36 requiring hysterectomy (46.6 cm2 Funded through the generosity of the Fetal Medicine Unit Charitable Fund in [37.2–72.6], P,.001). Acon rose with histopathologic Oxford and Hackensack University Medical Center. diagnosis: focal accreta (32.2 cm2 [17.2–57.3]), accreta Data from this article were presented at the Society for Reproductive Investigation (59.6 cm2 [40.1–89.9]), and percreta (46.6 cm2 [37.5– 62nd Annual Scientific Meeting, March 25–28, 2015, San Francisco, California, and the British Fetal and Maternal Medicine Society Meeting, April 71.5]; P,.001 analysis of variance for linear trend). 23–24, 2015, London, United Kingdom. Receiver operating characteristic analysis for prediction The authors thank the staff of the Fetal Medicine Unit at the Women’s Centre, of abnormally invasive placenta revealed that with an John Radcliffe Hospital, for all their help with recruiting and supporting the 2 Acon of 12.4 cm or greater, 100% sensitivity (95% con- women involved in the United Kingdom. The authors also acknowledge the following staff at Hackensack University Medical Center: Themba L. Nyirenda fidence interval [CI] 91.6–100) could be obtained with for excellent statistical advice; Ms. Sonia Da Silva Arnold, MS, Ms. Alison Forte, 92% specificity (95% CI 79.6–97.6); area under the curve and Ms. Magda Pawlik, APN, for assisting with obtaining informed consent, is 0.99 (95% CI 0.94–1.0). For prediction of abnormally abstracting of electronic medical record and coordination of patient care; Dr. Jesus Alvarez-Perez, Dr. Manuel Alvarez, and Dr. Yaakov Abdelhak, Maternal-Fetal invasive placenta requiring hysterectomy, 100% sensi- Medicine, the Center for Abnormal Placentation for their excellent surgical care; and tivity (95% CI 90.3–100) can be obtained with an Acon Dr. Elizabeth Kopp, Dr. Judy Gerardis, Dr. Maria Keanchong, Dr. Vinay Shah, of 17.4 cm2 or greater with 87% specificity (95% CI Dr. Andrew Rubenstein, and many of our 100+ community-based obstetrician 74.7–94.5; area under the curve 0.98 [0.93–1.0]). colleagues who supported this study and referred patients with placenta previa or suspected of abnormally invasive placenta from their practices for testing using Acon. CONCLUSION: The marker Acon provides a quantitative Corresponding author: Sally L. Collins, BMBCh, DPhil, The Fetal Medicine means for diagnosing abnormally invasive placenta and Unit, The Women’s Centre, The John Radcliffe Hospital, Headley Way, Head- assessing severity. If further validated, subjectivity could ington, Oxford OX3 9DU, UK; e-mail: [email protected]. be eliminated from the diagnosis of abnormally invasive Financial Disclosure placenta. The authors did not report any potential conflicts of interest. (Obstet Gynecol 2015;126:645–53) © 2015 by The American College of Obstetricians and Gynecologists. Published DOI: 10.1097/AOG.0000000000000962 by Wolters Kluwer Health, Inc. All rights reserved. ISSN: 0029-7844/15 LEVEL OF EVIDENCE: II

VOL. 126, NO. 3, SEPTEMBER 2015 OBSTETRICS & GYNECOLOGY 645

Copyright ª by The American College of Obstetricians and Gynecologists. Published by Wolters Kluwer Health, Inc. Unauthorized reproduction of this article is prohibited. bnormally invasive placentation, or placenta ac- younger than 18 years (United States) and inability to A creta, is a life-threatening obstetric condition in provide informed consent. which failure of placental separation at delivery can The women were managed according to their lead to catastrophic maternal hemorrhage. Prior cesar- local unit protocol. and delivery data were ean delivery and placenta previa are the main risk collected from antenatal records, operative and deliv- factors.1 The incidence of abnormally invasive pla- ery notes, and postpartum records. The presence and centa is increasing with the rise in cesarean delivery severity of abnormally invasive placentation were rates. assessed at delivery according to our clinical grading Maternal mortality and morbidity are reduced system (Table 1) by an attending obstetrician who had when women deliver in a tertiary hospital with seen more than 10 cases of abnormally invasive pla- a multidisciplinary team.2,3 Antenatal diagnosis rests centa and from histopathology results in which hys- on “typical” ultrasonographic findings.4 Magnetic res- terectomy was performed. All histopathology was onance imaging has yet to clearly demonstrate an undertaken by senior pathologists in each unit who improvement in pregnancy outcomes.5 Irrespective have special expertise in placentology. of the imaging modality, diagnosis is subjective with Patients underwent diagnostic imaging according accuracy depending on the experience of the opera- to the local unit protocol. In addition, static, trans- tor, which is limited by the rarity of the condition. abdominal three-dimensional power Doppler ultra- Recent advances in three-dimensional power Doppler sound volumes of the placental bed were obtained imaging offered a new marker of “numerous coherent according to a predefined protocol with the partici- vessels.”6 However, it too remains subjective and pant in a semirecumbent and a full bladder experience-dependent. using a RAB4-8-D 3D/4D curved array abdominal Although sensitivity is crucial, so too is the transducer (4–8.5 MHz) on a Voluson E8. Predeter- specificity of prenatal diagnosis. The price of a false- mined machine settings were used (for details of all positive diagnosis is high. A vertical laparotomy settings, see Collins et al13). To allow for differences in proceeding straight to hysterectomy is frequently used attenuation of the power Doppler signal resulting when an abnormally invasive placenta is anticipated.7 from variation in placental site and maternal adipos- Prophylactic occlusive balloons in the pelvic vascula- ity, the subnoise gain setting was selected according to ture risk significant complications.8–10 The American a previously validated technique.14,15 The participant College of Obstetricians and Gynecologists currently was asked to remain as still as possible, and volumes states that combined maternal and neonatal outcome were excluded if any “flash artefact” (secondary to is optimized in stable patients with a planned delivery fetal or transducer movement) was present. Initially at 34 weeks of gestation, resulting in iatrogenic neo- a standard 85° volume was collected in the sagittal natal morbidity.3,5,11,12 plane, midline under the uterovesical fold (presumed The aim of this study was to investigate whether site of previous cesarean delivery scar and therefore our novel ultrasound marker of largest area of most likely site of any abnormally invasive placenta). confluent three-dimensional power Doppler signal Subsequent volumes were then systematically col- (Acon) can accurately predict both the presence and lected until the entire basal plate of the placenta had severity of abnormally invasive placenta. been imaged, usually three to five volumes. The data were then stored anonymously using Sonoview and MATERIALS AND METHODS analyzed after delivery using the software package 4D Women were recruited from two referral centers: the View. Fetal Medicine Unit, John Radcliffe Hospital (Oxford, All the static volumes were opened in 4D View United Kingdom) and the Center for Abnormal and examined by one operator (S.L.C.). All images Placentation at Hackensack University Medical Cen- were coded (deidentified) and the analysis was con- ter (New Jersey). Patients were referred by their ducted blinded to pregnancy outcome and previous primary health care provider if they suspected abnor- imaging findings. The three-dimensional volumes mally invasive placentation: diagnosis of a placenta containing the uteroplacental interface were trans- previa and a history of previous uterine surgery or formed until only the power Doppler signal around ultrasound findings suggestive of abnormally invasive the uteroplacental interface remained. This was per- placenta on routine ultrasound scan. Written informed formed according to the process shown in Figure 1. A consent was obtained with local research ethics three-dimensional process is difficult to represent in approval. Exclusion criteria were multiple gestation, two dimensions; therefore, the images in Figure 1 have age younger than 16 years (United Kingdom) or been adjusted to make the concept clearer. The settings

646 Collins et al Diagnosing Abnormally Invasive Placenta OBSTETRICS & GYNECOLOGY

Copyright ª by The American College of Obstetricians and Gynecologists. Published by Wolters Kluwer Health, Inc. Unauthorized reproduction of this article is prohibited. Table 1. Clinical Grading System Used to Assess the Severity of the Abnormally Invasive Placenta

Grade Definition

1 At cesarean or vaginal delivery: complete placental separation at third stage; not abnormally invasive placenta 2 At cesarean delivery or laparotomy: no placental tissue seen invaded through the serosal surface of the uterus; only partial separation with synthetic oxytocin and gentle controlled cord traction, manual removal of placenta required for remaining tissue AND parts of placenta thought to be abnormally adherent by a senior, experienced clinician At vaginal delivery; manual removal of placenta required AND parts of placenta thought to be abnormally adherent by a senior, experienced clinician 3 At cesarean delivery or laparotomy: no placental tissue seen invaded through the serosal surface of the uterus; no signs of any separation with synthetic oxytocin and gentle controlled cord traction, manual removal of placenta required AND the whole placental bed thought to be abnormally adherent by a senior, experienced clinician At vaginal delivery: manual removal of placenta required AND the whole placental bed thought to be abnormally adherent by a senior, experienced clinician 4 At cesarean delivery or laparotomy: placental tissue seen to have invaded through the serosal surface of the uterus but NOT passing into any surrounding structures (including the posterior wall of the urinary bladder); a clear surgical plane can be identified between the bladder and uterus to allow nontraumatic reflection of the urinary bladder at hysterectomy 5 At cesarean delivery or laparotomy: placental tissue seen to have invaded through the serosal surface of the uterus AND invaded into the urinary bladder ONLY (consequently, a clear surgical plane cannot be identified between the bladder and uterus to allow nontraumatic reflection of the urinary bladder at hysterectomy) 6 At cesarean delivery or laparotomy: placental tissue seen to have invaded through the serosal surface of the uterus AND invaded into the pelvic side wall or any organ other than the urinary bladder, with or without invasion into the urinary bladder For the purposes of this scale, “uterus” includes the uterine body and uterine cervix. were identical for all the patients including power Ten volumes were selected to be included in the Doppler threshold and image magnification. The intra- and interobserver study, five were randomly three-dimensional power Doppler signal was overlaid taken from the confirmed abnormally invasive over the three-dimensional B mode volume (Fig. 1A placenta group and five from the not abnormally and B). The volume was manipulated in three dimen- invasive placenta group. Acon was estimated by two sions using the 4D View software and the power Dopp- observers (S.L.C. and G.N.S.) on three different oc- ler signal was sequentially reduced using the software’s casions blinded to their previous estimates and each “magi-cut,” color-only function. Initially the power other’sresults.Thetwo-waymixedintraclasscorre- Doppler signal representing the fetal circulation was lation coefficients were calculated to assess intra- and removed (Fig. 1C), then the remaining power Doppler interobserver variability. signal was removed until a 1-cm deep volume was left: Analyses of maternal and neonatal demographic 0.5 cm either side of the uteroplacental interface (Fig. characteristics were performed by Student’s t test or 2 1D). The volume was then reorientated until the plane x as appropriate. Acon was not normally distributed of the uteroplacental interface containing the power (Kolmogorov-Smirnov P,.001), being right-skewed Doppler signal of interest was perpendicular to the in both women in the case group and those in the eyeline of the observer (Fig. 1E). The vascularity at control group. Therefore, comparison of Acon the uteroplacental interface is thus viewed as if the between groups was conducted by nonparametric observer was looking down through the flat placental Mann-Whitney U and presented as box plots encom- bed from the myometrium toward the . If the passing the 25th to 75th centiles and whiskers indicat- starting volume is initially in midsagittal view, this is ing fifth to 95th centiles. Acon data analyzed by achieved by rotating by 90° around the y-axis. The B diagnostic categories, whether histopathologic or by mode was removed from view to improve visualiza- clinical grade, were log-transformed and an analysis of tion of the power Doppler signal. The largest area of variance of the linear trend between means was uti- confluent power Doppler signal (Area of Confluence lized. Regression analysis was used to examine the 2 [Acon], cm )wasthenquantifiedusingthe“generic relationship between the log-transformed Acon and area” measurement function provided by 4D View at the time of the scan. The receiver (Fig. 1E). operating characteristic curves were plotted for the

VOL. 126, NO. 3, SEPTEMBER 2015 Collins et al Diagnosing Abnormally Invasive Placenta 647

Copyright ª by The American College of Obstetricians and Gynecologists. Published by Wolters Kluwer Health, Inc. Unauthorized reproduction of this article is prohibited. Uterine serosal surface Myometrium

Maternal circulation Uteroplacental interface Placenta Fetal circulation removed Fetus Fetal circulation Umbilical cord A B C

All power Doppler signal removed except for within 1 cm of the uteroplacental interface

Rotate by 90˚ D around the x-axis. E Fig. 1. Two-dimensional representation of the three-dimensional method used to generate area of confluence. A. The volume in B mode. Lines have been drawn in on the two-dimensional image to demonstrate the uteroplacental interface and the serosal surface of the uterus. B. The B mode volume with the three-dimensional power Doppler signal overlaid. C. The power Doppler signal after the fetal circulation has been removed. D. All power Doppler signal removed except within 1 cm of the uteroplacental interface. E. The power Doppler signal with the flattest plane of the uteroplacental interface at 90˚ to the eyeline of the observer, the B mode signal removed, and the largest area of power Doppler confluence estimated (yellow outline). Collins. Diagnosing Abnormally Invasive Placenta. Obstet Gynecol 2015.

prediction of an abnormally invasive placenta by Acon unresponsive to conservative management, massive (all abnormally invasive placenta, n542) and the pres- leiomyomas, and uterine–bowel adhesions. Subse- ence of abnormally invasive placenta resulting in quent histopathology confirmed no abnormally cesarean hysterectomy (confirmed histopathologi- invasive placenta. Forty-two women had clinical cally, n536) and area under the curve was calculated evidence of abnormally invasive placenta at deliv- with the exact binomial method used to calculate the ery, 36 of whom required hysterectomy. All 36 were 95% confidence intervals (CIs). We report our data histopathologically confirmed as abnormally inva- with sensitivity set, a priori, at 100%. Our goal in sive placenta (five focal accreta, nine accreta, and developing an objective screening test is to provide 22 percreta). All the women with abnormally inva- assurance with the highest possible accuracy that sive placenta had at least one previous cesarean abnormally invasive placenta is not present. Statistical delivery except for two who had complex histories analyses were performed using SPSS 19.0, and results of multiple uterine surgeries. The median gestation were graphed using Graph Pad Prism. Results were for all included scans was 32 weeks. There was no considered to be significant when P,.05. difference in the gestational age at ultrasonography for the women diagnosed with abnormally invasive RESULTS placenta (median 32 1/7 weeks of gestation, range Ninety-three women were recruited over 2 years 17 1/7–33 6/7 weeks of gestation) and those without (April 2012 to March 2014); two were lost to abnormally invasive placenta (32 4/7 weeks of ges- follow-up, one was excluded for an incomplete scan tation, range 15 2/7239 2/7 weeks of gestation). of the uteroplacental interface, and one was There were no differences in the body mass excluded because the woman was unable to fill index, ethnicity, or parity of women with abnormally her bladder sufficiently to adequately visualize the invasive placenta compared with those without uterovesical interface. Images and delivery data abnormally invasive placenta (Table 2). The group were available for 89 women (Table 2). Forty- diagnosed with abnormally invasive placenta had seven women had no evidence of abnormally inva- more total previa (P,.001), underwent more hyster- sive placenta at delivery. Three of these women had ectomies (P,.001), and were delivered earlier ahysterectomyatthetimeforuterineatony (P,.01).

648 Collins et al Diagnosing Abnormally Invasive Placenta OBSTETRICS & GYNECOLOGY

Copyright ª by The American College of Obstetricians and Gynecologists. Published by Wolters Kluwer Health, Inc. Unauthorized reproduction of this article is prohibited. Table 2. Maternal Characteristics and Abnormally Invasive Placenta Risk Factors

Not Abnormally Invasive Placenta Abnormally Invasive Placenta Maternal Demographics (n547) (n542) P

Prepregnancy BMI (kg/m2) 26.765.4 (18.1–43.0) 26.165.3 (17.9–43.4) .66 Ethnicity White (non-Hispanic) 24 (50) 22 (53) .80 Hispanic 9 (19) 11 (26) 1.00 Black 4 (9) 4 (9) .70 Asian 3 (7) 4 (9) .20 Other (including mixed race) 7 (15) 1 (2) .06 Parity 0 5 (11) 2 (5) .40 1 22 (47) 15 (36) .40 2 12 (26) 10 (24) .60 3 or greater 8 (17) 15 (36) .05 Gestational age (wk) At delivery 37 1/7 (28 2/7–42) 34 3/7 (21 5/7–26 4/7) ,.01 At ultrasound scan 32 1/7 (17 1/7–33 6/7) 32 4/7 (15 2/7–39 2/7) .80 Hysterectomy performed 3 (6) 36 (86) ,.001 Admission to ICU 0 3 (7) .10 Abnormally invasive placenta risk factors Prior cesarean delivery (0–8) 0 10 (21) 2 (5) .03 1 22 (47) 21 (50) .80 2 9 (19) 9 (21) 1.00 3 or more 6 (13) 10 (24) .30 Placenta previa Major placenta previa (placenta over os) 17 (36) 32 (76) ,.001 Marginal placenta previa (2 cm or less from 7 (15) 7 (17) 1.00 os) Cesarean delivery scar and major 11 (23) 30 (71) ,.001 placenta previa Previous uterine surgery Evacuation of retained products of 9 (19) 13 (31) .23 conception Myomectomy 3 (6) 2 (5) 1.0 Transcervical resection of fibroids 1 (2) 1 (2) 1.0 Endometrial ablation 0 1 (2) .10 Assisted reproductive technology 1 (2) 3 (7) .30 BMI, body mass index; ICU, intensive care unit. Data are mean6standard deviation (range), n (%), or median (range) unless otherwise specified.

The intraclass correlation coefficient for the diagnosed with abnormally invasive placenta but values of Acon estimated by the two different observ- managed by placental removal without hysterec- ers (interobserver) was 0.92 (95% CI 0.85–0.97). The tomy (28.4 [21.9–33.6], P,.005; Fig. 2B). Acon rose intraclass correlation coefficients for three repeated with more severe histopathologic diagnosis focal estimates by each observer (intraobserver) were 0.94 accreta (32.2 cm2 [17.2–57.3]), accreta (59.6 cm2 (G.N.S.: 0.89–0.99) and 0.95 (S.L.C.: 0.90–1.0). [40.1–89.9]), and percreta (n522, 46.6 cm2 [37.5– Acon values were greater for the group clinically 71.5], P,.001, analysis of variance test for linear diagnosed with abnormally invasive placenta trend between means). When the histopathologic (n542, median 44.2, 25th–75th centile 31.4–61.7) diagnosis was compared between the 36 with abnor- than those without abnormally invasive placenta mally invasive placenta and the three without (4.5 [2.9–6.6], Mann-Whitney U, P,.001; Fig. 2A). abnormally invasive placenta who underwent hys- Acon values were even greater in the subset of the terectomy, Acon was greater for any form of accreta abnormally invasive placenta group whose pathol- (P,.001, Kruskal-Wallis with Dunn’smultiplecom- ogy was severe enough to require hysterectomy parison; Fig. 2C). Likewise, when Acon was plotted (46.6 [37.2–72.6]) compared with those clinically against the clinical grade assigned to each case, all

VOL. 126, NO. 3, SEPTEMBER 2015 Collins et al Diagnosing Abnormally Invasive Placenta 649

Copyright ª by The American College of Obstetricians and Gynecologists. Published by Wolters Kluwer Health, Inc. Unauthorized reproduction of this article is prohibited. Fig. 2. The median and 95% confidence intervals of the area of confluence (cm2) for the following comparisons. A. The clinical diagnosis of not abnormally invasive placenta (AIP) (n547, range 0.5–29.7 cm2) compared with AIP (n542, range 12.7–150.4 cm2, P,.001). B. Left two box plots illustrate patients without AIP who did not have a hysterec- tomy (no, n544, range 0.5–29.7 cm2)or who did have a hysterectomy (yes, n53, range: 2.2–9.4 cm2, P5nonsignificant). Right two box plots illustrate patients with AIP who did not have a hysterectomy (no, n56, range 12.7–33.6 cm2)orwhodid have a hysterectomy (yes, n536, range 18.7–150.4 cm2, P,.005). C.Patients histopathologically confirmed as not AIP (n53), focal accreta (n55, range 18.7– 61.1 cm2), accreta (n59, range 23.9– 102.6 cm2), or percreta (n522, range 24.9–150.4 cm2, P,.001). D.Clinical grade was greater for all grades relative to control (P,.001). Grade 2 (n512, median 32.2 cm2,range18.7–61.1); grade 3 (n56, median 54.7 cm2,range 12.7–89.3); grade 4 (n515, median 45.7 cm2,range24.9–102.6);grade5(n58, median 59.6 cm2,range43.0–150.4);andgrade 6(n52, median 126.6 cm2,range113.3–141.8).*SeeTable1fordefinitionofclinicalgrades. Collins. Diagnosing Abnormally Invasive Placenta. Obstet Gynecol 2015.

grades had an elevated Acon relative to control DISCUSSION (P,.001, analysis of variance test for linear trend Abnormally invasive placenta is a dangerous pathol- between means; Fig. 2D). There was no correlation ogy; it is difficult to manage clinically and is increasing between Acon and gestational age at the time of the worldwide.16 Our results suggest that a quantitative 2 ultrasound scan (r for abnormally invasive measure, designated here as largest Area of Confluence placenta50.12; for not abnormally invasive pla- (A ), can differentiate between the presence and 25 con centa r 0.08). absence of abnormally invasive placenta, and is associ- All 42 abnormally invasive placenta cases had ated with the histopathologic and clinical severity of A values of 12.4 cm2 or greater, whereas four of 47 con abnormally invasive placenta. That 100% sensitivity women in the control group had values greater than was achieved with an 8% false-positive rate suggests 12.4 cm2 (false-positives, range 12.8–29.7). The that this technique could allow reliable prenatal diag- receiver operating characteristic curve for prediction nosis. This would enable adequate preparation and of a clinical diagnosis of abnormally invasive placenta (all grades) shows an area under the curve of 0.99 intervention for cases of abnormally invasive placenta while minimizing the iatrogenic adverse events. How- (0.94–1.0; Fig. 3A; Table 3). With an Acon of 12.4 cm2 or greater, 100% sensitivity could be achieved ever, care must be taken when interpreting this false- for the prediction of all clinically diagnosed abnormally positive rate, because our sample was taken from invasive placenta with a 92% specificity and a positive a population with a high prevalence, women who were predictive value of 92% (8% false-positive rate). Of already considered to be at risk of abnormally invasive greater clinical relevance, the area under the curve placenta. It would be inappropriate for this test to be for abnormally invasive placenta requiring hysterec- used as a screening tool on a low-risk population as tomy compared with all other cases (including a result of the risk of the false-positive paradox (the clinically diagnosed abnormally invasive placenta man- false-positive rate being greater than the true-positive aged conservatively) was 0.98 (0.93–1.0). Therefore, an rate as a result of the rarity of the condition).17 2 Acon of 17.4 cm or greater yielded 100% sensitivity The observed increase in Acon with increasing with 87% specificity and a 13% false-positive rate. severity, both clinical and histopathologic, suggests

650 Collins et al Diagnosing Abnormally Invasive Placenta OBSTETRICS & GYNECOLOGY

Copyright ª by The American College of Obstetricians and Gynecologists. Published by Wolters Kluwer Health, Inc. Unauthorized reproduction of this article is prohibited. Fig. 3. Receiver operator curves for the prediction of abnormally invasive pla- centa using area of confluence. A. Receiver operating characteristic curve for any clinically diagnosed abnormally invasive placenta (all grades) relative to controls. Area under the curve is 0.99, P,.001. B. Receiver operator character- istic curve for any histopathologically confirmed abnormally invasive placenta significant enough to warrant hysterec- tomy relative to controls (not abnormally invasive placenta and clinically diag- nosed abnormally invasive placenta managed with manual removal of pla- centa). Area under the curve is 0.98, P,.001. Collins. Diagnosing Abnormally Invasive Pla- centa. Obstet Gynecol 2015.

that the numerical value of Acon may be able to pre- Hundred percentage sensitivity was achieved dict not only the presence of abnormally invasive pla- with an 8% false-positive rate. Although in many centa but also the degree of clinical risk. In the screening situations this would be too high, it is absence of histopathologic analysis, a clinical grading probably acceptable if it ensures no cases are missed system (Table 1) is required to compare severity of in a condition in which antenatal diagnosis has been outcomes between centers. Without this we cannot shown to significantly decrease maternal morbidity.2,3 compare the efficacy of diagnostic assessments. Most The technique itself also has weaknesses and studies have not separated their cases into accreta, unknown factors. The manual process currently used increta, or percreta and some have minimal histopath- to obtain Acon using 4D View requires considerable ologic data with which to confirm a subjective diag- practice and must be undertaken by operators both nosis, despite large numbers of patients.18 The use of experienced in placental imaging and manipulating a standardized clinical grading system would allow three-dimensional power Doppler images in 4D comparison both in research and clinical practice. View. Although we have demonstrated high repro- The strength of the study is the encouraging ducibility, both operators were not only experienced results despite the relatively small sample size. We in placental imaging, but were also instrumental in acknowledge several weaknesses. This was a “proof- designing the technique. This operator dependence of-concept” pilot study, which combined the values can be decreased by use of existing image analysis for Acon gathered at a variety of gestations ranging tools, which reliably map the vasculature and mor- from 19 to 39 weeks. Although there was no strong phometry of the placenta.19–21 We have already devel- correlation between the value of Acon and gestational oped a semiautomated technique to analyze the age, the latter is still likely an important factor. Eval- placenta, uteroplacental interface, and its vasculature 21,22 uation of Acon with larger numbers of at risk women in the first trimester. These techniques could be and over an extended gestational age range is applied to facilitate semiautomated estimation of Acon required to further evaluate the test and to determine thereby simplifying the process and decreasing oper- gestation-appropriate values. ator dependence.

Table 3. Receiver Operating Characteristic Curves for Prediction of Any Abnormally Invasive Placenta and Abnormally Invasive Placenta Requiring Cesarean Hysterectomy

Area of Confluence Abnormally Invasive Placenta Threshold Value (cm2) Sensitivity (95% CI) Specificity (95% CI) AUC (95% CI)

Any 12.4 100% (91.6–100) 91.5% (79.6–97.6) 0.99 (0.94–1.00) Requiring cesarean hysterectomy 17.4 100% (90.3–100) 86.8% (74.7–94.5) 0.98 (0.93–1.00) CI, confidence interval; AUC, area under the curve.

VOL. 126, NO. 3, SEPTEMBER 2015 Collins et al Diagnosing Abnormally Invasive Placenta 651

Copyright ª by The American College of Obstetricians and Gynecologists. Published by Wolters Kluwer Health, Inc. Unauthorized reproduction of this article is prohibited. Little is known of the process underlying the 5. Publications Committee, Society for Maternal-Fetal Medicine; development of the area of confluence. One theory is Belfort MA. Placenta accreta. Am J Obstet Gynecol 2010;203: 430–9. that it represents enlarged arteriovenous anastomoses 23 6. Shih JC, Palacios Jaraquemada JM, Su YN, Shyu MK, Lin CH, located in the subplacental myometrium, generating Lin SY, et al. Role of three-dimensional power Doppler in the a large circulating pool, which can be detected as an antenatal diagnosis of placenta accreta: comparison with gray- abnormal area of power Doppler signal in abnormally scale and color Doppler techniques. Ultrasound Obstet Gyne- col 2009;33:193–203. invasive . This agrees with the findings of 7. Dildy GA III. Postpartum hemorrhage: new management op- Tantbirojn et al, who showed trophoblast-induced re- tions. Clin Obstet Gynecol 2002;45:330–44. modeling deep in the myometrium.24 How A de- con 8. Dilauro MD, Dason S, Athreya S. Prophylactic balloon velops over time remains unclear, but serial data occlusion of internal iliac arteries in women with placenta across gestation may reveal not only the natural his- accreta: literature review and analysis. Clin Radiol 2012;67: tory, but whether abnormally invasive placenta pro- 515–20. gresses, for example, from an accreta to percreta. This 9. Teare J, Evans E, Belli A, Wendler R. Sciatic nerve ischaemia after iliac artery occlusion balloon catheter placement for pla- is not trivial; current American College of Obstetri- centa percreta. Int J Obstet Anesth 2014;23:178–81. cians and Gynecologists’ recommendations suggest 12 10. Gagnon J, Boucher L, Kaufman I, Brown R, Moore A. Iliac delivery at 34 weeks of gestation. Although our data artery rupture related to balloon insertion for placenta accreta do not support a strong correlation between the size of causing maternal hemorrhage and neonatal compromise. Can J Anaesth 2013;60:1212–7. Acon and gestational age, the majority of our data was collected in the late second and early third trimester. 11. De Luca R, Boulvain M, Irion O, Berner M, Pfister RE. Incidence of early neonatal mortality and morbidity after late-preterm and With a larger sample size, CIs for Acon can be deter- term cesarean delivery. Pediatrics 2009;123:e1064–71. mined across gestation and in all grades of abnormally 12. Placenta accreta. Committee Opinion No. 529. American Col- invasive placenta, thereby enabling establishment of lege of Obstetricians and Gynecologists. Obstet Gynecol 2012; clinically useful cutoff values that could reliably pre- 120:207–11. dict not only the presence of abnormally invasive pla- 13. Collins SL, Stevenson GN, Noble JA, Impey L. Developmental centa, but also its likely clinical severity. changes in spiral artery blood flow in the human placenta observed with colour Doppler ultrasonography. Placenta In summary we report a new, quantitative ultra- 2012;33:782–7. sound technique enabling reliable prenatal identifica- 14. Collins SL, Stevenson GN, Noble JA, Impey L, Welsh AW. tion of abnormally invasive placenta, a condition in Influence of power Doppler gain setting on Virtual Organ which prenatal diagnosis has been shown to reduce Computer-aided AnaLysis indices in vivo: can use of the indi- 25 vidual sub-noise gain level optimize information? Ultrasound the high maternal mortality and morbidity. The Obstet Gynecol 2012;40:75–80. application of existing image analysis tools to auto- 15. Sanderson J, Wu L, Mahajan A, Meriki N, Henry A, mate calculation of Acon could eliminate operator Welsh AW. Selection of the sub-noise gain level for acquisition dependence and permit its use for screening outside of VOCAL data sets: a reliability study. Ultrasound Med Biol of specialized centers. Larger, longitudinal studies will 2014;40:562–7. allow exploration of the pathology of excess vascular- 16. Khong TY. The pathology of placenta accreta, a worldwide epidemic. J Clin Pathol 2008;61:1243–6. ity in abnormally invasive placenta, the optimal ges- 17. Vacher HL. Quantitative literacy—drug testing, cancer screen- tation for diagnosis, and yield further information on ing, and the identification of igneous rocks. J Geosci Edu 2003; the technique’s performance. 51:2. 18. Palacios Jaraquemada JM, Bruno CH. Magnetic resonance imaging in 300 cases of placenta accreta: surgical correla- REFERENCES tion of new findings. Acta Obstet Gynecol Scand 2005;84: 1. Rosenberg T, Pariente G, Sergienko R, Wiznitzer A, Sheiner E. 716–24. Critical analysis of risk factors and outcome of placenta previa. 19. Stevenson GN, Collins SL, Impey L, Noble A. OP10.09: Arch Gynecol Obstet 2011;284:47–51. a novel semi-automated (SA) technique for 3D ultrasound mea- 2. Eller AG, Bennett MA, Sharshiner M, Masheter C, surement of placental volume. Ultrasound Obstet Gynecol Soisson AP, Dodson M, et al. Maternal morbidity in cases 2010;36:82. of placenta accreta managed by a multidisciplinary care team 20. Stevenson GN, Collins SL, Impey L, Noble JA. Surface param- compared with standard obstetric care. Obstet Gynecol 2011; eterisation of the utero-placental interface using 3D power 117:331–7. Doppler ultrasound. In: Biomedical imaging: from nano to 3. Al-Khan A, Gupta V, Illsley NP, Mannion C, Koenig C, macro, 2011 IEEE International Symposium on Biomedical Bogomol A, et al. Maternal and fetal outcomes in placenta Engineering; 2011 Mar 30–Apr 2; Chicago, IL. New York accreta after institution of team-managed care. Reprod Sci (NY): IEEE; 2011. p. 891–4. 2014;21:761–71. 21. Stevenson GN, Collins SL, Welsh AW, Impey LW, Noble JA. 4. Wong HS, Cheung YK, Zuccollo J, Tait J, Pringle KC. Evalu- A technique for the estimation of fractional moving blood vol- ation of sonographic diagnostic criteria for placenta accreta. ume by using three-dimensional power Doppler US. J Clin Ultrasound 2008;36:551–9. 2015;274:230–7.

652 Collins et al Diagnosing Abnormally Invasive Placenta OBSTETRICS & GYNECOLOGY

Copyright ª by The American College of Obstetricians and Gynecologists. Published by Wolters Kluwer Health, Inc. Unauthorized reproduction of this article is prohibited. 22. Collins SL, Stevenson GN, Noble JA, Impey L. Elsevier Tro- 24. Tantbirojn P, Crum CP, Parast MM. Pathophysiology of pla- phoblast Research Award Lecture: Searching for an early preg- centa creta: the role of decidua and extravillous trophoblast. nancy 3-D morphometric ultrasound marker to predict fetal Placenta 2008;29:639–45. – growth restriction. Placenta 2013;34(suppl):S85 9. 25. Solheim KN, Esakoff TF, Little SE, Cheng YW, Sparks TN, 23. Schaaps JP, Tsatsaris V, Goffin F, Brichant JF, Delbecque K, Caughey AB. The effect of cesarean delivery rates on the Tebache M, et al. Shunting the intervillous space: new concepts future incidence of placenta previa, placenta accreta, and in human uteroplacental vascularization. Am J Obstet Gynecol maternal mortality. J Matern Fetal Neonatal Med 2011;24: 2005;192:323–32. 1341–6.

Earn CME Credits for Your Contribution as a Reviewer to Obstetrics & Gynecology

In recognition of their time, effort, and expertise, reviewers of manuscripts for Obstetrics & Gynecology are eligible to receive continuing medical education credits.

ACCME Accreditation The American College of Obstetricians and Gynecologists is accredited by the Accreditation Council for Continuing Medical Education (ACCME) to provide continuing medical education for physicians.

AMA PRA Category 1 Credit(s)™ The American College of Obstetricians and Gynecologists designates this manuscript review activity for a maximum of 3 AMA PRA Category 1 Credits.™ Physicians should claim only the credit commensurate with the extent of their participation in the activity.

College Cognate Credit(s) The American College of Obstetricians and Gynecologists designates this manuscript review activity for a maximum of 3 Category 1 College Cognate Credits. The College has a reciprocity agreement with the AMA that allows AMA PRA Category 1 Credits™ to be equivalent to College Cognate Credits.

College Fellows will be awarded credit for a maximum of five reviews yearly. Those who are not College Fellows will receive e-mail documentation approximately 1 month after completion of the review, which can be submitted to an accrediting body for credits.

rev 11/2014

VOL. 126, NO. 3, SEPTEMBER 2015 Collins et al Diagnosing Abnormally Invasive Placenta 653

Copyright ª by The American College of Obstetricians and Gynecologists. Published by Wolters Kluwer Health, Inc. Unauthorized reproduction of this article is prohibited.