Introducing an efficient model for the prediction of placenta accreta spectrum using the MCP regression approach based on sonography indexes: how efficient is sonography in diagnosing accreta?

This is a prospective cross-sectional study performed in Hafez Hospital affiliated to Shiraz University of Medical Sciences, Shiraz, Iran. In this study all women with a history of at least one previous Cesarean section, during March 20th 2016 to February 19th 2017, were considered for inclusion in the study. Among these, all women in their second and third trimester of pregnancy who referred to the perinatology ward of the medical care center for the evaluation of site of placenta and placental adhesion, were included in the study. All women who referred during their first trimester were asked to refer during their second trimester as well.

Individuals without a history of Cesarean sections, and those who did not refer for their follow-ups were excluded from the study. Figure 1 shows the flow diagram related to patients’ recruitment.

Every individual underwent systematic two-dimensional grey-scale US imaging (GE Voluson R 730, GE medical systems, Zipf, Austria). Initially, site of placenta was evaluated using abdominal transducers. In cases of suspicion of low lying placenta or placenta previa, trans-vaginal sonography was done, during which patients had an empty bladder and the distance between placental edge and the internal cervical orifice was measured. The term “placenta previa” was used when the placental edge overlaped or was within 2 cm of the internal cervical orifice in late pregnancy, and the term “low-lying” was used if the placental edge was located further than 2 cm but within 3.5 cm from the internal cervical orifice. In these cases, trans-vaginal US were repeated, during which the bladder was filled with 300 ml of liquid for better evaluation of the uterine serosa-bladder interfaces. Accordingly, participants were categorized in the “high risk” group for PAS if the placenta was previa or low-lying.

For placentas that contained abnormal vessels or lacunae structures, color Doppler sonography was done in order to map the vascularization of the intraplacental and uterine serosa-bladder interface and to measure the velocity flow of the inter-lacunae turbulant.

Patients classified as high risk for PAS were further re-evaluated just before planned delivery.

In each of the US and Doppler examinations the following parameters were examined: 1) presence of multiple placental lacunae. These irregularly shaped hypoechoic vascular spaces larger than 1*1 cm in size within the placental parenchyma, may give the placenta a “Swiss cheese” appearance, often containing turbulent flow that is visible on gray scale imaging. According to Finberg’s grading scale for lacunae [15], the structures were categorized as followed: grade 1+ for one to three lacunae structures, which are usually small, grade 2+ for four to six lacunae structures, which are typically larger, and grade 3+ which shows diffuse lacunae throughout the placenta, 2) losing or irregularity in the echolucent area between the placenta and the myometrium (clear space), 3) thinning or interruption of the hyperechoic interface between the uterine serosa and bladder wall (bladder line), 4) presence of an exophytic mass caused by intrusion of the placental tissue into the bladder, 5) sub-placental hypervascularity, 6) utero-vesical vascularity, 7) bridging vessels that extended from the placenta into the bladder wall via the myometrium. In case of existence of feeding vessels into the lacunae, velocity of blood flow within the lacunae structures was also measured.

Sonography imaging was done mainly during the third trimester, except in women who presented with history of rupture of membranes or vaginal bleeding or any emergency condition that would place that individual at risk of termination of pregnancy, during which US was done at any gestational age. In cases with multiple US examinations, the last scan before delivery was considered for study.

All known and suspected cases of PAS were admitted to the high risk pregnancy department at 33–34 weeks of gestation and a formal multidisciplinary management and treatment program was initiated for them by a trained team [16].

According to our protocol, surgery was done with a surgical oncology team, general surgery team, and a urology team stationed at the high risk center, in which blood products were easily accessible and facilities such as postoperative ICU care was available. Moreover, a neonatal care department and a neonatal team were also present at the center. Surgery was initiated with adequate hydration and among patients who were suspected of adhesion, according to preoperative sonography, a fundal incision was made. With a fundal hysterotomy the child was delivered. The umbilical cord was closed at the nearest site to the placenta. If the patient was hemodynamically stable, without any effort to separate the placenta, a quick evaluation was done in 1–2 min. If the placenta had no invasion to the outer wall of the uterus and there was no bulging or bridging vessels existing between the uterus and the intestines or the bladder, moreover the patient remained hemodynamically stable, if the placenta had completely separated during this time, a standard typical Cesarean section would be performed. If the placenta was separated from all regions and was only attached in a few centimeters from the middle, if the surgery team would determine that by ligation of vessels or even removal of a few centimeters of the uterus, they are able to separate the placenta and save the uterus, then this approach was done. In cases in which during after birth, the patient was hemodynamically unstable, hemodynamic resuscitation was initiated and hysterectomy was done. Moreover, in cases where, the patient was hemodynamically stable, however the placenta had invaded the outer wall of the uterus or in cases where dilated or bridging vessels were seen between the uterus and the bladder or intestines, after birth and clamping of umbilical vessels, hysterectomy was done.

There was no attempt to remove the placenta manually.

For the purpose of this study, invasive placentation was suspected based on clinical assessment of abnormal adherence and evidence of gross placental invasion at time of surgery, after which histopathological study of the excised uterus confirmed the diagnosis of PAS.

All participant had declined conservative treatment (preserving the uterus with the placenta left in situ) or didn’t fulfill criteria for this treatment approach.

The uterus and the attached placenta were first weighed and measured. The external surface of the uterus, especially the anterior and lower segments, was evaluated for any hematoma, ruptured and bulging areas. Areas suspicious of percreta were inked and then the uterus was bivalved using a knife. The specimens were fixed in 10% neutral-buffered formalin overnight. Serial bread-loaf sections of the uterus were checked for areas of increta/percreta. Multiple sections were taken from suspicious areas of accreta and myometrial invasion. Moreover, two sections from the cervix or lower uterine segment (in supra-cervical hysterectomy) were taken to represent placenta previa. Four full wall thickness sections of non-attached areas of placental disc and sections of membrane and umbilical cord were taken for evaluation of placental abnormality. All sections were stained with hematoxylin and eosin and were then evaluated by an expert pathologist.

The absence of decidua between the placental villi and myometrium was considered placenta accreta, deeper invasion into the myometrium was considered increta, and complete invasion through the uterine was considered percreta [17].

Location of placenta in sonography was classified as: anterior high, fundal region posterior high, lateral high, anterior low lying, posterior low lying, lateral low, anterior previa, and posterior previa.

Data on age, number of Cesarean sections, gestational age at which diagnosis of PAS was considered, gestational age of delivery, gestational age of first Cesarean section, risk factors including: previous rupture of uterus, dilation and curettage, other operations, myomectomy, and sonography indices including abnormal placental lacuna, loss of clear zone, bladder wall interruption, myometrial thinning, placental bulging, exophytic mass, utero-vesical hypervascularity, subplacental hypervascularity, existence of bridging vessels, lacunar flow, and final diagnosis were registered for each patient.

High pressure in placental lacuna was considered as blood velocity higher than 15 cm/s [18].

In order to increase coherency and objectivity of study, all definitions of sonography indices used in the study, were done according to the unified definitions proposed by the EW-AIP 2016 [12]. Moreover, to remove any inter-observer bias, all sonography evaluations were done by a single radiologist who was blinded to the classification of patients (high risk or low risk groups).

In our study we considered grade 2 and grade 3 lacuna according to the the EW-AIP description as “new lacuna”.

All univariate analysis was done using the SPSS® software for windows®, version 18, (SPSS Inc., Chicago, IL, USA). Initially individuals were classified as high risk and low risk groups (based on the existence of low lying placenta or placenta previa), after which they were categorized as PAS and non-PAS (based on pathology reports) and compared accordingly. For comparison of qualitative data the Chi-square test and the Fisher’s exact test and for comparison of normally distributed quantitative data, the independent-test was used. Furthermore, for comparison of quantitative data without a normal distribution the Mann-Whitney test was used.

The Kappa test was used to evaluate the overlap and agreement of each sonography index with the gold standard modality, for the diagnosis of PAS.

The receiver operator characteristic (ROC) curve analysis was used to define sensitivity and specificity of each sonography index compared to the gold standard diagnostic modality for the diagnosis of PAS. A cut-off point was also defined based on lacunar flow among individuals with lacunar structures to estimate the occurrence of PAS. The Youden index was used to define the optimal cut-off point based on lacunar flow for diagnosing PAS. The Youden index considers a point on the ROC curve optimum which has the maximum sensitivity and specificity [19].

Due to the nature of PAS rates among pregnant individuals, and due to the relatively small sample size and large number of variables, using traditional logistic regression was not appropriate in this setting. To investigate simultaneous effects of 15 variables on PAS, Minimax Concave Penalty (MCP) was used. Like other penalized models, degree of shrinkage in MCP is regulated by the tuning parameter which is a positive constant and estimates using the cross validation technique. Classification accuracy of the proposed model was investigated by obtained probability of PAS for each case and optimal cut-off point was determined using the ROC curve analysis. Statistical analysis was performed using SPSS 18.0 and ncvreg package in R 3.3.1 software.

In the agreement analysis and the regression model abnormal lacuna structures were re-defined into a different variable as new lacuna. New lacuna was categorized into two categories as followed: those with grade 0 and 1 were considered as no lacuna and those with grades 2 and 3 as positive for new lacuna.