Research design and tissue collection considerations for investigation of placenta accreta spectrum.

When it comes to the interpretation of data found in PAS reports, there are a number of factors to keep in mind. Beyond gestational age-matching, it is important to understand that there are several potential sources of variability in PAS cases. We have already given our reasons for excluding accreta, IVF-related PAS, and any pregnancy complication known to impact or be attributable to the placenta in order diminish sources of variability/confounders to interpretation. In some studies, researchers use a mixture of all three PAS types in their measures, which raises questions as to whether they are mixing results from tissues with different attributes or composition. In such studies researchers simply describe samples as PAS, without any reference at all to sub-type defined by degree of invasion ( Supplementary Table 1 ). While it is likely that the frequency of various cell types differs between case and control as we have seen for PAS and previa [ 40 ], is it also possible that there are significant differences between the cases themselves? This emphasizes the importance of knowing both the location of the tissue sampled and minimizing underlying subject variability with respect to potential etiologies. To reduce the variability produced by these various factors, it is better to analyze similar cases, e.g. all placenta percreta or all increta vs gestational age-matched placenta previa, even though this may decrease the number of cases available for study. An interesting variation is the primiparous PAS case where there appears to be no previa or evidence from the medical history of uterine damage [ 9 , 54 ]. Such cases are rare, but would be well worth studying in-depth, possibly via PAS treatment center consortia. It would be useful to know whether these spontaneous cases have a distinct molecular profile from “classic” PAS, i.e. the combination of cesarean scar and previa, and from IVF-related PAS. Are these cases of a different origin or are they the occurrence of PAS in a location where there was an undiscovered uterine abnormality, now concealed or obliterated by the abnormal invasion? We do not yet know how to define the timeline of an individual PAS case, whether abnormal invasion started in the first trimester (as seen in cesarean scar pregnancy [ 55 ]) or the second trimester, whether formation of the decidua is defective and therefore permits over-invasion, or whether the invasive process destroys what was a normal decidua earlier in pregnancy. We do not know whether myometrial invasion is still advancing during the third trimester, or arrested earlier in gestation. While also relatively rare, study of second trimester, or even earlier PAS cases with an eye towards examining these issues is warranted, but will also likely require collaboration between centers.

The rate of occurrence of suitable cases is not high, especially if you are not at a specialized clinical center for PAS. In these circumstances, it is important to consider all the possible samples you may want to collect, even though your protocol may not immediately require them. How many times have you wished you had collected (fill in the blank)? When considering tissue samples from a PAS pregnancy, it is useful to take tissue not only from the site of deepest invasion, but also, if possible, from a location away from the deep invasion site where the uterine penetration of the placenta resembles that of a normal placenta, most notably, presence of decidua (e.g. [ 35 , 36 ]). It is important to remember that the deepest invasion site and “normal” samples from the same placenta can be regarded as paired (subject serves as their own control) and may provide a better control comparison than with a sample from a different subject. It is also important to note however that there is currently limited data on the comparison of two such samples against an unrelated sample. Thus, caution is needed as limited evidence suggests that cells from the ‘normal’ collection site (i.e. PAS but remote from site of invasion, with decidua) differ from those of control samples without PAS [ 35 – 37 ]. Although maternal and fetal (cord) blood samples can be obtained before/during surgery, sometimes this is not always possible, in which case, cord blood can often be obtained during gross dissection from the umbilical vein at the insertion of the cord. Sampling tissue at the site of deepest invasion will mean different things depending on the type of PAS being examined. In the case of placenta accreta or increta, the less invasive of the PAS types, it will be important for histopathology and immunohistochemistry to obtain samples which span the serosal-myometrial interface to villous tissue, allowing calculation of invasion depth as a percentage of myometrial depth [ 2 ], as well as investigation of the tissues surrounding the invasion site. For preparation of extravillous trophoblast, the basal plate, the maternal-facing layer of placental tissue adjacent to the myometrium is the most important as these cells form the leading edge of invasion [ 38 , 39 ]. For percreta samples, there is usually no myometrium and frequently the uterine serosa is damaged or absent if the villous tissue has broken through. In this case, the leading edge is the only sample that can be obtained from the invasion site. For those samples where the tissue is being extracted, recording the site of sampling is very useful for later understanding of the pathology, especially when combined with well-oriented OCT- or paraffin-embedded sections. As all the samples described above will have been obtained following delivery and sometimes lengthy surgery, it is imperative that these be cooled and processed as soon as possible after acquisition. Bulk tissue samples can be flash-frozen and blood samples separated into component fraction relatively quickly. Fresh tissue can be kept on ice until processing, if single cell RNA-seq, or cell-specific isolation from the mixed tissue is undertaken. Tissue sections intended for immunohistochemistry should be immersed immediately in your cold, selected preservation medium such as 10% neutral buffered formalin for a period of up to but no more than 24 hr. Preparing cells from PAS tissue samples can be challenging depending on the cell type of interest. Cytotrophoblast are not a problem because of their overwhelming numbers. Extravillous trophoblast or immune cells however may be difficult to isolate given their smaller numbers, especially in term placental tissue [ 13 ]. Unfortunately, unlike other placental cell preparations, scaling up the preparation is not an option given the need to obtain cells from the relatively limited leading edge area at the invasion site. After cell dispersion using your favorite enzymatic digestion mix (we use a collagenase/trypsin/nuclease mix) and removal of red blood cells by lysis, we have found that the best method for purifying specific cell types is by immunological means [ 40 ]. Thus, we have isolated extravillous trophoblast using the distinctive HLA-G marker as an antigen [ 41 , 42 ]. Similarly, cytotrophoblast can be isolated using antigens such as integrin ß4 [ 43 ]. It is important to remember the possibility for contamination by small quantities of cells also expressing the capture antigen, for example HLA-G-expressing immune cells [ 44 ]. For scRNA-seq, fresh cell preparation is much to be desired, although frozen cell mixtures will suffice, provided cell composition numbers are not relevant to the study or for which compensation can be derived. The small numbers of cells such as extravillous trophoblast isolated from PAS samples (0.25–0.5 × 10 5 cells/g tissue; [ 40 ]) limits the purposes to which these cells can be allocated. RNA and DNA are less problematic because they can be isolated from the same samples and are further amplified during assay. However, obtaining sufficient protein for procedures such as Western blotting can be challenging and may require the entirety of a placenta cell preparation - an entire case. Techniques such as mass spectrometry, requiring less sample, may need to be considered, especially for the detection of multiple proteins. After tissue acquisition, there are a number of pieces of information necessary to complete the description of a case. Standard obstetric information is of course required, including sex of the neonate, gestational age at delivery and birthweight/birthweight centile. In addition to confirming PAS by histopathology, it is advantageous to use pathological grading according to the FIGO or SPP guidelines [ 9 ]. Table 2 contains the minimum information to be obtained after delivery.

The choice of controls is perhaps the most difficult of research design decisions to make in PAS. One type of control is tissue taken from the same PAS placenta but remote from the site of deepest invasion, a “normal” sample. This does come with obvious advantages; it is a paired sample not only genetically and immunologically, but also for gestational age. However, we do not as yet know, as noted above, whether the uterine conditions or endocrine milieu may affect this sample in a similar but perhaps lesser manner than the tissue at the primary invasion site. Nevertheless, it can be simply acquired and provides one reference point against which to assess the PAS deepest invasion sample. Pathological series have used CD-hysterectomy samples derived from conditions such as atony [ 45 ] for IHC comparison, but the rarity of these cases and their unpredictable occurrence renders it difficult to collect fresh tissue samples from such cases. Clinical judgement will differ as to the optimal gestational age for delivery, but for increta and percreta samples this is usually <36 weeks gestational age. Recommendations for delivery are generally between 34–36 weeks of gestation [ 46 ]. However in our experience, 40% of increta and percreta cases whose deliveries were targeted for this gestational age range were delivered earlier, usually for bleeding. Under these circumstances, many researchers opt for samples which can be gestational age-matched. In this situation, two different sets of controls have been utilized. The first has been preterm births, many of which fall into the gestational age window during which PAS pregnancies are usually delivered. Most researchers also require that these pregnancies have no signs of infection and be delivered by CD without labor. It is understood that vaginal birth, laboring birth (even if eventually resulting in CD), and those showing signs of infection are conditions that alter the tissue and cellular parameters under investigation and are thus precluded [ 47 – 50 ]. The other useful control, which we have chosen to use, are placenta previa pregnancies. These pregnancies are often delivered early, like PAS, due to episodes of second and third trimester bleeding and also concern for abruption [ 51 ]. These pregnancies are delivered by CD for “mechanical”, as opposed to biological reasons; the need for CD delivery is simply necessity, given the blockage of the cervical os. The previa differs therefore from the preterm birth, where there is a probability that aberrant biochemical or endocrine factors have forced delivery before term and, again, may influence the tissue and cellular parameters under investigation. The previa is thus more likely to permit gestational age-matching, is less likely to be affected by biochemical or endocrine factors and, most importantly, is similar to the majority of PAS pregnancies which are also placenta previa. We have shown that gestational-age-matched previa without PAS differs little from non-previa at the molecular level [ 40 ]. Other evidence suggests that tissue from placenta previa may be different in some ways from normal term placenta [ 52 , 53 ], but this is likely to be much more limited than that for placentas obtained from preterm birth. We have here identified a number of issues relevant to the investigation of PAS: the need for high-quality histopathological diagnosis, separation of PAS cases by severity for analytical purposes, reducing inter-subject variability and choice of controls. As examples, Supplementary Table 1 reviews ~20 years of reports where PAS bio-samples were the foci of investigations, and the table codifies for each publication where the issues above might be relevant.

The presumption here is that you wish to obtain fresh tissue from PAS pregnancies, placental and/or uterine. In most centers equipped to treat PAS, clinical pathologists handle the hysterectomy specimens and confirm diagnosis by histopathology. Beyond collaboration with the clinical personnel that manage these cases, validation of PAS diagnosis requires the involvement of a clinical pathologist, preferably a perinatal or placental pathologist. Studies using basic science approaches that include ‘clinically’ diagnosed cases are not acceptable. It is highly recommended that investigators use the FIGO or SPP classification systems [ 2 , 3 ] to characterize subjects. Both SPP and FIGO gradings parallel the less precise classification of accreta, increta and percreta, but refine diagnoses relative to PAS severity. FIGO utilizes antepartum imaging and surgical findings for classification, while still emphasizing the need for histopathological confirmation. The SPP classification was developed by perinatal pathologists for consistency of diagnosis, correlation with antenatal imaging, and includes an online pathology image library for each diagnostic grade of PAS [ 3 ]. For molecular studies the SPP classification is preferred, as the FIGO system is inherently clinical, and accreta without hysterectomy cannot be diagnosed other than by clinical criteria. In contrast, SPP provides guidelines for histopathological diagnosis of superficial accreta without hysterectomy. Both systems evolved from the need for PAS diagnoses that reflect disease severity in terms of clinical outcomes such as surgical complications, transfusion products and subsequent hospitalization or surgery. SPP shows closer correlation with these clinically relevant outcomes than the traditional accreta, increta, percreta [ 6 , 7 ]. The FIGO system correlates closely with histopathological findings [ 8 ] and, in the only test we know of that directly compared the comparability of SPP and FIGO systems, yield consistent and reproducible classifications that are well-correlated [ 9 ]. In the best possible research environment, both FIGO and SPP should be applied and the clinician and the pathologist contribute to distinguishing an iatrogenically caused appearance of percreta (e.g. dissection of the adherent placenta from the bladder dome often leads to perforation of the uterine serosa, leading to what some studies have suggested is misdiagnosis of percreta as cesarean scar dehiscence, uterine window or iatrogenic) [ 10 ]. Intra-operative photographs should be taken, guided by the need for the pathologist to distinguish iatrogenic from pathologic damage. There are two follow-on problems with the primary requirement of histopathological confirmation . It is absence of decidua and degree of myometrial invasion that determine diagnosis. Therefore, the first problem is that gold-standard diagnosis requires hysterectomy or excised/resected tissue containing the maternal-placental interface and underlying myometrium [ 1 – 3 ]. This 1) makes obtaining control tissue samples extremely difficult (see below); 2) precludes many cases of accreta where uterine conservation is prevalent (unless special histopathological procedures such as those recommended in the SPP classification are undertaken); and 3) is incompatible with some forms of medical management (e.g. embolization) [ 11 , 12 ]. In molecular investigations we have excluded accreta altogether, using only histopathologically confirmed cases of increta/percreta for molecular studies [ 13 ]. Our reasons are two-fold. First, in our population, accreta was disproportionately associated with ART when compared with increta/percreta and had significantly lower rates of the “classic” combination of placenta previa and cesarean scar [ 14 ]. Subsequent studies with larger sample sizes confirms ART pregnancies resulting in PAS have lower rates of prior CD and placenta previa [ 15 , 16 ]. Hence, at least in our current medical environment, epidemiological risks factors for accreta differ from increta and percreta, occurring more frequently in subjects of no/low parity, using ART, and less frequently involving placenta previa or cesarean scar. This suggests different pathophysiological mechanisms may be involved in the more severe forms of PAS versus accreta. Second, where invasion is superficial, as in accreta, the type and amount of tissue available that is invasive is more difficult to collect in volumes sufficient for, e.g., isolating specific cell-types, while tissues containing mixed cell types are more likely to contain ‘normal’ cells than in cases where adequate tissue volume can be obtained from deeper and more widespread invasion. An additional rationale is that more severe forms of PAS may reveal greater differences in gene/protein expression than the mildest form. In some clinical centers, there can be attempts to remove the placenta surgically, depending on invasion depth, followed by some form of uterine re-section [ 17 , 18 ]. In these cases, it may be difficult to obtain tissue from the most deeply invasive area or to acquire tissue samples which can be assessed as to the relationship between placenta, myometrium and serosa. This will in turn hinder histopathological determination of PAS status. Without an understanding of the exact tissue source location, it may be difficult to interpret results, since there may be changes in tissue composition and cellular distribution between different locations in the PAS placenta. Cellular distributions likely vary by degree of invasion. Systematic comparisons of cell-type distributions from samples derived from accreta vs. increta. vs percreta have not been done, but one can easily imagine PAS severity-related differences in extravillous trophoblast and endothelial cell numbers due to invasive pathology and increase in vascularity. The second research design consideration for basic science studies of PAS is the exclusion of confounders, one generally shared by other studies of complications with a placental contribution (e.g. GDM, preeclampsia, IUGR). Nonetheless, this is critical for studies in which high-throughput molecular approaches are used. If 50% of cases have other gestational complications impacting the placenta, how is one to interpret, for example, global mRNA expression or single-cell RNA-seq data? If there are to be breakthroughs in discerning the molecular events involved in PAS, other conditions impacting the placenta need to be eliminated insofar as possible. Therefore, in selection of PAS cases, pre-existing maternal health conditions (e.g. hypertension, diabetes, auto-immune diseases), as well as pregnancy complications known to involve the placenta should be excluded (e.g. preeclampsia, GDM). We do not include IVF pregnancies in our molecular studies. It is now well documented that the incidence of placenta previa, and of PAS are increased with IVF treatment. This is not to claim all ART-related treatment should be excluded, but the potential variability introduced by inclusion of IVF should be considered. It is not just the plethora of underlying causes of infertility that contribute to variability. The association of PAS with IVF is influenced by the methods used to treat infertility. Frozen embryo transfer (FET) is more closely associated with PAS outcome than other commonly used IVF methods, with differing approaches to endometrial preparation/stimulation modulating the effect [ 19 , 20 ]. It is also now known that non-previa implantation is associated with ‘milder’ PAS, presumably accreta [ 21 ] and that FET with pre-implantation sub-optimal endometrial thickness increases the FET-associated risk for PAS outcome several-fold [ 22 ]. Therefore, unless one is specifically comparing IVF-related PAS to non-IVF-related PAS, and most critically in molecular approaches, IVF-related cases should ideally be excluded or, if not, subjected to sub-analyses to ensure they do not differ from non-IVF related PAS. When fresh tissue collection/preservation is a research goal, characterization of cases prior to delivery is efficient, as well as permitting time for consent-related issues to be discussed in advance of the planned delivery, itself a highly stressful maternal event. This may limit non-previa accreta cases from prospective tissue collection, as diagnostic imaging, while having acceptable sensitivity and specificity for increta and percreta [ 23 , 24 ], presents fewer of the clear signs of abnormal invasion and is relatively poor at diagnosing non-previa accreta [ 21 ] This requires collaborative links with the physicians who are diagnosing, planning and performing the PAS surgeries. Antenatal diagnosis is via ultrasound and/or MRI. While final diagnosis can only be made after delivery, imaging techniques are sufficiently reliable to identify cases most likely to require hysterectomy. Liaison with the clinical Pathology Department as to when and what tissue can be obtained is required. We developed a protocol for tissue collection in conjunction with the Pathology Department leadership, which was then available to the pathology assistants performing gross dissection and fixation. Our tissue collection was carried out during the post-operative gross dissection of the uterus, immediately following the surgical hysterectomy, and prior to immersion of the tissues in fixative. It permitted sampling within 30–60 minutes of removal of the uterus, selection of tissue adjacent to and remote from the area of deepest invasion and photographing of the strips of tissue that were sampled (for depth of invasion). The best option therefore is for a team member to be present in Pathology after delivery of the neonate, awaiting the uteroplacental tissue removed during hysterectomy. The other information required concerns the clinical and demographic data related to the subjects. It is more expansive than the conventional subject characteristics table. For PAS, given relative rarity and the difficulty of study, maternal variables relevant to PAS risk should be extensively investigated, if possible, and included. Such data bolsters the diagnosis and contributes to general clinical knowledge. There are likewise many databases used for large scale epidemiological investigation of PAS that could be usefully deployed in sub-analyses. Most epidemiological/outcomes analyses include clinically diagnosed PAS, and do not distinguish between accreta, increta and percreta referring to all cases in general as PAS (e.g. [ 25 , 26 ]). Most such databases will not contain sufficient detail to employ FIGO or SPP classification. However, sub-analyses by PAS severity inthe sub-group in which histopathological confirmation is available, would benefit both the epidemiology and ART literature, the latter of which also includes large numbers of clinically as opposed to histopathologically diagnosed PAS (e.g. [ 27 , 28 ]). In short, analyses lumping together clinically and histopathologically confirmed cases, and failure to analyze with disease severity in mind blurs what might be differences in risk factors or biological pathways contributing to disease severity ( Supplementary Table 1 ). Thus investigators should carefully document the diagnoses and at least include in their subject characteristics how many cases were accreta, increta, percreta, how many were caesarean-hysterectomy versus vaginal deliveries and, if possible the FIGO or SPP grade. Because pre-term delivery is prevalent in PAS, neonatal birth weights should be given as centiles instead of or in addition to raw values. Many early, and even current literature on PAS characterizes the neonates as SGA or significantly lower birthweight [ 29 , 30 ], when in fact PAS neonates do not have excess SGA nor LGA when validated, gestational age/sex specific birth weight centiles are compared [ 31 – 33 ]. Placental location should be given where possible. Previa should only be reported if the placenta, at least partially, covers the cervical os, and if it is partial vs. central previa, these should be distinguished. If the placenta is marginal, the distance to the os and laterality should be reported (e.g. left, right, posterior marginal previa, 2 cm from os). Placental location is generally noted in ultrasound reports and would be confirmed where cesarean-hysterectomy is involved. Caution should be used, as scans obtained at 25 weeks or less may show complete, partial or marginal previa. However up to 95% of the marginal/partial and up to 75% of the previas completely covering the cervical os on 2nd trimester scans will resolve. This is because as the placenta grows it tends to grow towards areas of higher perfusion, and also because as the lower segment of the uterus grows, the placental margins become more distant from the os [ 34 ]. For an excellent consideration of conditions related to uterine iatrogenic procedures or structural anomalies that potentially alter development of the utero-placental interface (PAS predisposing conditions in the table below), see Neville et al. [ 9 ]. Table 1 lists the various parameters which can be used to judge whether a subject should be designated as a PAS case, placenta previa case or other control.

It is important to understand fully the challenges presented by PAS before designing and undertaking research. In the absence of crucial parameters, it is frequently impossible to judge whether research conducted on tissue from PAS pregnancies is comparable or whether to have confidence in the data so obtained. Comprehensive research design requires an appreciation of the clinical conditions, epidemiological associations as well as an understanding of the anatomical and structural complications of this disorder. While still a relatively rare problem compared to the major obstetric pathologies, the rising incidence continues to attract attention. This rarity however also necessitates careful consideration of experimental design so as to obtain sufficient sample numbers, made even more scarce by the need to avoid variability. We hope the planning elements, caveats, indicators and questions provided in this report will advance meaningful PAS research.

Histopathological diagnosis of Placenta Accreta Spectrum (PAS) requires tissue sampling at the placental-maternal interface, including the underlying myometrium. While initially defined clinically in the 1930s as “undue adherence of the placenta”, by 1969 placenta accreta was histopathologically defined as absence of decidua with direct attachment of villous tissue to myometrial fibers [ 1 ]. Superficial attachment of trophoblast to myometrium became defined as placenta accreta, while deeper invasion was termed placenta increta and invasion up to or through the serosa, percreta. These categorical descriptions were widely used from the 1970s until relatively recently. Significant refinement of histopathological criteria into grades has been codified by the Society of Pediatric Pathologists [ 2 ]. A similar grading system developed by the International Federation of Gynaecology and Obstetrics (FIGO) using clinical and imaging parameters [ 3 ] as well as histopathology. Use of either approach will increase comparability between studies, especially those focusing on molecular analyses, where reproducible classification of cases is critical. A >50-fold increase in the incidence of PAS since the mid-20th century correlates with increasing rates of caesarean delivery (CD), the use of ART and older maternal age [ 4 , 5 ]. This, in turn, has triggered an increase in the number of reports addressing this damaging pathology. Most have focused on identification of risk factors, diagnostic imaging, clinical outcomes and management/interventions. It is only more recently that basic science investigations have become a significant focus, and herein lies the critical role of utilizing reliable, reproducible, diagnostic standards. As a lesser known but complex pathology, there are a number of papers being published, or submitted for publication which do not adequately classify PAS cases, utilize inappropriate controls or fail to consider confounders in the research design. A lack of rigor in study design and subject inclusion/exclusion diminishes the confidence one might otherwise have in these studies. This report is composed of several sections. The first considers issues related to research design and variables to be considered in selection of cases/controls. The next section concerns the acquisition of tissue and the considerations for its treatment and preservation. Subsequently we discuss the types of controls available for PAS investigations, and finally some gaps in knowledge.