Prenatal identification of invasive placentation using ultrasound: systematic review and meta-analysis

Authors


  • All authors contributed equally to the manuscript.

ABSTRACT

Objective

The accuracy of prospective sonographic prenatal detection of invasive placentation is unclear. The objective of this study was to conduct a systematic review and meta-analysis to assess the performance of ultrasound in at-risk women for prenatal identification of invasive placentation.

Methods

MEDLINE, EMBASE, The Cochrane Database of Systematic Reviews, Database of Abstracts of Reviews of Effects (DARE) and The Cochrane Central Register of Controlled Trials (CENTRAL) were searched using the search terms ‘placenta accreta’, ‘placenta increta’, ‘placenta percreta’, ‘ultrasound’, ‘magnetic resonance imaging (MRI)’, ‘invasive placenta’ and ‘infiltrative placenta’. Two authors independently abstracted data from the articles. Sensitivity, specificity, positive and negative likelihood ratios (LR+ and LR–), the diagnostic odds ratio (DOR) and their 95% CIs for each study were calculated. Forest plots and summary receiver–operating characteristics curves were produced. Between-study heterogeneity was explored both graphically and statistically. The MOOSE (meta-analysis of observational studies in epidemiology) guidelines were followed.

Results

Twenty-three studies involving 3707 pregnancies at risk for invasive placentation were included. The overall performance of ultrasound for the antenatal detection of invasive placentation was as follows: sensitivity, 90.72 (95% CI, 87.2–93.6)%; specificity, 96.94 (95% CI, 96.3–97.5)%; LR+, 11.01 (95% CI, 6.1–20.0); LR–, 0.16 (95% CI, 0.11–0.23); and DOR, 98.59 (95% CI, 48.8–199.0). Among the different ultrasound signs, color Doppler had the best predictive accuracy (sensitivity, 90.74 (95% CI, 85.2–94.7)%; specificity, 87.68 (95% CI, 84.6–90.4)%; LR+, 7.77 (95% CI, 3.3–18.4); LR–, 0.17 (95% CI, 0.10–0.29); and DOR, 69.02 (95% CI, 22.8–208.9)).

Conclusions

Ultrasound has a high accuracy for prenatal diagnosis of disorders of invasive placentation in high-risk women. The use of color Doppler improves the test performance. Copyright © 2013 ISUOG. Published by John Wiley & Sons Ltd.

INTRODUCTION

Morbidly adherent placenta is a spectrum of conditions characterized by an abnormal adherence of the placenta to the implantation site. Three major variants of adherent placentation can be recognized according to the degree of trophoblastic invasion through the myometrium and the uterine serosa: placenta accreta, placenta increta and placenta percreta. Placenta accreta probably represents the most common variant of abnormally adherent placenta, and all varieties of invasive placentation are associated with a significant increase in maternal morbidity, especially due to blood loss, damage to local organs, need for urgent hysterectomy and postoperative complications[1]. Placenta previa and previous uterine surgery represent the major risk factors for invasive placentation[2-6]. Prenatal diagnosis of invasive placentation is associated with a reduced risk of maternal complications such as peripartum blood loss, need for transfusions and rate of hysterectomy, as it allows a preplanned treatment of the condition[7-13].

Several terms are used to describe morbidly adherent placenta; ‘invasive placentation’ is the term used in this manuscript to include placenta accreta and its variants. Ultrasonography is usually employed as the primary modality for antenatal diagnosis of invasive placentation. Prenatal magnetic resonance imaging is reported to be complementary to ultrasound, as it may help in diagnosing invasive placentation, especially in those cases in which ultrasound is not conclusive, to assess the degree of invasion (i.e. ambiguous ultrasound findings, posterior placenta previa)[14, 15]. The performance of antenatal ultrasound and of different sonographic signs is not consistent across published studies. This is most probably owing to a combination of limited sample size, retrospective design and variability of inclusion criteria and definition of invasive placentation.

The aim of this review was to systematically assess the performance of ultrasound in the prenatal diagnosis of placenta accreta and its variants and to explore the role of the different specific ultrasound signs in predicting disorders of invasive placentation.

METHODS

Search strategy

This review was performed according to a protocol designed a priori and recommended for systematic reviews and meta-analyses[16-18]. MEDLINE, EMBASE and The Cochrane Library including The Cochrane Database of Systematic Reviews (CDSR), Database of Abstracts of Reviews of Effects (DARE) and The Cochrane Central Register of Controlled Trials (CENTRAL) were searched electronically on 7th February 2013, utilizing combinations of the relevant medical subject heading (MeSH) terms, keywords and word variants for ‘placenta accreta’, ‘placenta increta’, ‘placenta percreta’, ‘ultrasound’, ‘magnetic resonance imaging (MRI)’, ‘invasive placenta’ and ‘infiltrative placenta’ (Table S1 online). The search and selection criteria were restricted to the English language. Reference lists of relevant articles and reviews were hand-searched for additional reports.

Study selection

Studies were assessed according to the following criteria: population, outcome, prenatal diagnosis of placenta accreta by ultrasound and study design. In this review, the general term ‘invasive placentation’ will refer to placenta accreta and its variants (increta/percreta). For the purpose of this study, invasive placentation was defined based on histopathological diagnosis of trophoblastic invasion through the myometrium or clinical assessment of abnormal adherence/evidence of gross placental invasion at the time of surgery in the absence of histopathological evidence. The overall sensitivity and specificity of prenatal ultrasound in the diagnosis of invasive placentation and the predictive value of various sonographic signs were noted.

The sonographic signs included in this review were the ones most commonly reported to be associated with invasive placentation and comprise: (1) vascular lacunae within the placenta, (2) loss of normal hypoechoic retroplacental zone, (3) interruption of the bladder line and/or focal exophytic masses extending into the bladder space and (4) color Doppler abnormalities such as abnormal blood vessels at the myometrium–bladder interface[14]. Interruption of the bladder line and the presence of exophytic masses extending into the bladder space were considered together and labeled as ‘abnormalities of the uterus–bladder interface’. When multiple color Doppler signs were reported and their overall presence in positive and negative cases could not be extrapolated, only the one showing the best predictive performance was included. In cases in which the overall performance of ultrasound and the number of imaging criteria used to diagnose invasive placentation were not stated, the sign showing the best predictive value was used as a surrogate for the final diagnosis.

Prospective and retrospective cohorts, case–control studies, case reports and case series were analyzed. Only studies reporting a prospective diagnosis of invasive placentation and/or the evaluation of single ultrasound signs in the second and/or third trimesters of pregnancy and studies for which the number of true positives, false positives, true negatives and false negatives were available were included in the final analysis. Opinion articles and studies carried out only in the first trimester of pregnancy were excluded. Case reports and case series with fewer than five cases and larger case series with a lack of information on false negatives were also excluded in order to avoid publication bias.

We decided to exclude several published reports. Some reported the inclusion criterion ‘suspicion of accreta’ on routine ultrasound. This is problematic, since there are no objective criteria for this condition. No information was given on how many accreta cases were missed because of a lack of suspicion on routine ultrasound. Absence of definition of ‘accreta’ also led to exclusion.

Data extraction and quality assessment

Two reviewers (F.D., C.I.) independently extracted data. Inconsistencies were discussed by the reviewers and consensus reached. For those articles in which targeted information was not reported but the methodology was such that the information might have been recorded initially, the authors were contacted requesting the data. The quality of the studies was assessed using the revised tool for the quality assessment of diagnostic accuracy studies (QUADAS-2)[19]. Each item is scored ‘yes’ or ‘no’, or ‘unclear’ if there is insufficient information to make an accurate judgment[18].

Statistical analysis

Meta-DiSc 1.4 (http://www.hrc.es/investigacion/metadisc_en.htm; Hospital Universitario Ramón y Cajal, Madrid, Spain) was used to analyze the data[20]. Heterogeneity was identified using Cochran's Q test and the I2 statistic, in which P < 0.05 and I2 ≥ 50% indicate significant heterogeneity[20, 21]. According to the results of heterogeneity testing, we chose an appropriate statistical model (random or fixed effects model) to pool the sensitivity, specificity, positive likelihood ratio (LR+), negative likelihood ratio (LR–) and diagnostic odds ratio (DOR). Pooled sensitivity, specificity, LR+, LR– and DOR were calculated according to reconstructed 2 × 2 tables[22-25]. The DOR is defined as the ratio of the odds of the test being positive if the subject has a disease relative to the odds of the test being positive if the subject does not have the disease (and is also related to the likelihood ratios as: LR+/LR–)[26]. Additionally, summary receiver–operating characteristics (sROC) curves were plotted. The area under the curve (AUC) and the Q* index were also computed to evaluate the overall performance of the diagnostic test accuracy. The AUC of an sROC curve is a measure of the overall performance of a diagnostic test in accurately differentiating those cases with and those without the condition of interest. The Q* index is defined by the point at which sensitivity and specificity are equal, which is closest to the ideal top-left corner of the sROC space. Both values range between 0 and 1, with higher values indicating better test performance[27].

RESULTS

General characteristics of the studies

The search yielded 447 possible citations; of these, 360 were excluded by reviewing the title or the abstract, as they did not meet the selection criteria. The remaining 87 full-text manuscripts were retrieved, and 23 studies were included in the final analysis. These 23 studies included 3707 pregnancies at risk for invasive placentation, mainly based on the presence of anterior placenta and a history of uterine surgery. The incidence of invasive placentation in this review was 9.3%. A summary of the included studies is given in Table 1.

Table 1. Characteristics of studies included in the systematic review
ReferenceStudy designInclusion criteriaTrimester at scanReference standardDiagnostic criteria (n)Women scanned (n)Invasive placentation (n)
  1. Only first author of each study is shown. Clin, clinical findings; CS, Cesarean section; Path, pathology; PP, placenta previa; Prosp, prospective; Retro, retrospective; Surg, surgical findings; ?, not stated.
Cali (2013)[38]ProspPP, previous uterine surgery2–3Path≥ 218741
Fishman (2011)[44]RetroPP3Path≥ 215423
Esakoff (2011)[45]RetroPP3Path?10819
Hamada (2011)[30]ProspPP and/or previous CS3Path≥ 2705
Mansour (2011)[41]ProspPP, previous uterine surgery3Surg≥ 23515
Lim (2011)[42]RetroPrevious CS and/or third-trimester bleeding2–3Path≥ 1139
El Behery (2010)[46]ProspPP, previous CS or uterine surgery3Path, Surg≥ 1357
Shih (2009)[37]ProspPP and/or previous CS or uterine surgery3Path≥ 117039
Chou (2009)[47]RetroPP, previous CS2–3Path?446
Wong (2008)[52]RetroPP and/or previous CS or uterine surgery2–3Path, Surg≥ 2669
Masselli (2008)[39]RetroPP, previous CS2–3Path≥ 25012
Dwyer (2008)[40]RetroPP and/or previous CS or uterine surgery?Path, Clin≥ 23215
Miura (2008)[48]ProspPP and/or previous CS3Path≥ 4124
Japaraj (2007)[49]ProspPP, previous CS3Path, Surg≥ 1217
Warshak (2006)[43]RetroPP and/or previous uterine CS, uterine surgery1–2–3Path≥ 245339
Yang (2006)[31]RetroPP, previous CS2,3Path≥ 15123
Comstock (2004)[28]ProspLow anterior placenta, previous CS2–3Path≥ 1200215
Moodley (2004)[50]ProspPP3Path, Surg≥ 2303
Chou (2000)[34]RetroPP and/or previous CS or uterine surgery2–3Path?8017
Twickler (2000)[35]ProspPP, previous CS3Path, Surg≥ 12010
Levine (1997)[36]ProspPP, previous CS or uterine surgery3Path, Surg≥ 2197
Lerner (1995)[51]ProspPP, previous CS3Path≥ 2215
Finberg (1992)[32]ProspPP, previous CS2–3Path, Surg≥ 23415

Quality assessment based on QUADAS guidelines was conducted on all 23 studies included for systematic review (Figure 1). Most of the studies were of high quality and there was an overall low risk of bias and low concern regarding the applicability of the studies[19]. As there was evidence of significant heterogeneity between the studies included, a random effects model was used. The heterogeneity test results for sensitivity and specificity are illustrated in Figures 2 and 3.

Figure 1.

Summary of results of Quality Assessment of Diagnostic Accuracy Studies (QUADAS) tool for articles included in the present analysis. Proportion of studies with low (image), high (image) or unclear (image): (a) risk of bias; (b) concerns regarding applicability.

Figure 2.

Forest plots of overall sensitivity and specificity of ultrasonography in the prenatal diagnosis of invasive placentation according to the current analysis. Only first author's name is given for each reference; names of authors of prospective studies are in capitals and names of authors of retrospective studies are in capital/lower-case letters.

Figure 3.

Forest plots of sensitivity and specificity of specific ultrasound signs in the prenatal diagnosis of invasive placentation according to the current analysis: (a) placental lacunae, (b) loss of retroplacental clear space, (c) bladder-border abnormalities and (d) color Doppler abnormalities.

Diagnostic accuracy

The overall performance of ultrasound for the antenatal detection of invasive placentation was as follows: sensitivity, 90.72 (95% CI, 87.2–93.6)%; specificity, 96.94 (95% CI, 96.3–97.5)%; LR+, 11.01 (95% CI, 6.1–20.0); LR–, 0.16 (95% CI, 0.11–0.23); and DOR, 98.59 (95% CI, 48.8–199.0) (Table 2 and Figure 2). The sROC for the performance of prenatal ultrasound for the detection of invasive placenta is shown in Figure 4, while the corresponding values for specific ultrasound signs are reported in Figure 5 and in Table 2.

Table 2. Pooled values for sensitivity, specificity, positive and negative likelihood ratios (LR+ and LR–) and diagnostic odds ratio (DOR) for ultrasound overall and the different ultrasound signs in the identification of invasive placentation
Diagnostic methodStudies (n)Patients (n)Sensitivity (95% CI) (%)Specificity (95% CI) (%)LR+ (95% CI)LR– (95% CI)DOR (95% CI)
Ultrasound (overall)23370790.72 (87.2–93.6)96.94 (96.3–97.5)11.01 (6.1–20.0)0.16 (0.11–0.23)98.59 (48.8–199.0)
Placental lacunae13272577.43 (70.9–83.1)95.02 (94.1–95.8)4.52 (2.5–8.1)0.29 (0.20–0.43)24.32 (9.13–64.8)
Loss of hypoechoic space10263366.24 (58.3–73.6)95.76 (94.9–96.5)5.64 (2.3–14.1)0.38 (0.20–0.69)21.98 (6.8–70.6)
Abnormalities of uterus–bladder interface9257949.66 (41.4–58.0)99.75 (99.5–99.9)30.56 (8.1–115.5)0.51 (0.34–0.77)93.70 (35.5–247.5)
Color Doppler abnormalities1271490.74 (85.2–94.7)87.68 (84.6–90.4)7.77 (3.3–18.4)0.17 (0.10–0.29)69.02 (22.8–208.9)
Figure 4.

Summary receiver–operating characteristics curve (—) with 95% CI (---) for ultrasonography in the prenatal diagnosis of invasive placentation (area under the curve ± SE = 0.956 ± 0.011; Q* ± SE = 0.898 ± 0.015). Size of data points is proportional to study size.

Figure 5.

Summary receiver–operating characteristics curves (—) with 95% CI (---) for specific ultrasound signs in the prenatal diagnosis of invasive placentation: (a) placental lacunae (area under the curve (AUC) ± SE = 0.889 ± 0.032; Q* ± SE = 0.820 ± 0.034), (b) loss of retroplacental clear space (AUC = 0.884 ± 0.049; Q* = 0.815 ± 0.050), (c) bladder-border abnormalities (AUC = 0.934 ± 0.037; Q* = 0.870 ± 0.045) and (d) color Doppler abnormalities (AUC = 0.948 ± 0.020; Q* = 0.888 ± 0.027). Size of data points is proportional to study size. image indicates two overlapping data points.

The AUC for diagnostic accuracy was 0.956 (standard error (SE), 0.011) for ultrasound overall, 0.889 (SE, 0.032) for presence of placental lacunae, 0.884 (SE, 0.049) for loss of the hypoechoic space between the placenta and the myometrium, 0.934 (SE, 0.037) for abnormalities at the level of the uterus–bladder interface and 0.948 (SE, 0.020) for color Doppler abnormalities (Figure 5). Among the different ultrasound signs, color Doppler had the best predictive accuracy (sensitivity, 90.74 (95% CI, 85.2–94.7)%; specificity, 87.68 (95% CI, 84.6–90.4)%; LR+, 7.77 (95% CI, 3.3–18.4); LR–, 0.17 (95% CI, 0.10–0.29); and DOR, 69.02 (95% CI, 22.8–208.9)).

DISCUSSION

The findings from this review show that prenatal ultrasound has predictive accuracy in diagnosing invasive placentation in a population at high risk. Among the sonographic signs of invasive placentation, color Doppler had the best combination of sensitivity and specificity. Quality assessment of the studies showed that the study quality was generally high, high sensitivity and specificity being seen in both retrospective and prospective studies.

Women who had had previous uterine surgery and placenta previa were assessed for invasive placentation. The prevalence of invasive placentation was 9.3% in this review. This prevalence is heavily influenced by the largest study, of over 2000 women, in which the prevalence of invasive placentation was much lower (0.75%)[28]. This was most probably owing to the fact that the authors included women with previous Cesarean section and low-lying placenta detected in the second trimester of pregnancy, not all of which may have continued to be low-lying in the third trimester. If we exclude this study, the pooled prevalence of invasive placentation was 19.3%. We think this figure represents a more realistic estimation of invasive placentation in women with a low placenta in the third trimester who had had a previous Cesarean delivery.

Prenatal diagnosis of invasive placentation has been shown to reduce the rate of maternal morbidity, as it allows planned management of the condition through the use of interventional radiology techniques or a conservative surgical approach saving the uterus[8-13, 29]. Conservative surgical approach often involves the use of a fundal/classical incision to deliver the baby without disturbing the placenta. Complications from the placement of vascular occlusive balloons have also been reported. It is important that prenatal diagnosis of invasive placentation is accurate and the false-positive rate of the diagnosis is kept to a minimum.

The overall values of sensitivity and specificity for the ultrasound diagnosis of invasive placentation reported in this review are based on a number of sonographic criteria. We hypothesize that reduction in the number of sonographic criteria needed to label a scan as suggestive of invasive placentation may increase the sensitivity but is likely to reduce the specificity of the test. Conversely, an increase in the number of criteria needed to label a case as positive would reduce sensitivity but would improve specificity.

Assessment of individual signs should be viewed with caution. Observation of one sign is likely to increase the chance of detecting others, since the signs are not looked for in isolation. In this review, we found a sensitivity of 77.43 (95% CI, 70.9–83.1)% and a specificity of 95.02 (95% CI, 94.1–95.8)% for the presence of lacunae. The pathophysiology of placental lacunae is not clear, although their presence has been associated with an increased likelihood of invasive placentation[30-34]. Lacunae may be present even in women with placenta previa without myometrial invasion[30]. The invasion of trophoblastic tissue through the myometrium and the absence of decidua basalis in invasive placentation progressively lead to a reduction in myometrial thickness and a loss of the hypoechoic space between the myometrium and the placenta[35]. The low sensitivity of this ultrasound sign may arise because the lower uterine segment appears as a thin line during the late third trimester on conventional transabdominal ultrasound, and evaluation of the interface between the myometrium and the placenta may be difficult.

Higher degrees of placental invasion lead to the destruction of the outer third of the myometrium and uterine serosa with subsequent involvement of the bladder. This condition may be diagnosed with ultrasound by examining the border between the bladder and myometrium, which is normally echogenic and smooth[14]. Exophytic masses in the bladder are likely to be seen only with placenta percreta. Observation of this sign is a reliable ‘rule-in’ sign for making the diagnosis, but its absence does not exclude lesser degrees of placental adherence.

Many reported series on invasive placentation are retrospective in nature and few were blinded. With the benefit of hindsight, it may be easier to spot signs in images on prenatal ultrasound. Most series included in the current review were prospective. The rest were also prospectively conducted, but ultrasound signs were examined retrospectively. The availability, not only of true positives but also of true and false negatives, gives us confidence in the observed results. However, the results are not applicable to fundal or posterior placenta with invasive placentation. The results are only applicable to women with placenta previa and a history of a Cesarean delivery or uterine surgery. This is the reason for low scores for generalizability observed in the quality assessment. Low anterior placenta with invasive placentation poses the biggest challenge to the surgeon. Entry into the uterus for delivery of the baby is affected by the presence of the placenta. It is impossible not to disturb the placenta at conventional lower segment Cesarean delivery. We argue that the group of anterior placenta previa with a previous Cesarean section forms the largest group amongst women with invasive placentation, the group most likely to experience complications and the group in whom prenatal diagnosis is likely to have the biggest impact. Abnormalities on color Doppler and presence of abnormal vessels performed best as predictors of disorders of invasive placentation in high-risk women. However, this is not an objective criterion, and needs to be clarified.

In conclusion, the results of this study show that among the different ultrasound signs the presence of abnormal vasculature on color Doppler ultrasound has the best combination of sensitivity and specificity, and that abnormality of the uterus–bladder interface has the best specificity for the prediction of invasive placentation. The presence of placental lacunae and loss of the clear space between the placenta and the myometrium does not perform as well. In women with a low anterior placenta who have had uterine surgery, third-trimester ultrasound is highly sensitive and specific in diagnosing invasive placentation prenatally. In women with previous Cesarean section and low anterior placenta detected in the third trimester, the prevalence of invasive placentation is roughly one in five. Future research should be directed at developing objective criteria for color Doppler abnormalities and for determining the best surgical technique for delivery.

ACKNOWLEDGMENTS

We would like to thank Ms Tiffany Hoare, NHS liaison librarian, St George's University of London, for her help with the literature search. We would also like to thank Prof. Palacios-Jaquaremada, Dr Cali, Dr Wong and Dr Mansour for their contribution to this systematic review in terms of additional data supplied and support.

Supporting Information On The Internet

The following supporting information may be found in the online version of this article:

Table S1 Search strategy using MEDLINE, EMBASE, Cinahl and The Cochrane Library (since inception) including The Cochrane Database of Systematic Reviews (CDSR), Database of Abstracts of Reviews of Effects (DARE) and The Cochrane Central Register of Controlled Trials (CENTRAL)

Ancillary