Beruflich Dokumente
Kultur Dokumente
1–8, 2013
Copyright Ó 2013 World Federation for Ultrasound in Medicine & Biology
Printed in the USA. All rights reserved
0301-5629/$ - see front matter
http://dx.doi.org/10.1016/j.ultrasmedbio.2013.05.017
d Original Contribution
(Received 9 January 2013; revised 15 April 2013; in final form 24 May 2013)
Abstract—The aim of this study was to evaluate the diagnostic value of ultrasound (US) as compared with
magnetic resonance imaging (MRI) in the detection of placenta accreta. Sensitivity, specificity, summary receiver
operating characteristic curves and areas under the curve (AUCs) were described and calculated using Meta-Disc
Statistical Software, Version 1.4 (Unit of Clinical Biostatistics, Ram on y Cajal Hospital, Madrid, Spain). In the 13
studies included, US sensitivity was 83% (95% confidence interval [CI] 77%–88%), US specificity was 95% (95%
CI: 93%–96%) and the diagnostic odds ratio (DOR) was 63.41 (95% CI: 29.04–138.48). In the MRI studies, sensi-
tivity was 82% (95% CI: 72%–90%), specificity was 88% (95% CI: 81%–94%) and the DOR was 22.95 (95% CI:
3.19–165.11). Summary receiver operating characteristic analysis indicated that the diagnostic value of US in
detection of placenta accreta is not significantly different from that of MRI. Both US and MRI were highly sensitive
and specific in the detection of placenta accreta to support effective diagnostic methods. (E-mail: xielm72@yahoo.
cn) Ó 2013 World Federation for Ultrasound in Medicine & Biology.
Key Words: Ultrasound, Magnetic resonance imaging, Placenta accreta, Meta-analysis.
1
2 Ultrasound in Medicine and Biology Volume -, Number -, 2013
and (vi) increased vascularity proximal to the bladder (c) case reports or case series; (d) insufficient data to re-
wall on color Doppler (Chou et al. 2000; Finberg and assess sensitivity and specificity from an individual
Williams 1992; Thia et al. 2007). In contrast, the study; and (f) duplicate data.
diagnostic criteria extracted with MRI include (i) Studies using US with or without Doppler imaging
abnormal uterine bulging; (ii) heterogeneous signal of were included; similarly, studies using MRI with or
the placenta on T2-weighted (T2 W) images; (iii) dark in- without gadolinium contrast were included. All retrieved
traplacental bands on T2 W images; (iv) focal interrup- articles were reviewed by two observers who applied the
tions in the myometrial wall; (v) tenting of the bladder; inclusion and exclusion criteria described above.
and (vi) direct visualization of the invasion of pelvic Disagreements were resolved by consensus.
structures by placental tissue (Baughman et al. 2008;
Lax et al. 2007). Data extraction
The reported diagnostic accuracy of US and MRI Data were extracted independently from all studies
varies widely for placenta accreta. Meta-analysis is that fit the inclusion criteria by the same observer of study
a statistical method that allows the results of individual selection. When observer disagreements occurred, a third
studies to be evaluated systematically (Irwig et al. reader reviewed the article, and the disagreement was
1994). Therefore, the purpose of this article was to resolved by discussion. For each study included, certain
summarize the available publications and to compare data were extracted: first author, year of publication,
the diagnostic value of US and MRI in detection of study design (prospective or retrospective), patient char-
placenta accreta. acteristics (average gestational age), technical aspects
(e.g., probe type and probe frequency used in the US
METHODS studies, magnetic field strength and contrast agent used
in the MRI studies), image interpretation (blinded or
Search strategy not). For each study, the numbers of TPs, FPs, FNs and
A computerized literature search (Deville et al. TNs for US and MRI were recorded.
2000) of PubMed, EMBASE and the Cochrane Central
Register of Controlled Trials Database was conducted Quality assessment
to identify relevant published articles on the diagnostic The methodological quality of the articles was as-
accuracy of US or MRI in detection placenta accreta. sessed using the updated quality assessment tool
We used a combination of the following terms: (i) QUADAS-2 (Quality Assessment of Diagnostic Accu-
‘‘US’’ OR ‘‘ultrasound’’ OR ‘‘MRI’’ OR ‘‘magnetic reso- racy Studies) guidelines (Whiting et al. 2011) to evaluate
nance imaging’’; (ii) ‘‘placenta accreta’’ OR ‘‘placenta in- the risk of bias and applicability. Each item was rated as
creta’’ OR ‘‘placenta percreta’’; and (iii) ‘‘sensitivity’’ OR ‘‘low,’’ ‘‘high,’’ or ‘‘unclear.’’ The ‘‘unclear’’ category was
’’specificity’’ OR ‘‘accuracy’’ OR ‘‘diagnosis’’ OR used only when insufficient data were reported. In the
‘‘detection.’’ No date limits were specified, and the search publication that included two index tests (US and MRI),
was current as of March 20, 2013. To expand our search, it was necessary to assess each index test in QUADAS
the reference lists from all retrieved articles were DOMAIN 2 (index test). Certain items were omitted
screened to identify additional studies. because they were not relevant to this topic.
Table 1. TP, FP, FN, TN and other features of ultrasound (13 studies)
Average gestational Number Probe Probe Color andpower Image
Study age (wk) of patients TP FP FN TN type frequency (MHz) Doppler imaging interpretation
TP 5 true positive; FP 5 false positive; FN 5 false negative; TN 5 true negative; TA 5 transabdominal probe; TV 5 transvaginal probe.
* Prospective study design.
y
Retrospective study design.
z
Unknown study design.
significant). All statistics (sensitivity, specificity, diag- the field of interest of this review (n 5 338); (ii) articles
nostic odds ratio [DOR], SROC, AUC and Q*) refer to were published as reviews, letters, case reports or case
the diagnostic value of each index test. series (n 5 19); (iii) the data from the articles could not
Heterogeneity was assessed by calculating the I2 be used to construct or calculate TPRs, FPRs, FNRs or
statistic, which is a quantitative measure of the amount TNRs (n 5 13); (d) publications were not written in
of heterogeneity with an upper limit of 100%. I2 values English (n 5 3); and (e) the articles were published
of 25%, 50% and 75% were considered indicative of more than once (n 5 3). A total of 13 studies
low, moderate and high heterogeneity, respectively (Chalubinski et al. 2013; Chou et al. 2000; Dwyer et al.
(Higgins et al. 2003). When I2 exceeded 50%, a random 2008; Elhawary et al. 2013; Esakoff et al. 2011;
effect model was used in the meta-analysis. In all other Finberg and Williams 1992; Japaraj et al. 2007; Lam
cases, a fixed effect model was used. et al. 2002; Lerner et al. 1995; Levine et al. 1997; Lim
Statistical analyses were performed using Meta- et al. 2011; Masselli et al. 2008; Warshak et al. 2006)
Disc Statistical Software Version 1.4 (Unit of Clinical met all inclusion criteria and were selected for data
Biostatistics, Ram on y Cajal Hospital, Madrid, Spain) extraction and meta-analysis.
(Zamora et al. 2006).
Details on examination
The TP, FP, FN and TN results and other details
RESULTS
related to the 13 studies are summarized in Tables 1
Literature search and 2. Eleven of the 13 studies were retrospective, one
A comprehensive literature search revealed 389 study was prospective and the one was not defined. Gesta-
primary studies. After review of the titles, abstracts and tional ages of the patients included in the studies ranged
full texts, 376 articles were excluded. The reasons for from 11 to 37 wk. The diagnostic accuracy of US was
exclusion were as follows: (i) the articles were not in assessed in 13 papers, whereas six papers assessed the
Table 2. TP, FP, FN, TN and other features of magnetic resonance imaging (6 studies)
Average gestational Number Gadolinium Field Image
Study age (wk) of patients TP FP FN TN contrast material strength (T) interpretation
Table 3. QUADAS-2 results for included studies performed with ultrasound (US) or magnetic resonance imaging (MRI)
Risk of bias Applicability concerns
accuracy of MRI. In the six trials comparing US with combined the methods. Color Doppler imaging and
MRI, the relative accuracies of US and MRI were directly power Doppler imaging were used in 11 studies. The
compared within the same group of patients in four transducer frequency in US varied from 3.5 to
studies (Dwyer et al. 2008; Elhawary et al. 2013; Lim 7.5 MHz. In the MRI studies, the six papers included
et al. 2011; Masselli et al. 2008). In two studies, 183 patients who were available for the analysis.
patients underwent MRI evaluation because of Gadolinium contrast materials were used in two studies
suspicious findings on US or because US provided (Lam et al. 2002; Warshak et al. 2006).
inconclusive evidence of placenta accreta (Lam et al.
2002; Warshak et al. 2006). In the US studies, the 13 Quality assessment
papers included 1115 patients who were available for A quality assessment of the included studies based
the analysis. Several studies used transabdominal or on the updated QUADAS-2 is outlined in Table 3. Over-
transvaginal probes alone, whereas other studies all, the quality of the studies was satisfactory.
Meta-analysis DISCUSSION
The pooled sensitivity, pooled specificity and DOR
Ultrasound is the most commonly used imaging
of the two diagnostic methods obtained on the basis of
modality for the diagnosis of placental accreta because
a random effect model, are illustrated in Figures 1–4.
it is an inexpensive, non-invasive and time-saving
The pooled sensitivities of US and MRI were 0.83 and
method. Color Doppler and power Doppler, as well as
0.82, respectively, and there was no statistically
the use of a transvaginal transducer, appeared to improve
significant difference in sensitivity between US and
the performance of conventional ultrasound in the assess-
MRI (p . 0.05). The pooled specificities were 0.95 and
ment of placenta accreta. Color and power Doppler high-
0.88 for US and MRI, respectively; there were no
lighted areas of increased vascularity caused by dilated
statistically significant differences in specificity
blood vessels crossing the placenta and uterine wall
between US and MRI (p . 0.05). The DORs of US and
(Chou et al. 2002), which further validated the conven-
MRI were 63.41 and 22.95, respectively.
tional gray-scale sonographic features. A transvaginal
To compare US and MR, we used SROC analysis
transducer can improve the near-field resolution of the
(Figs. 5, 6). The AUCs for US and MRI were 0.9485
interface between the placenta and the lower uterine
and 0.8963, respectively, and the Q* indices were
segment, especially in cases of placenta previa or poste-
0.8884 and 0.8273, respectively. Whether the AUC of
rior placenta (Lerner et al. 1995). Higher-frequency
SROC curves statistically differed between the two diag-
probes have been reported to improve spatial resolution
nostic tests can be determined by Z-value statistics anal-
for superficial structures, thereby affecting the accuracy
ysis (Hanley and McNeil 1982). The Z-value for US and
of US (Benacerraf et al. 2000). Given that it is more
MRI was 0.8193, and the corresponding p-value was
accessible to patients and physicians, US will continue
0.4126.
to be the mainstay for prenatal diagnosis of placenta
Fig. 3. Sensitivity of diagnosis of placenta accreta with magnetic resonance imaging. CI 5 confidence interval.
6 Ultrasound in Medicine and Biology Volume -, Number -, 2013
Fig. 4. Specificity of diagnosis of placenta accreta with magnetic resonance imaging. CI 5 confidence interval.
accreta (Teo et al. 2009). However, US may be unreliable All but one of the studies found that US and MRI are
when imaging a posterior placenta; in such a case, MRI effective tools in the diagnosis of placenta accreta. Lam
can achieve superior diagnostic accuracy (Bakri et al. et al. (2002) reported the sensitivity of US and MRI as
1993; Thorp et al. 1992). 33% and 38%, respectively. However, the findings of an
More recently, MRI with and without gadolinium individual study have limited generalizability because
has been explored as an adjuvant modality for further of the relatively small number of patients. To derive
improving prenatal diagnosis of placenta accreta. It is more rigorous estimates of the diagnostic performance
believed that the gadolinium delineates the outer of US and MRI, we pooled published studies.
placental surface proximal to the myometrium more To the best of our knowledge, this systematic review
clearly; as a consequence, the confusion between hetero- with meta-analysis is the first attempt to compare the
geneous signals originating in the placenta from those diagnostic value of US and MRI in the detection of
caused by maternal blood vessels is eliminated, thereby placenta accreta. Considering the fact that diagnostic
enhancing the specificity of MRI. However, the use of ga- performance is determined by sensitivity as well as spec-
dolinium in pregnancy is controversial given the potential ificity, a meta-analysis that considers the two factors is
for negative effects (Dwyer et al. 2008; Warshak et al. necessary. The relationship between sensitivity and spec-
2006). ificity is non-linear, and SROC analysis is thus helpful in
transforming the information linearly to compare the
diagnostic methods subsequently. Analysis of the SROC
curves revealed that the AUC of US was 0.9485 with
a Q* index of 0.8884, whereas the AUC of MRI was
0.8963 with a Q* index of 0.8273. The above results indi-
cate that both US and MRI have a high degree of sensi-
tivity and specificity in diagnosing placenta accreta.
Significant difference analysis of the two diagnostic
methods was performed using the Z-value test from the
AUCs of the SROC curves. In this study, the Z-value
was 0.8193 with a corresponding p-value of 0.4126, indi-
cating the absence of a significant difference in diagnostic
value between US and MRI.
Massive hemorrhage and associated morbidities,
including cystotomy, ureteral injury, pulmonary embolus,
admission to the intensive care unit and re-operation, are
common with placenta accreta. When placenta accreta is
suspected on US or MRI, the attending obstetricians
should prepare and counsel the patients and assemble
a multidisciplinary team (including obstetricians, anes-
Fig. 5. Summary receiver operating characteristic (SROC) thetists, interventional radiologists, urologists, visceral
curve for ultrasound. The size of the circle is proportional to
the weight of the individual study. AUC 5 0.9485; surgeons and neonatologists) to handle a potentially
SE(AUC) 5 0.0148; Q* 5 0.8884; SE(Q*) 5 0.0198. life-threatening situation in the operation room (Japaraj
AUC 5 area under the curve; SE 5 standard error. et al. 2007). It is reported that approximately 90% of
US vs. MRI in diagnosis of placenta accreta d X. MENG et al. 7
CONCLUSIONS
There is no significant difference in diagnostic value
between US and MRI in placenta accreta. Both US and
MRI are highly specific and sensitive in diagnosing or
excluding the presence of placenta accreta, and US
should be the first choice for patients with limited time
and financial support.
Irwig L, Tosteson AN, Gatsonis C, Lau J, Colditz G, Chalmers TC, morbidity associated with multiple repeat cesarean deliveries. Ob-
Mosteller F. Guidelines for meta-analysis evaluating diagnostic stet Gynecol 2006;107:1226–1232.
tests. Ann Intern Med 1994;120:667–676. Sofiah S, Fung YC. Placenta accreta: Clinical risk factors, accuracy of
Japaraj RP, Mimin TS, Mukudan K. Antenatal diagnosis of placenta pre- antenatal diagnosis and effect on pregnancy outcome. Med J
via accreta in patients with previous cesarean scar. J Obstet Gynae- Malaysia 2009;64:298–302.
col Res 2007;33:431–437. Solheim KN, Esakoff TF, Little SE, Cheng YW, Sparks TN,
Lam G, Kuller J, McMahon M. Use of magnetic resonance imaging and Caughey AB. The effect of cesarean delivery rates on the future inci-
ultrasound in the antenatal diagnosis of placenta accreta. J Soc Gy- dence of placenta previa, placenta accreta, and maternal mortality.
necol Investig 2002;9:37–40. J Matern Fetal Neonatal Med 2011;24:1341–1346.
Lax A, Prince MR, Mennitt KW, Schwebach JR, Budorick NE. The Teo TH, Law YM, Tay KH, Tan BS, Cheah FK. Use of magnetic reso-
value of specific MRI features in the evaluation of suspected nance imaging in evaluation of placental invasion. Clin Radiol 2009;
placental invasion. Magn Reson Imaging 2007;25:87–93. 64:511–516.
Lerner JP, Deane S, Timor-Tritsch IE. Characterization of placenta ac- Thia EW, Lee SL, Tan HK, Tan LK. Ultrasonographical features of
creta using transvaginal sonography and color Doppler imaging. morbidly-adherent placentas. Singapore Med J 2007;48:799–802.
Ultrasound Obstet Gynecol 1995;5:198–201. Thorp JM Jr, Councell RB, Sandridge DA, Wiest HH. Antepartum diag-
Levine D, Hulka CA, Ludmir J, Li W, Edelman RR. Placenta accreta: nosis of placenta previa percreta by magnetic resonance imaging.
Evaluation with color Doppler US, power Doppler US and MR Obstet Gynecol 1992;80:506–508.
imaging. Radiology 1997;205:773–776. Warshak CR, Eskander R, Hull AD, Scioscia AL, Mattrey RF,
Lim PS, Greenberg M, Edelson MI, Bell KA, Edmonds PR, Benirschke K, Resnik R. Accuracy of ultrasonography and magnetic
Mackey AM. Utility of Ultrasound and MRI in Prenatal Diagnosis resonance imaging in the diagnosis of placenta accreta. Obstet Gy-
of Placenta Accreta: A Pilot Study. Am J Roentgenol 2011;197: necol 2006;108:573–581.
1506–1513. Warshak CR, Ramos GA, Eskander R, Benirschke K, Saenz CC,
Masselli G, Brunelli R, Cascian E, Polettini E, Piccioni MG, Kelly TF, Moore TR, Resnik R. Effect of predelivery diagnosis in
Anceschi M, Gualdi G. Magnetic resonance imaging in the evalua- 99 consecutive cases of placenta accreta. Obstet Gynecol 2010;
tion of placental adhesive disorders: Correlation with color Doppler 115:65–69.
ultrasound. Eur Radiol 2008;18:1292–1299. Whiting PF, Rutjes AW, Westwood ME, Mallett S, Deeks JJ,
Moses LE, Shapiro D, Littenberg B. Combining independent studies of Reitsma JB, Leeflang MM, Sterne JA, Bossuyt PM. QUADAS-2:
a diagnostic test into a summary ROC curve: Data analytic A revised tool for the quality assessment of diagnostic accuracy
approaches and some additional considerations. Stat Med 1993;12: studies. Ann Intern Med 2011;155:529–536.
1293–1316. Wortman AC, Alexander JM. Placenta accreta, increta, and percreta.
O’Brien JM, Barton JR, Donaldson ES. The management of placenta Obstet Gynecol Clin North Am 2013;40:137–154.
percreta: Conservative and operative strategies. Am J Obstet Gyne- Zamora J, Abraira V, Muriel A, Khan K, Coomarasamy A. Meta-DiSc:
col 1996;175:1632–1638. A software for meta-analysis of test accuracy data. BMC Med Res
Silver RM, Landon MB, Rouse DJ, Leveno KJ, Spong CY, Thom EA, Methodol 2006;6:31.
Moawad AH, Caritis SN, Harper M, Wapner RJ, Sorokin Y, Zelop CM, Harlow BL, Frigoletto FD Jr, Safon LE, Saltzman DH. Emer-
Miodovnik M, Carpenter M, Peaceman AM, O’Sullivan MJ, gency peripartum hysterectomy. Am J Obstet Gynecol 1993;168:
Sibai B, Langer O, Thorp JM, Ramin SM, Mercer BM. Maternal 1443–1448.