| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Genitourinary Imaging |
1 From the Departments of Radiology (K.K., H.H., Y.K., K.K.Y., Y.L.), Biostatistics (M.R.S.), and Gynecologic Oncology (C.B.P.), University of California-San Francisco, Box 0628, 505 Parnassus Ave, San Francisco, CA 94143-0628. Received September 9, 1998; revision requested October 26; revision received November 25; accepted April 16, 1999. K.K. supported in part by a grant from the French Society of Radiology. K.K.Y. supported in part by a GE-AUR Radiology Research Associate Fellowship. Address reprint requests to H.H. (e-mail: Hedvig.Hricak@radiology.ucsf.edu).
 |
Abstract |
|---|
 TOP Abstract IntroductionMATERIALS AND METHODSRESULTSDISCUSSIONReferences |
|---|
MATERIALS AND METHODS: Data were obtained from a MEDLINE literature search and from manual reviews of article bibliographies. Articles were selected that included results in patients with proved endometrial cancer and imaging-histopathologic correlation and that presented data that allowed calculation of contingency tables. Data for the imaging evaluation of myometrial and cervical invasion were abstracted independently by two authors. Data on year of publication, International Federation of Gynecology and Obstetrics (FIGO) stage distribution, and methodologic quality were also collected. A subgroup analysis was performed to compare contrast medium–enhanced MR imaging with nonenhanced MR imaging, US, and CT.
RESULTS: Six studies met the inclusion criteria for CT; 16, for US; and 25, for MR imaging. Summary receiver operating characteristic analysis showed no significant differences in the overall performance of CT, US, and MR imaging. In the assessment of myometrial invasion, however, contrast-enhanced MR imaging performed significantly better than did nonenhanced MR imaging or US (P < .002) and demonstrated a trend toward better results, as compared with CT. The lack of data on the assessment of cervical invasion at CT or US prevented meta-analytic comparison with data obtained at MR imaging. Results were not influenced by year of publication, FIGO stage distribution, or methodologic quality.
CONCLUSION: Although US, CT, or MR imaging can be used in the pretreatment evaluation of endometrial cancer, contrast-enhanced MR imaging offers "one-stop" examination with the highest efficacy.
Index terms: Computed tomography (CT), comparative studies, 854.12111 • Lymphatic system, neoplasms, 99.33 • Magnetic resonance (MR), comparative studies, 854.1214 • Pelvic organs, CT, 854.12111 • Pelvic organs, MR, 854.1214, 854.12143 • Pelvic organs, US, 854.1298 • Ultrasound (US), comparative studies, 854.1298 • Uterine neoplasms, 854.32 • Uterus, endometrium, 854.32
|
Introduction |
|---|
|
TOPAbstractIntroductionMATERIALS AND METHODSRESULTSDISCUSSIONReferences |
|---|
The diagnostic performance of magnetic resonance (MR) imaging, computed tomography (CT), and ultrasonography (US) in assessing myometrial invasion and cervical extension has been extensively evaluated. Results from studies (7,8) in which MR imaging was compared with CT suggest MR imaging is more accurate than CT. Investigators (9,10) who have compared MR imaging with endovaginal US have reported similar performances for both techniques. However, to our knowledge, only one group of investigators (11) has compared the three techniques in the same patient population (n = 26). In addition, when taken individually, each of these studies lacks the statistical power to determine whether differences between the techniques are significant or represent merely random variation.
Meta-analysis is a statistical method that allows the results of individual studies to be combined and that allows the literature to be reviewed systematically (12,13). We undertook a meta-analysis to compare the three modalities and to define evidence-based imaging guidelines for staging endometrial cancer.
|
MATERIALS AND METHODS |
|---|
|
TOPAbstractIntroductionMATERIALS AND METHODSRESULTSDISCUSSIONReferences |
|---|
To identify additional relevant references, the reference lists of the articles retrieved were checked manually. Unpublished data were not included. A total of 84 articles were found: Sixty-eight were published in English; seven, in Japanese; four, in French; three, in Italian; and two, in German.
Study Selection
The studies included in the meta-analysis met the following inclusion criteria: (a) the patients had a histologic diagnosis of endometrial carcinoma; (b) the standard of reference was surgical staging with histopathologic results; (c) the presented data were obtained by observers who were blinded to the pathologic results; (d) the presented data allowed for the calculation of true-positive, true-negative, false-positive, and false-negative results for imaging tests; and (e) the data or a subset of the data was not published more than once.
Forty-seven of 84 references fulfilled the inclusion criteria (Table 1). Thirty-seven of 84 references were excluded. Reasons for exclusion were incomplete or inconclusive data (n = 25), presurgical radiation therapy after imaging (n = 3), lack of surgical staging with histopathologic results (n = 3), results presented in two publications (n = 3), imaging interpretation with knowledge of pathologic results (n = 2), and patient selection based on imaging quality (selection bias, n = 1).
|
The following data were recorded for each article: (a) sample size; (b) stage distribution, by using FIGO staging guidelines (Table 2); (c) imaging modality (CT, US, MR imaging) and technique (helical vs nonhelical CT, transabdominal vs combined transabdominal and transvaginal US, frequency of US transducer, contrast medium–enhanced vs nonenhanced MR imaging, magnetic field strength [in tesla] of MR imaging equipment); (d) design of imaging interpretation (prospective, retrospective, not specified); (e) diagnostic criteria used for image interpretation; and (f) true-positive, true-negative, false-positive, and false-negative study results for the staging elements of myometrial invasion, cervical invasion, presence of extrauterine disease, and presence of lymph node metastasis.
|
Methodologic Quality Rating
The methodologic quality rating was based on four criteria, as follows: (a) adequate description of the study group (selection criteria were defined, and patient characteristics were presented), (b) adequate description of diagnostic criteria (imaging staging criteria were defined, and staging results were presented accordingly), (c) no potential work-up bias (results of imaging did not influence the decision to perform surgery), and (d) congruence of presented data.
Studies were classified as "high quality" if they met all methodologic requirements, "medium quality" if they met three methodologic requirements, "fair quality" if they met two methodologic requirements, and "not acceptable" if they met only one methodologic requirement.
Data Analysis
The meta-analytic method used in our study was based on summary receiver operating characteristic (ROC) curves (14,15). Sensitivity and specificity were recalculated for each reference study by using the conventional corrections for zero counts (16). Because of the interdependence of sensitivity and specificity (Pearson correlation coefficients of 0.5 for CT, -0.22 for US, and -0.23 for MR imaging), quantitative integration of data by means of standard meta-analytic methods for the comparison of CT, US, and MR imaging was considered inappropriate.
To compare the three imaging modalities, we used summary ROC analysis, which accounts for the interdependence between sensitivity and specificity. Summary ROC, a mathematic transformation of sensitivity and specificity, has been described by Moses et al (14). In addition, transformed data of all studies were combined through a robust regression (Huber M-regression) analysis (17) in a regression line. The robust regression analysis reduces the effect of heterogeneity among studies by attributing appropriate weights to each study according to its deviation from the normal distribution. The regression line is then back-transformed into a summary ROC curve. By combining the data from all studies, summary ROC curves were independent of the diagnostic thresholds that were used to diagnose deep myometrial invasion.
Following the guidelines for fitting summary ROC curves, we obtained corresponding single-number summaries. These were the points on the summary ROC curve where sensitivity and specificity are equal (Q* values). Testing for differences between CT, US, and MR imaging was based on Q* values and on their associated standard errors.
Covariate Adjustment
To determine whether imaging results were significantly affected by heterogeneity in the studies, we extracted covariates, including the severity of disease (FIGO stage distribution in the study population), year of publication, results of methodologic quality rating, and subtype of imaging technique (contrast-enhanced vs nonenhanced MR imaging, transabdominal vs combined transabdominal and transvaginal US). Covariate adjustment analysis was performed by applying a series of statistical tests that followed the method described by Moses et al (14).
|
RESULTS |
|---|
|
TOPAbstractIntroductionMATERIALS AND METHODSRESULTSDISCUSSIONReferences |
|---|
Subsets of data were obtained for the assessment of myometrial invasion in 42 studies, of cervical invasion in 14 studies, of extrauterine disease in six studies, and of lymph node metastasis in one study. Because of the limited number of investigators who evaluated extrauterine disease and lymph node metastasis, meta-analytic comparison between imaging modalities for these staging elements was not possible, and these data sets were, therefore, excluded from the analysis. Studies were performed prospectively (n = 19) or retrospectively (n = 3), or the method of data collection was not specified (n = 25).
Overall Performance
Data sets on myometrial invasion and cervical involvement were combined for the evaluation of overall staging performance for each imaging modality. Because of the interdependence of sensitivity and specificity, we used summary ROC analysis to compare the three imaging modalities. Summary ROC analysis revealed no statistically significant differences in the performance of CT, US, and MR imaging (Fig 1). Q* values were 0.80 (95% CI: 0.62, 0.98) for CT, 0.86 (95% CI: 0.81, 0.91) for US, and 0.87 (95% CI: 0.85, 0.89) for MR imaging.
|
Myometrial Invasion
Data on myometrial invasion met the inclusion criteria for meta-analysis in six of six studies with CT (203 patients, articles published between 1982 and 1995), in 16 of 16 studies with US (611 patients, articles published between 1987 and 1996), and in 20 of 25 studies with MR imaging (742 patients, articles published between 1987 and 1997) (Table 3). All studies of CT included the use of iodinated contrast media administered intravenously; none of the studies included use of helical CT.
|
Among the 20 articles on MR imaging, nine data sets with and 18 data sets without the use of intravenous contrast media were available for meta-analytic comparison. Magnetic field strength was 1.5 T (n = 12); 1.0 T (n = 1); 0.5 T (n = 5); 0.35 T (n = 2); 0.02 T (n = 1); unknown (n = 4); or varied from 0.15 to 1.5 T because the study was multiinstitutional (n = 2).
Q* values were 0.79 (95% CI: 0.61, 0.96) for CT, 0.85 (95% CI: 0.81, 0.88) for US, 0.86 (95% CI: 0.83, 0.89) for MR imaging, 0.83 (95% CI: 0.79, 0.87) for nonenhanced MR imaging, and 0.91 (95% CI: 0.89, 0.92) for contrast-enhanced MR imaging.
Summary ROC curves were obtained for the comparison among CT, US, and MR imaging (Fig 2); between contrast-enhanced MR imaging and nonenhanced MR imaging (Fig 3); and among CT, US, and contrast-enhanced MR imaging (Fig 4). For the comparison among CT, US, and MR imaging, the summary ROC curves were of similar shape and showed no significant differences in performance (Fig 2). Comparison between MR imaging techniques showed significantly better results for contrast-enhanced versus nonenhanced MR imaging (P < .001) (Fig 3). Differences in Q* values did reach statistical significance when contrast-enhanced MR imaging was compared with US (P = .002). The summary ROC curves also showed a trend toward higher diagnostic accuracy for contrast-enhanced MR imaging than for CT (P = .18) (Fig 4).
|
|
|
|
|
DISCUSSION |
|---|
|
TOPAbstractIntroductionMATERIALS AND METHODSRESULTSDISCUSSIONReferences |
|---|
As well-established "primary treatment in a cancer patient gives the best opportunity for cure" (55), preoperative knowledge of local-regional tumor extent may indirectly affect patient survival. Therefore, gaining knowledge of morphologic prognostic factors of cancer and of their inclusion in treatment planning could be an important step in further improving prognosis. Prognostic factors that influence the choice of treatment are patient factors (age, health), tumor extent (depth of myometrial invasion, cervical invasion, and lymph node invasion), and histologic factors (tumor grade, histologic type, lymphatic vascular space invasion) (2).
Histologic factors are related to depth of myometrial invasion, endocervical extension, and lymph node metastasis. A higher tumor grade correlates with lower survival rates and higher rates of myometrial invasion (eg, the incidence of deep myometrial invasion is about 7% for grade I tumors versus 46% for grade III tumors) (1). For grade II tumors with deep myometrial invasion and for all grade III tumors, the lymph node invasion rate is about 31% for pelvic lymph nodes and 23% for paraaortic lymph nodes (2).
Gross evaluation of myometrial invasion becomes less reliable with increasing tumor grade (gross inspection has 80% reliability vs 94% reliability for frozen-section analysis) (56,57). The 5-year mortality rates of 51% for clear cell carcinoma and of 21% for papillary serous carcinoma are worse than the 6% 5-year mortality rate for the more common adenocarcinomas (58). The worse outcome of these histologic types also correlates with a higher incidence of lymph node metastasis and myometrial invasion (eg, 50% of patients with clinical stage I papillary serous carcinoma have deep myometrial invasion) (59). Invasion of the cervical stroma (FIGO stage IIB) and of the lymph nodes (FIGO stage IIIC) are well-known adverse prognostic factors that are strongly correlated with myometrial invasion and tumor grade (1,60). However, diagnosis of cervical invasion by means of endocervical curettage is unreliable because of false-positive rates of up to 25% and false-negative rates of up to 10% (61).
In 1988, the inadequacies of pretreatment diagnostic assessment of endometrial cancer led to recommendations for routine surgical staging (5). However, since the introduction of surgical staging in 1988, the technology for diagnostic radiology has markedly advanced with the improvement of spatial resolution of CT and with the development of endovaginal US, MR imaging, and MR contrast media. Accurate pretreatment assessment of endometrial cancer at imaging can potentially optimize surgical and nonsurgical treatment. Pretreatment knowledge of myometrial or cervical invasion affects the performance and extent of lymphadenectomy (3,62,63). Presence of cervical invasion alters the type of hysterectomy that is performed (64), and the poor outcome of associated parametrial invasion (65) indicates irradiation, rather than immediate surgery, is the best treatment in these patients (2).
Despite the breadth of information provided with modern imaging methods, no imaging guidelines or algorithms are available for the use of imaging in the pretreatment assessment of endometrial cancer. In part, this is due to the lack of multimodality studies with adequate sample size and statistical power in which state-of-the-art imaging techniques are compared. When a single imaging modality is evaluated, the investigator tends to overestimate its value because of a variety of biases. This meta-analysis was undertaken to address some of the shortcomings in the existing literature.
In our meta-analysis of 10–15 years of literature, we attempted to minimize some of the well-known limitations of meta-analysis. If primary studies suffer from inadequate scientific quality, combining them in a meta-analytic comparison will yield only poor results. We attempted to minimize this problem by applying strict inclusion and exclusion criteria to every study and by taking into account a methodologic quality rating. To avoid confounding due to the use of different thresholds for the diagnosis of myometrial invasion, we used the summary ROC method rather than the pooling method that combines the patient populations to determine a combined value of sensitivity and specificity.
Another potential pitfall of meta-analysis is that differences in patients may preclude application of the results of a meta-analysis to the target population. We attempted to avoid this pitfall by assessing the severity of the disease and by performing a covariate analysis, which demonstrated no differences in results when the FIGO stage distribution of each study population was considered.
By including only published data, we did not exclude publication bias, which tends to cause overestimation of diagnostic performance because of the greater likelihood of publication of positive rather than negative results. Although meta-analysis does not replace large prospective studies, the results from a meta-analysis in which results of smaller trials are combined may not differ significantly from the results of larger prospective trials (13).
Findings of this study showed that, in the evaluation of myometrial invasion, contrast-enhanced MR imaging performed significantly better than nonenhanced MR imaging and US (P < .002) and showed a trend toward better results when compared with CT (P = .18).
Although commonly used in clinical settings, to our knowledge there have not been sufficient published data evaluating the performance of CT in endometrial cancer staging. Due to the lack of published data on helical CT in staging endometrial cancer, the performance of CT may be underestimated in this meta-analysis.
Comparison of US studies that use the transabdominal approach and studies that use the endovaginal approach was not possible because of insufficient data on US studies that use only the transabdominal approach. As 88% of the included US studies combined the transabdominal approach with the endovaginal approach, this combination appeared to represent the recommended standard technique in the published literature. Compared with CT, US has a significantly lower sensitivity for paraaortic and pelvic lymph node detection (66,67).
Recently, a meta-analysis in which the use of MR imaging, CT, and lymphography in patients with cervical cancer was compared showed a trend toward better performance with MR imaging than with CT or lymphography in the diagnosis of pelvic lymphadenopathy (68). Although the meta-analysis included only patients with cervical cancer, these results might still apply to patients with endometrial cancer. Therefore, in patients with a large uterine mass and a resultant higher probability of lymph node metastasis, MR imaging or CT, and not US, should be used as the primary imaging modality. Incomplete data on lymph node status in patients with endometrial cancer precluded meta-analytic comparison.
Although six studies compared contrast-enhanced with nonenhanced MR imaging (10,41,42,44,46,48), the difference in accuracy between these techniques reached statistical significance in only one study (10). This meta-analysis increased the statistical significance (from P = .03 in Yamashita's study [10] to P < .001 in this analysis) and therefore confirmed, with higher confidence, the previously published superiority of contrast-enhanced over nonenhanced MR imaging. Staging of endometrial cancer at MR imaging should, therefore, include the use of intravenous contrast media.
The lack of reported data on the assessment of cervical invasion at CT and US may be related to diagnostic difficulties with these imaging modalities. Results of MR imaging for this indication ranged from 66% to 100% (mean, 86%) for sensitivity; from 92% to 100% (mean, 97%) for specificity; and from 87% to 95% for the CIs for the Q* value.
Among 10 articles that included assessment of cervical invasion at MR imaging, the authors of only one article (49) differentiated cervical stromal invasion (the only cervical prognostic factor) from superficial cervical invasion (not a prognostic factor). The authors demonstrated an increase in sensitivity from 56% to 100% and no change in specificity if only cervical stromal invasion was considered a positive sign of cervical invasion. The meta-analytic results for the assessment of cervical invasion at MR imaging could not reflect this differentiation and might, therefore, underestimate the real sensitivity of MR imaging. If the patient is at high risk for cervical invasion, on the basis of findings at clinical examination, contrast-enhanced MR imaging might represent the appropriate cost-effective solution for evaluating myometrial, cervical, and lymph node involvement.
Although the algorithmic approach to diagnostic tests offers the potential for evidence-based assessment, guidelines need to be adopted. Imaging should be used because it offers an important adjunct to treatment decision making. The choice of the primary imaging modality should be tailored to local expertise; to the patient's pretest probability for myometrial, cervical, or lymph node disease; and to the patient's habitus.
For endometrial cancer staging, the choice of an imaging modality depends on tumor grade, histologic type, and findings at clinical examination.
On the basis of our meta-analysis of imaging findings and well-established prognostic factors that are assessed at endometrial biopsy, endocervical curettage, and clinical examination, the following clinical practice guidelines can be formulated for staging endometrial cancer. (a) No imaging is required for patients with grade I tumors and a nonenlarged uterus at physical examination because the pretest probability of myometrial, cervical, or nodal involvement is low. If results from the physical examination are inconclusive or if there is concomitant pelvic disease, US, CT, or MR imaging can be used at the initial radiologic investigation. (b) Patients with high-grade papillary or clear cell tumors should undergo CT or MR imaging because there is a high pretest probability of nodal involvement. (c) Patients with possible cervical involvement at physical examination or with positive or inconclusive results from endocervical curettage should undergo MR imaging, since this is the only modality that has been shown to accurately depict cervical invasion. (d) In patients who require multifactorial assessment, contrast-enhanced MR imaging is the only modality that can be used to accurately evaluate myometrial, cervical, and nodal involvement.
|
Footnotes |
|---|
Author contributions: Guarantors of integrity of entire study, K.K., H.H.; study concepts, K.K., H.H.; study design, K.K., K.K.Y.; definition of intellectual content, K.K., K.K.Y., H.H.; literature research, K.K., Y.K.; data acquisition, K.K., Y.K.; data analysis, K.K., K.K.Y.; H.H., M.R.S., Y.L.; statistical analysis, M.R.S.; manuscript preparation, K.K.; manuscript editing, H.H., C.B.P., K.K.Y.; manuscript review, K.K.Y., H.H., M.R.S.
|
References |
|---|
|
TOPAbstractIntroductionMATERIALS AND METHODSRESULTSDISCUSSIONReferences |
|---|
This article has been cited by other articles:
|
|
 |
|
  M. M. Baekelandt, M. Castiglione, and On behalf of the ESMO Guidelines Working Group Endometrial carcinoma: ESMO Clinical Recommendations for diagnosis, treatment and follow-up Ann. Onc., May 1, 2009; 20(suppl_4): iv29 - iv31. [Full Text] [PDF]
|
|
|
|
 |
|
 G. Lin, K.-K. Ng, C.-J. Chang, J.-J. Wang, K.-C. Ho, T.-C. Yen, T.-I Wu, C.-C. Wang, Y.-R. Chen, Y.-T. Huang, et al. Myometrial Invasion in Endometrial Cancer: Diagnostic Accuracy of Diffusion-weighted 3.0-T MR Imaging--Initial Experience Radiology, March 1, 2009; 250(3): 784 - 792. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 P. Torricelli, S. Ferraresi, F. Fiocchi, G. Ligabue, V. M. Jasonni, I. Di Monte, and F. Rivasi 3-T MRI in the Preoperative Evaluation of Depth of Myometrial Infiltration in Endometrial Cancer Am. J. Roentgenol., February 1, 2008; 190(2): 489 - 495. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 J. N. Bakkum-Gamez, J. Gonzalez-Bosquet, N. N. Laack, A. Mariani, and S. C. Dowdy Current Issues in the Management of Endometrial Cancer Mayo Clin. Proc., January 1, 2008; 83(1): 97 - 112. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
E. Sala, S. Wakely, E. Senior, and D. Lomas MRI of Malignant Neoplasms of the Uterine Corpus and Cervix Am. J. Roentgenol., June 1, 2007; 188(6): 1577 - 1587. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 A. Sarna, B. G. Coleman, S. K. Srinivas, and D. A. Torigian Computed Tomographic and Ultrasonographic Findings of Endometrial Carcinoma Appearing as a Fungating Inguinal Mass J. Ultrasound Med., November 1, 2005; 24(11): 1579 - 1582. [Full Text] [PDF]
|
|
|
|
 |
|
 M. M. Leitao Jr and D. S. Chi Fertility-Sparing Options for Patients with Gynecologic Malignancies Oncologist, September 1, 2005; 10(8): 613 - 622. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
R. Manfredi, P. Mirk, G. Maresca, P. A. Margariti, A. Testa, G. F. Zannoni, D. Giordano, G. Scambia, and P. Marano Local-Regional Staging of Endometrial Carcinoma: Role of MR Imaging in Surgical Planning Radiology, May 1, 2004; 231(2): 372 - 378. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 K. Fong, P. Causer, M. Atri, A. Lytwyn, and R. Kung Transvaginal US and Hysterosonography in Postmenopausal Women with Breast Cancer Receiving Tamoxifen: Correlation with Hysteroscopy and Pathologic Study RadioGraphics, January 1, 2003; 23(1): 137 - 150. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
L. A. Hardesty, J. H. Sumkin, C. Hakim, C. Johns, and M. Nath The Ability of Helical CT to Preoperatively Stage Endometrial Carcinoma Am. J. Roentgenol., March 1, 2001; 176(3): 603 - 606. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
K. A. Frei, K. Kinkel, H. M. Bonél, Y. Lu, C. Zaloudek, and H. Hricak Prediction of Deep Myometrial Invasion in Patients with Endometrial Cancer: Clinical Utility of Contrast-enhanced MR Imaging-A Meta-analysis and Bayesian Analysis Radiology, August 1, 2000; 216(2): 444 - 449. [Abstract] [Full Text]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| RADIOLOGY | RADIOGRAPHICS | RSNA JOURNALS ONLINE |
