Diagnostic Performance of Nanoparticle-Enhanced Magnetic Resonance Imaging in the Diagnosis of Lymph Node Metastases in Patients With Endome
http://www.100md.com
《临床肿瘤学》
the Departments of Radiology, Gyneoncology and Histopathology, St Bartholomew's Hospital
Department of Radiology, Royal Marsden Hospital, London, United Kingdom
Department of Radiology, Massachusetts General Hospital, Boston, MA
ABSTRACT
PURPOSE: Lymph node metastases affect management and prognosis of patients with gynecologic malignancies. Preoperative nodal assessment with computed tomography or magnetic resonance imaging (MRI) is inaccurate. A new lymph node–specific contrast agent, ferumoxtran-10, composed of ultrasmall particles of iron oxide (USPIO), may enhance the detection of lymph node metastases independent of node size. Our aim was to compare the diagnostic performance of MRI with USPIO against standard size criteria.
METHODS: Forty-four patients with endometrial (n = 15) or cervical (n = 29) cancer were included. MRI was performed before and after administration of USPIO. Two independent observers viewed the MR images before lymph node sampling. Lymph node metastases were predicted using size criteria and USPIO criteria. Lymph node sampling was performed in all patients.
RESULTS: Lymph node sampling provided 768 pelvic or para-aortic nodes for pathology, of which 335 were correlated on MRI; 17 malignant nodes were found in 11 of 44 patients (25%). On a node-by-node basis, sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) by size criteria were 29%*, 99%, 56%, and 96%, and by USPIO criteria (reader 1/reader 2) were 93%/82%* (*P = .008/.004), 97%/97%, 61%/59%, and 100%/99%, respectively (where [*] indicates the statistical difference of P = x/x between the two results marked by the asterisk). On a patient-by-patient basis, sensitivity, specificity, PPV, and NPV by size criteria were 27%*, 94%, 60%, and 79%, and by USPIO criteria (reader 1/reader 2) were 100%/91%* (*P = .031/.06), 94%/87%, 82%/71%, and 100%/96%, respectively. The statistic was 0.93.
CONCLUSION: Lymph node characterization with USPIO increases the sensitivity of MRI in the prediction of lymph node metastases, with no loss of specificity. This may greatly improve preoperative treatment planning.
INTRODUCTION
In patients with cervical and endometrial cancer, the presence of lymph node metastases suggests poor prognosis, with a marked decrease in 5-year survival rate.1,2 Lymph node involvement is also an important factor in the choice of adjuvant radiotherapy in both endometrial and cervical cancer.3 Surgical lymph node assessment is the gold standard for the diagnosis of lymph node metastases.4 However, surgical lymphadenectomy is specialized and increases the time and cost of the procedure, with an increased risk of immediate and delayed complications to the patient.5-7 Therefore, a noninvasive technique that accurately identifies lymph node metastases would be beneficial.
Computed tomography (CT) and magnetic resonance imaging (MRI) may be used to predict lymph node metastases. However, there are no reliable discernible morphologic criteria to differentiate benign nodes from those containing metastases. Size criteria may predict the likelihood of nodal metastases, using a threshold value above which the node is considered malignant. However, nodes that are smaller than the threshold may harbor metastatic deposits. Conversely, a node that is larger than the threshold value may be benign. This problem is reflected in the diagnostic performance of MRI and CT in lymph node staging, in which sensitivities for lymph node metastases in gynecologic cancers range from 43% to 73%.8-11
A lymph node–specific MR contrast agent has been developed that allows the identification of malignant nodal infiltration independent of the lymph node size. This novel MR contrast agent is classified as a nanoparticle (mean diameter, 30 nm), and is composed of an iron oxide core, coated with low molecular weight dextran. The class of these MR contrast agents is collectively known as ultrasmall particles of iron oxide (USPIO). The particles, which are administered intravenously, are taken up by macrophages in the reticuloendothelial system, predominantly within the lymph nodes.12 Uptake of USPIO results in marked loss of signal intensity (darkening) of the node on T2- and T2*-weighted sequences because of a susceptibility artifact caused by the iron. Metastatic tissue within a node displaces the normal macrophages, thus preventing uptake of contrast agent, and the node continues to remain high in signal intensity.
The aim of this study was to assess the diagnostic performance of USPIO-enhanced MRI (USPIO-MRI) in predicting nodal metastases within pelvic and para-aortic lymph nodes in patients with cancer of the endometrium and cervix.
PATIENTS AND METHODS
Patients
The local research ethics committee approved the study. Off-license use of USPIO (Sinerem, Guerbet, France) was approved by the Medical Controls Agency, United Kingdom. All patients were older than 18 years and provided written informed consent. All patients were recruited at two large tertiary care academic centers.
Patients were enrolled if they had confirmed cervical or endometrial carcinoma and surgical lymphadenectomy was planned. Exclusion criteria included a history of allergy to iron treatment or a previous allergy to USPIO. Forty-four patients completed the protocol (Table 1). Fourteen additional patients (eight with cervical carcinoma and six with endometrial carcinoma) were recruited but did not complete the protocol: six were treated nonsurgically, five had surgery but no lymphadenectomy, and three did not complete the USPIO infusion (see Administration of Contrast).
Imaging
All patients were imaged as part of their standard preoperative assessment. T1-weighted axial images were obtained though the upper abdomen and pelvis (repetition time in msec [TR], 500 to 540; echo time in msec [TE], 8 to 9; slice width, 4 to 8 mm; interslice gap, 0 to 2 mm; number of acquisitions, two; field of view, 30 x 30 to 42 x 42 cm; matrix, 256 x 256); T2-weighted axial, sagittal, and oblique axial images were obtained through the pelvis (TR, 3,333 to 7,058; TE, 80 to 120 effective time; echo train length [ETL], 8; slice width, 3 to 4 mm; interslice gap, 0 to 1 mm; number of acquisitions, two; field of view, 24 x 24 cm; matrix, 256 x 256 to 512 x 512). Additional nodal imaging sequences included T2*-weighted gradient echo images in axial and oblique (aligned along the psoas muscle and external iliac vessels) planes (TR, 500 to 800; TE, 15 to 27; flip angle [FA], 30; slice width, 3.5 to 5 mm; interslice gap, 0 mm; number of acquisitions, two; field of view, 24 x 24 to 42 x 42 cm; matrix, 256 x 256 to 512 x 512). At 24 to 36 hours after the administration of USPIO, the T2- and T2*-weighted sequences used for the interpretation of lymph node status were repeated. Imaging was performed on a Signa 1.5T (GE Medical Systems, Milwaukee, WI) in 31 patients, Gyroscan Intera 1.5T system (Philips, Eindhoven, the Netherlands) in 11 patients, and a Symphony 1.5T (Siemens, Erlangen, Germany) in two patients.
Administration of Contrast
After the preliminary MRI scan, USPIO contrast medium (Sinerem) was administered intravenously. Each vial of contrast, containing 210 mg of USPIO, was reconstituted with 10 mL of normal saline. The dose of 2.6 mg iron/kg (0.13 mL/kg) was then diluted in 100 mL of normal saline and infused over 30 minutes under supervision. Three patients did not complete the USPIO infusion (one patient had a vasovagal reaction with fainting, one patient had suspected contrast reaction as evidenced by the onset of a tight cough at the beginning of the infusion, and one patient developed a marked rash and flushing). These patients had no significant adverse clinical events and recovered fully after a short period of observation. Six other patients (who completed the protocol) had minor adverse effects including erythematous rash (n = 5), muscle cramps (n = 1), diarrhea (n = 1), and indigestion (n = 1). All symptoms resolved spontaneously 1 to 4 hours after the infusion.
Image Interpretation
Two observers (readers 1 and 2), with no previous experience reading USPIO-MRI, interpreted the images prospectively. Learning curve patients (reader 1, three patients; reader 2, two patients) were not blinded and were excluded from the readers' results. A third reader, experienced in USPIO-MRI, was asked to review a random number of scans (29 of 44) independently to assess interobserver variability.
Size Criteria
The short-axis and long-axis diameters of each identifiable node were measured using electronic calipers on the scanner console. Node diagnosis was made using two standard size criteria: for short-axis criteria, short-axis diameter more than 10 mm indicated metastasis13; for size ratio criteria, nodes less than 8 mm short axis were considered benign, nodes more than 10 mm short axis were considered metastatic, and for nodes with a short axis between 8 and 10 mm, if the ratio of the short to long axis was more than 0.8 (ie, a round node), then the node was considered positive.14,15
Several other threshold values for short-axis diameter were also applied above which the node was considered positive: more than 5, 8, and 9 mm, respectively.16
USPIO Criteria
Nodes were evaluated on the pre- and post-USPIO images. USPIO criteria for benignity were homogeneous or slightly heterogeneous complete darkening (Fig 1); central darkening with a smooth rim of nondarkening (Fig 2); and darkening of nodal tissue with central area of nondarkening, which corresponded to fatty hilum on T1 images. USPIO criteria for malignant infiltration were only faint darkening, but with a decrease in signal intensity of less than 50% (Fig 3); focal area of high signal that did not correspond to fatty hilum on T1 (Fig 4); and complete uniform retention of high signal (Fig 5). For receiver operating characteristic (ROC) curve analysis, a five-point scale was used to indicate the level of confidence in the diagnosis: 1 indicated definitely benign, 2 indicated probably benign, 3 indicated indeterminate, 4 indicated probably malignant, and 5 indicated definitely malignant. Nodes in groups 1 and 2 were classified as benign and masses in groups 3 to 5 were classified as malignant for sensitivity and specificity calculations.
Surgery
Nodal status was discussed with the surgeon before lymphadenectomy. Careful identification of the nodes was made at the time of surgery, using a combination of the MR images and diagrams or with the attendance of the radiologist in the operating room. Resected nodes were anatomically labeled. In three patients, image-guided percutaneous biopsy or fine needle aspirate of a lymph node was performed, in accordance with clinical preference (Table 1).
Histopathology
Lymph nodes were cut into parallel slices of 2 to 3 mm thickness and all nodal tissue was routinely processed and embedded in paraffin. Sections were stained with hematoxylin and eosin and reported by a histopathologist with a specialist interest in gynecologic malignancy. One hundred twenty-two nodes (16%) were reviewed using serial sectioning with examination of at least five additional levels, one of which was stained immunohistochemically for cytokeratin, using the broad-spectrum anticytokeratin antibody MNF-116 (Dako, Cambridge, UK) in a 1:50 dilution using a standardized avidin-biotin complex system.
Statistics
Data were entered on an Statistical Package for the Social Sciences version 11 software package (SPSS Inc, Chicago, IL). The McNemar test was used to compare the sensitivity (in node-positive patients) and specificity (in node-negative patients) obtained with and without USPIO. Intrapatient correlation calculations and variance correction for clustered binary outcome variables were undertaken according to the methods proposed by Fleiss17 and Donner and Klar.18 Given that these methods become inaccurate when sample sizes are small, exact binomial confidence limits were therefore calculated for sensitivity and the positive predictive value. Exact binomial confidence limits were calculated for the patient by patient analysis (Table 2).
In differentiating between benign and malignant nodes, ROC curve analysis was also performed. The confidence scores of the readers 1 and 2 in identifying benign and malignant nodes were compared with the final histologic diagnosis using ROC analysis. ROC curves were also calculated for short-axis diameter in millimeters and the size ratio criteria. The mean areas under the ROC curves were determined. The ROC curve analysis was performed on nodes with complete data sets, thereby excluding the five learning curve cases in which confidence scores were only available from one reader in each sample. Thus 279 nodes were suitable for inclusion in the ROC curve analysis. Interobserver variability for USPIO-MRI was calculated using the statistic. Interobserver variability for nodal size was measured using the mean coefficient of variability.
RESULTS
A total of 768 lymph nodes sampled from the pelvic side wall and para-aortic regions were available for pathology, 17 of which were malignant (2.2%). A mean of 17.5 nodes were harvested per patient (range, one to 41); 11 of 44 patients (25%) had one or more positive nodes and 335 nodes could be anatomically correlated with the magnetic resonance images. The diagnostic performance of size criteria and USPIO criteria were assessed on a node-by-node basis (Table 3) and on a patient-by-patient basis (where a patient is considered to have nodal involvement if one or more nodes are diagnosed as malignant; Table 2). On a node-by-node basis, the sensitivity for the detection of malignant nodes using the size ratio was 0.29 (95% CI, 0.10 to 0.56); the sensitivity using USPIO criteria for reader 1 was 0.93 (95% CI, 0.66 to 1.00; P = .008), and the sensitivity for reader 2 was 0.82 (95% CI, 0.57 to 0.96; P = .004). On a patient-by-patient basis, the sensitivity using size ratio was 0.27 (95% CI, 0.06 to 0.61); the sensitivity using USPIO criteria for reader 1 was 1.00 (95% CI, 0.66 to 1.00; P = .03), and the sensitivity for reader 2 was 0.91 (95% CI, 0.59 to 1.00; P = .06).
ROC curve analysis (Fig 6) shows the area under the curve for short-axis diameter was 0.832 (95% CI, 0.686 to 0.978); for size ratio, the area under the curve was 0.621 (95% CI, 0.391 to 0.851). For USPIO-MRI reader 1, the area under the curve was 0.981 (95% CI, 0.966 to 0.997) and for reader 2, the area under the curve was 0.991 (95% CI, 0.981 to 1.001). Therefore, overall diagnostic performance of readers 1 and 2 using USPIO-MRI is better than that obtained using size ratio.
In four of the 44 patients, several small parametrial nodes were identified within the tissue block of the resected uterus, separate from lymph node dissection. These four patients had a primary diagnosis of cervical carcinoma. Seven of these nodes contained metastatic deposits. These nodes could not be identified in retrospect on the MR images either before or after USPIO administration. If these nodes are included in the node-by-node analysis, then the sensitivity for size criteria using the size ratio was 0.21 (95% CI, 0.07 to 0.42); sensitivity for reader 1 using USPIO criteria was 0.72 (95% CI, 0.46 to 0.90; P = .008), and sensitivity for reader 2 was 0.58 (95% CI, 0.37 to 0.78; P = .004), with a total of 10 false-negative nodal diagnoses (seven parametrial nodes and three pelvic side-wall nodes containing micrometastases, confirmed on immunostains). On a patient-by-patient analysis, when including the parametrial nodes, the sensitivity using the size ratio was 0.25 (95% CI, 0.05 to 0.57); using USPIO criteria, sensitivity for reader 1 was 0.90 (95% CI, 0.55 to 1.00; P = .03), and sensitivity for reader 2 was 0.83 (95% CI, 0.52 to 0.98; P = .06). Altogether, there were two patients with false-negative results: one patient with a false-negative result had two side-wall nodes containing micrometastases as well as three positive parametrial nodes (this was a learning curve patient for reader 1); the second patient with a false-negative result had a single, 4-mm paracervical/parametrial node that could not be seen in retrospect. The remainder of the 25 nodes dissected in this patient were correctly predicted to be benign on USPIO-MRI.
False-positive diagnoses on USPIO-MRI (three patients, reader 1; five patients, reader 2) occurred in the following situations. A false-positive result was obtained for a patient in whom extensive local and distant recurrent disease was diagnosed on follow-up CT scan within 3 months of her surgery. This suggests that the positive diagnosis on USPIO-MRI may have been correct but the node suggestive of metastasis was not surgically resected, even though 41 nodes were collected. A false-positive result was obtained in two patients in whom at least one node was histologically positive but in whom USPIO-MRI diagnosed additional nodes. A false-positive result was obtained in four patients in whom no nodal metastasis was present histologically and in whom there was no recurrent nodal disease within 6 months of follow-up. In these latter four patients, misinterpretation was due to a fatty hilum (one patient) heterogeneous signal within a node (two patients); in the fourth patient, all three readers diagnosed metastatic para-aortic nodes with a confidence score of 5, but the patient had concomitant infected pelvic lymphoceles. It is possible that the presence of lymphoceles, secondary to an extensive lymph node dissection in the previous 6 months, may have affected contrast uptake because of disruption of the normal lymph channels. Currently, we do not know what effect previous lymph node dissection of lymphocele has on USPIO uptake.
Interobserver variability (Table 4) was good to very good. The mean coefficient of variability for lymph node size measurements between reader 1 and reader 2 was 0.06 (standard deviation, 0.05).
DISCUSSION
Our results demonstrate a significant increase in the sensitivity, with no loss of specificity, for the noninvasive detection of malignant lymph nodes in patients with cervical and endometrial cancer. This was seen on a node-by-node basis, for which the sensitivity increased from 29% using standard size criteria to 93% using USPIO criteria. This increase in sensitivity was seen to a greater extent on a patient-by-patient basis, for which the sensitivity increased from 27% using standard size criteria to 100% using USPIO criteria. Only one patient with cervical cancer metastatic to lymph nodes along the pelvic side wall had a false-negative result using USPIO criteria by one reader. One patient with a single 4-mm metastatic node within the parametrial tissues also had a false-negative result. All patients with endometrial cancer metastatic to lymph nodes were correctly identified.
Nodal size is the current standard criterion for the prediction of metastases to lymph nodes on cross-sectional imaging. There is some variability in the size threshold used by different groups, but the most widely used threshold value for identifying malignant lymph nodes in the literature is 10 mm short-axis diameter.8,19,20 Our results demonstrate a clear improvement when compared with this threshold level. More recently, a size ratio has been used whereby indeterminate-sized nodes (8 to 10 mm) that are round are considered malignant.14 Our results show a clear improvement when compared with this threshold. Setting a lower short-axis diameter threshold value for diagnosis of malignant nodes improves sensitivity but with a corresponding decrease in the specificity. This is reflected in the ROC curve analysis in which the area under the curve for using short-axis diameter as the diagnostic tool was 0.834 compared with 0.991 using USPIO criteria (although close confidence intervals were observed), reflecting the small number of positive nodes in our group of patients.
Contrast administration was generally well tolerated; flushing or a rash was the most frequent symptom. Contrast administration was stopped in 5% of patients, as a precaution, but no patient had any significant sequelae as a result of the infusion. The MRI sequences used in our study are widely available on most MR systems independent of vendor platforms, and image analysis does not require any additional software.
False-negative diagnoses were predominantly due to our failure to identify nodes in the parametrial tissues, both before and after USPIO administration, as well as in retrospect. This may be due to limitations of our MRI technique, in which we set out specifically to diagnose the pelvic side-wall nodes, which are well seen on MRI and are the traditional focus for radiologists attempting preoperative lymph node diagnosis. Future improvements in technique may help prevent this problem. However, the parametrial tissues frequently contain an extensive plexus of serpiginous veins, and the identification of parametrial nodes is likely to remain challenging. Currently, the parametrial tissues are removed at primary surgery in most patients with cervical carcinoma; thus, preoperative diagnosis at this site may not have an impact on decision making. No positive parametrial nodes were diagnosed at histology in patients with endometrial carcinoma. This may be due to a difference in surgical technique: endometrial carcinoma is not commonly treated by radical hysterectomy; it is usually treated by simple hysterectomy, and thus the parametrial tissues are not available for histologic assessment. Micrometastases are also likely to remain a challenge because of limitations in the spatial resolution of MRI systems.
False-positive diagnoses may reflect incomplete nodal dissection, in which a node may have been inaccessible to or not seen by the surgeon. In these patients, it may be postulated that USPIO-MRI provides a more complete anatomic map of the lymph node involvement. However, false-positive interpretation also occurred where a node seemed heterogeneous or where a fatty hilum was misinterpreted as a deposit. In addition, the presence of bilateral infected pelvic lymphoceles in one patient may have altered the contrast uptake into the nodes. These misinterpretations may be related to learning curves or possibly due to technical factors.
Our findings support earlier published results of USPIO-MRI in other tumors, either in head and neck or breast, or in patients with a tumor in the abdomen or pelvis.21-24 A recently published study restricted to patients with prostate cancer showed the sensitivity for nodal staging using USPIO criteria was 90.5%—significantly greater than that achieved using size criteria, which had a sensitivity of 35.4%.15 On a patient-by-patient basis, the same study had a sensitivity of 100% for detecting patients with one or more positive nodes.
USPIO-MRI has the potential to improve the surgical and nonsurgical management of patients with cervical and endometrial carcinoma in several ways. First, the preoperative localization of malignant nodes may allow tailoring of the surgical lymphadenectomy, allowing rapid collection of any node suggestive of metastasis for frozen section, and by directing the surgeon to a node suggestive of metastasis at a site not usually accessed. This tailored approach may reduce both surgical time and morbidity without a reduction in diagnostic yield. Second, when USPIO-MRI is negative, surgical lymph node sampling could be avoided altogether, with a high degree of confidence. This could be of particular value when surgical risks such as obesity and diabetes are present. Third, USPIO-MRI could accurately map the extent of lymph node involvement in both the pelvis and para-aortic regions to define the radiotherapy fields. This is particularly relevant with the development of intensity-modulated radiotherapy, in which the shape of the field may be closely tailored to the individual patient.
In conclusion, USPIO-MRI improves the preoperative characterization of lymph node metastases compared with standard MRI. This information may provide the clinician with important information in planning the optimal surgical or radiotherapy treatment.
Appendix
Andrea Rockall obtained ethics approval, Medical Control Agency approval, and BUPA Foundation Grant (London, UK); recruited patients; interpreted images, analyzed data, and was the primary author of the manuscript. S. Aslam Sohaib recruited patients at the Royal Marsden Hospital, interpreted images, prepared the manuscript, and analyzed data. M. Harisinghani provided advice on MRI sequences, interpreted images, and provided advice on the manuscript. S. Babar recruited patients and assisted with data collection. N. Singh performed node histopathology. I. Jacobs, D. Oram, A. Jeyarajah, and J. Shepherd sampled surgical lymph nodes and provided advice on the manuscript. R. Reznek acted as the senior supervisor and performed data interpretation.
Authors' Disclosures of Potential Conflicts of Interest
The authors indicated no potential conflicts of interest.
Acknowledgment
We thank Peter A. Armstrong for reading and commenting on the manuscript and Roger A'Hern for statistical advice.
NOTES
Supported by the BUPA Foundation, London, United Kingdom.
Presented in abstract form at the 2nd Annual Meeting of the International Cancer Imaging Society, Paris, France, October 2002, 15th Annual European Congress of Radiology, Vienna, Austria, March 7-11, 2003, and the 89th Scientific Assembly and Annual Meeting of the Radiological Society of North America, Chicago, IL, November 30-December 5, 2003.
Authors' disclosures of potential conflicts of interest are found at the end of this article.
REFERENCES
Stehman FB, Bundy BN, DiSaia PJ, et al: Carcinoma of the cervix treated with radiation therapy: I. A multi-variate analysis of prognostic variables in the Gynecologic Oncology Group. Cancer 67:2776-2785, 1991
Manetta A, Delgado G, Petrilli E, et al: The significance of paraaortic node status in carcinoma of the cervix and endometrium. Gynecol Oncol 23:284-290, 1986
Cohn DE, Peters WA III, Muntz HG, et al: Adenocarcinoma of the uterine cervix metastatic to lymph nodes. Am J Obstet Gynecol 178:1131-1137, 1998
Shepherd JH: Revised FIGO staging for gynaecological cancer. Br J Obstet Gynaecol 96:889-892, 1989
Hertel H, Kohler C, Elhawary T, et al: Laparoscopic staging compared with imaging techniques in the staging of advanced cervical cancer. Gynecol Oncol 87:46-51, 2002
Schlaerth JB, Spirtos NM, Carson LF, et al: Laparoscopic retroperitoneal lymphadenectomy followed by immediate laparotomy in women with cervical cancer: A gynecologic oncology group study. Gynecol Oncol 85:81-88, 2002
Agostini A, Carcopino X, Franchi F, et al: Port site metastasis after laparoscopy for uterine cervical carcinoma. Surg Endosc 17:1663-1665, 2003
Reinhardt MJ, Ehritt-Braun C, Vogelgesang D, et al: Metastatic lymph nodes in patients with cervical cancer: Detection with MR imaging and FDG PET. Radiology 218:776-782, 2001
Yu KK, Hricak H, Subak LL, et al: Preoperative staging of cervical carcinoma: Phased array coil fast spin-echo versus body coil spin-echo T2-weighted MR imaging. AJR Am J Roentgenol 171:707-711, 1998
Bipat S, Glas AS, van der Velden J, et al: Computed tomography and magnetic resonance imaging in staging of uterine cervical carcinoma: A systematic review. Gynecol Oncol 91:59-66, 2003
Heller PB, Maletano JH, Bundy BN, et al: Clinical-pathologic study of stage IIB, III, and IVA carcinoma of the cervix: Extended diagnostic evaluation for paraaortic node metastasis—A Gynecologic Oncology Group study. Gynecol Oncol 38:425-430, 1990
Vassallo P, Matei C, Heston WD, et al: AMI-227-enhanced MR lymphography: Usefulness for differentiating reactive from tumor-bearing lymph nodes. Radiology 193:501-506, 1994
Reinhardt MJ, Technau-Ihling K, Altehoefer C, et al: Lymphangiography causes false-positive findings on 18F-FDG PET imaging. Anticancer Res 23:2941-2944, 2003
Jager GJ, Barentsz JO, Oosterhof GO, et al: Pelvic adenopathy in prostatic and urinary bladder carcinoma: MR imaging with a three-dimensional TI-weighted magnetization-prepared-rapid gradient-echo sequence. AJR Am J Roentgenol 167:1503-1507, 1996
Harisinghani MG, Barentsz J, Hahn PF, et al: Noninvasive detection of clinically occult lymph-node metastases in prostate cancer. N Engl J Med 348:2491-2499, 2003
Vinnicombe SJ, Norman AR, Nicolson V, et al: Normal pelvic lymph nodes: Evaluation with CT after bipedal lymphangiography. Radiology 194:349-355, 1995
Fleiss JL: Statistical methods for rates and proportions, 2nd edition. New York, NY, John Wiley and Sons, 1981
Donner A, Klar N: Methods for comparing event rates in intervention studies when the unit of allocation is a cluster. Am J Epidemiol 140:279-289, 1994
Lin WC, Hung YC, Yeh LS, et al: Usefulness of (18)F-fluorodeoxyglucose positron emission tomography to detect para-aortic lymph nodal metastasis in advanced cervical cancer with negative computed tomography findings. Gynecol Oncol 89:73-76, 2003
Grigsby PW, Siegel BA, Dehdashti F: Lymph node staging by positron emission tomography in patients with carcinoma of the cervix. J Clin Oncol 19:3745-3749, 2001
Anzai Y, Piccoli CW, Outwater EK, et al: Evaluation of neck and body metastases to nodes with ferumoxtran 10-enhanced MR imaging: Phase III safety and efficacy study. Radiology 228:777-788, 2003
Stets C, Brandt S, Wallis F, et al: Axillary lymph node metastases: A statistical analysis of various parameters in MRI with USPIO. J Magn Reson Imaging 16:60-68, 2002
Sigal R, Vogl T, Casselman J, et al: Lymph node metastases from head and neck squamous cell carcinoma: MR imaging with ultrasmall superparamagnetic iron oxide particles (Sinerem MR): Results of a phase-III multicenter clinical trial. Eur Radiol 12:1104-1113, 2002
Pannu HK, Wang KP, Borman TL, et al: MR imaging of mediastinal lymph nodes: Evaluation using a superparamagnetic contrast agent. J Magn Reson Imaging 12:899-904, 2000(Andrea G. Rockall, Syed A)
Department of Radiology, Royal Marsden Hospital, London, United Kingdom
Department of Radiology, Massachusetts General Hospital, Boston, MA
ABSTRACT
PURPOSE: Lymph node metastases affect management and prognosis of patients with gynecologic malignancies. Preoperative nodal assessment with computed tomography or magnetic resonance imaging (MRI) is inaccurate. A new lymph node–specific contrast agent, ferumoxtran-10, composed of ultrasmall particles of iron oxide (USPIO), may enhance the detection of lymph node metastases independent of node size. Our aim was to compare the diagnostic performance of MRI with USPIO against standard size criteria.
METHODS: Forty-four patients with endometrial (n = 15) or cervical (n = 29) cancer were included. MRI was performed before and after administration of USPIO. Two independent observers viewed the MR images before lymph node sampling. Lymph node metastases were predicted using size criteria and USPIO criteria. Lymph node sampling was performed in all patients.
RESULTS: Lymph node sampling provided 768 pelvic or para-aortic nodes for pathology, of which 335 were correlated on MRI; 17 malignant nodes were found in 11 of 44 patients (25%). On a node-by-node basis, sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) by size criteria were 29%*, 99%, 56%, and 96%, and by USPIO criteria (reader 1/reader 2) were 93%/82%* (*P = .008/.004), 97%/97%, 61%/59%, and 100%/99%, respectively (where [*] indicates the statistical difference of P = x/x between the two results marked by the asterisk). On a patient-by-patient basis, sensitivity, specificity, PPV, and NPV by size criteria were 27%*, 94%, 60%, and 79%, and by USPIO criteria (reader 1/reader 2) were 100%/91%* (*P = .031/.06), 94%/87%, 82%/71%, and 100%/96%, respectively. The statistic was 0.93.
CONCLUSION: Lymph node characterization with USPIO increases the sensitivity of MRI in the prediction of lymph node metastases, with no loss of specificity. This may greatly improve preoperative treatment planning.
INTRODUCTION
In patients with cervical and endometrial cancer, the presence of lymph node metastases suggests poor prognosis, with a marked decrease in 5-year survival rate.1,2 Lymph node involvement is also an important factor in the choice of adjuvant radiotherapy in both endometrial and cervical cancer.3 Surgical lymph node assessment is the gold standard for the diagnosis of lymph node metastases.4 However, surgical lymphadenectomy is specialized and increases the time and cost of the procedure, with an increased risk of immediate and delayed complications to the patient.5-7 Therefore, a noninvasive technique that accurately identifies lymph node metastases would be beneficial.
Computed tomography (CT) and magnetic resonance imaging (MRI) may be used to predict lymph node metastases. However, there are no reliable discernible morphologic criteria to differentiate benign nodes from those containing metastases. Size criteria may predict the likelihood of nodal metastases, using a threshold value above which the node is considered malignant. However, nodes that are smaller than the threshold may harbor metastatic deposits. Conversely, a node that is larger than the threshold value may be benign. This problem is reflected in the diagnostic performance of MRI and CT in lymph node staging, in which sensitivities for lymph node metastases in gynecologic cancers range from 43% to 73%.8-11
A lymph node–specific MR contrast agent has been developed that allows the identification of malignant nodal infiltration independent of the lymph node size. This novel MR contrast agent is classified as a nanoparticle (mean diameter, 30 nm), and is composed of an iron oxide core, coated with low molecular weight dextran. The class of these MR contrast agents is collectively known as ultrasmall particles of iron oxide (USPIO). The particles, which are administered intravenously, are taken up by macrophages in the reticuloendothelial system, predominantly within the lymph nodes.12 Uptake of USPIO results in marked loss of signal intensity (darkening) of the node on T2- and T2*-weighted sequences because of a susceptibility artifact caused by the iron. Metastatic tissue within a node displaces the normal macrophages, thus preventing uptake of contrast agent, and the node continues to remain high in signal intensity.
The aim of this study was to assess the diagnostic performance of USPIO-enhanced MRI (USPIO-MRI) in predicting nodal metastases within pelvic and para-aortic lymph nodes in patients with cancer of the endometrium and cervix.
PATIENTS AND METHODS
Patients
The local research ethics committee approved the study. Off-license use of USPIO (Sinerem, Guerbet, France) was approved by the Medical Controls Agency, United Kingdom. All patients were older than 18 years and provided written informed consent. All patients were recruited at two large tertiary care academic centers.
Patients were enrolled if they had confirmed cervical or endometrial carcinoma and surgical lymphadenectomy was planned. Exclusion criteria included a history of allergy to iron treatment or a previous allergy to USPIO. Forty-four patients completed the protocol (Table 1). Fourteen additional patients (eight with cervical carcinoma and six with endometrial carcinoma) were recruited but did not complete the protocol: six were treated nonsurgically, five had surgery but no lymphadenectomy, and three did not complete the USPIO infusion (see Administration of Contrast).
Imaging
All patients were imaged as part of their standard preoperative assessment. T1-weighted axial images were obtained though the upper abdomen and pelvis (repetition time in msec [TR], 500 to 540; echo time in msec [TE], 8 to 9; slice width, 4 to 8 mm; interslice gap, 0 to 2 mm; number of acquisitions, two; field of view, 30 x 30 to 42 x 42 cm; matrix, 256 x 256); T2-weighted axial, sagittal, and oblique axial images were obtained through the pelvis (TR, 3,333 to 7,058; TE, 80 to 120 effective time; echo train length [ETL], 8; slice width, 3 to 4 mm; interslice gap, 0 to 1 mm; number of acquisitions, two; field of view, 24 x 24 cm; matrix, 256 x 256 to 512 x 512). Additional nodal imaging sequences included T2*-weighted gradient echo images in axial and oblique (aligned along the psoas muscle and external iliac vessels) planes (TR, 500 to 800; TE, 15 to 27; flip angle [FA], 30; slice width, 3.5 to 5 mm; interslice gap, 0 mm; number of acquisitions, two; field of view, 24 x 24 to 42 x 42 cm; matrix, 256 x 256 to 512 x 512). At 24 to 36 hours after the administration of USPIO, the T2- and T2*-weighted sequences used for the interpretation of lymph node status were repeated. Imaging was performed on a Signa 1.5T (GE Medical Systems, Milwaukee, WI) in 31 patients, Gyroscan Intera 1.5T system (Philips, Eindhoven, the Netherlands) in 11 patients, and a Symphony 1.5T (Siemens, Erlangen, Germany) in two patients.
Administration of Contrast
After the preliminary MRI scan, USPIO contrast medium (Sinerem) was administered intravenously. Each vial of contrast, containing 210 mg of USPIO, was reconstituted with 10 mL of normal saline. The dose of 2.6 mg iron/kg (0.13 mL/kg) was then diluted in 100 mL of normal saline and infused over 30 minutes under supervision. Three patients did not complete the USPIO infusion (one patient had a vasovagal reaction with fainting, one patient had suspected contrast reaction as evidenced by the onset of a tight cough at the beginning of the infusion, and one patient developed a marked rash and flushing). These patients had no significant adverse clinical events and recovered fully after a short period of observation. Six other patients (who completed the protocol) had minor adverse effects including erythematous rash (n = 5), muscle cramps (n = 1), diarrhea (n = 1), and indigestion (n = 1). All symptoms resolved spontaneously 1 to 4 hours after the infusion.
Image Interpretation
Two observers (readers 1 and 2), with no previous experience reading USPIO-MRI, interpreted the images prospectively. Learning curve patients (reader 1, three patients; reader 2, two patients) were not blinded and were excluded from the readers' results. A third reader, experienced in USPIO-MRI, was asked to review a random number of scans (29 of 44) independently to assess interobserver variability.
Size Criteria
The short-axis and long-axis diameters of each identifiable node were measured using electronic calipers on the scanner console. Node diagnosis was made using two standard size criteria: for short-axis criteria, short-axis diameter more than 10 mm indicated metastasis13; for size ratio criteria, nodes less than 8 mm short axis were considered benign, nodes more than 10 mm short axis were considered metastatic, and for nodes with a short axis between 8 and 10 mm, if the ratio of the short to long axis was more than 0.8 (ie, a round node), then the node was considered positive.14,15
Several other threshold values for short-axis diameter were also applied above which the node was considered positive: more than 5, 8, and 9 mm, respectively.16
USPIO Criteria
Nodes were evaluated on the pre- and post-USPIO images. USPIO criteria for benignity were homogeneous or slightly heterogeneous complete darkening (Fig 1); central darkening with a smooth rim of nondarkening (Fig 2); and darkening of nodal tissue with central area of nondarkening, which corresponded to fatty hilum on T1 images. USPIO criteria for malignant infiltration were only faint darkening, but with a decrease in signal intensity of less than 50% (Fig 3); focal area of high signal that did not correspond to fatty hilum on T1 (Fig 4); and complete uniform retention of high signal (Fig 5). For receiver operating characteristic (ROC) curve analysis, a five-point scale was used to indicate the level of confidence in the diagnosis: 1 indicated definitely benign, 2 indicated probably benign, 3 indicated indeterminate, 4 indicated probably malignant, and 5 indicated definitely malignant. Nodes in groups 1 and 2 were classified as benign and masses in groups 3 to 5 were classified as malignant for sensitivity and specificity calculations.
Surgery
Nodal status was discussed with the surgeon before lymphadenectomy. Careful identification of the nodes was made at the time of surgery, using a combination of the MR images and diagrams or with the attendance of the radiologist in the operating room. Resected nodes were anatomically labeled. In three patients, image-guided percutaneous biopsy or fine needle aspirate of a lymph node was performed, in accordance with clinical preference (Table 1).
Histopathology
Lymph nodes were cut into parallel slices of 2 to 3 mm thickness and all nodal tissue was routinely processed and embedded in paraffin. Sections were stained with hematoxylin and eosin and reported by a histopathologist with a specialist interest in gynecologic malignancy. One hundred twenty-two nodes (16%) were reviewed using serial sectioning with examination of at least five additional levels, one of which was stained immunohistochemically for cytokeratin, using the broad-spectrum anticytokeratin antibody MNF-116 (Dako, Cambridge, UK) in a 1:50 dilution using a standardized avidin-biotin complex system.
Statistics
Data were entered on an Statistical Package for the Social Sciences version 11 software package (SPSS Inc, Chicago, IL). The McNemar test was used to compare the sensitivity (in node-positive patients) and specificity (in node-negative patients) obtained with and without USPIO. Intrapatient correlation calculations and variance correction for clustered binary outcome variables were undertaken according to the methods proposed by Fleiss17 and Donner and Klar.18 Given that these methods become inaccurate when sample sizes are small, exact binomial confidence limits were therefore calculated for sensitivity and the positive predictive value. Exact binomial confidence limits were calculated for the patient by patient analysis (Table 2).
In differentiating between benign and malignant nodes, ROC curve analysis was also performed. The confidence scores of the readers 1 and 2 in identifying benign and malignant nodes were compared with the final histologic diagnosis using ROC analysis. ROC curves were also calculated for short-axis diameter in millimeters and the size ratio criteria. The mean areas under the ROC curves were determined. The ROC curve analysis was performed on nodes with complete data sets, thereby excluding the five learning curve cases in which confidence scores were only available from one reader in each sample. Thus 279 nodes were suitable for inclusion in the ROC curve analysis. Interobserver variability for USPIO-MRI was calculated using the statistic. Interobserver variability for nodal size was measured using the mean coefficient of variability.
RESULTS
A total of 768 lymph nodes sampled from the pelvic side wall and para-aortic regions were available for pathology, 17 of which were malignant (2.2%). A mean of 17.5 nodes were harvested per patient (range, one to 41); 11 of 44 patients (25%) had one or more positive nodes and 335 nodes could be anatomically correlated with the magnetic resonance images. The diagnostic performance of size criteria and USPIO criteria were assessed on a node-by-node basis (Table 3) and on a patient-by-patient basis (where a patient is considered to have nodal involvement if one or more nodes are diagnosed as malignant; Table 2). On a node-by-node basis, the sensitivity for the detection of malignant nodes using the size ratio was 0.29 (95% CI, 0.10 to 0.56); the sensitivity using USPIO criteria for reader 1 was 0.93 (95% CI, 0.66 to 1.00; P = .008), and the sensitivity for reader 2 was 0.82 (95% CI, 0.57 to 0.96; P = .004). On a patient-by-patient basis, the sensitivity using size ratio was 0.27 (95% CI, 0.06 to 0.61); the sensitivity using USPIO criteria for reader 1 was 1.00 (95% CI, 0.66 to 1.00; P = .03), and the sensitivity for reader 2 was 0.91 (95% CI, 0.59 to 1.00; P = .06).
ROC curve analysis (Fig 6) shows the area under the curve for short-axis diameter was 0.832 (95% CI, 0.686 to 0.978); for size ratio, the area under the curve was 0.621 (95% CI, 0.391 to 0.851). For USPIO-MRI reader 1, the area under the curve was 0.981 (95% CI, 0.966 to 0.997) and for reader 2, the area under the curve was 0.991 (95% CI, 0.981 to 1.001). Therefore, overall diagnostic performance of readers 1 and 2 using USPIO-MRI is better than that obtained using size ratio.
In four of the 44 patients, several small parametrial nodes were identified within the tissue block of the resected uterus, separate from lymph node dissection. These four patients had a primary diagnosis of cervical carcinoma. Seven of these nodes contained metastatic deposits. These nodes could not be identified in retrospect on the MR images either before or after USPIO administration. If these nodes are included in the node-by-node analysis, then the sensitivity for size criteria using the size ratio was 0.21 (95% CI, 0.07 to 0.42); sensitivity for reader 1 using USPIO criteria was 0.72 (95% CI, 0.46 to 0.90; P = .008), and sensitivity for reader 2 was 0.58 (95% CI, 0.37 to 0.78; P = .004), with a total of 10 false-negative nodal diagnoses (seven parametrial nodes and three pelvic side-wall nodes containing micrometastases, confirmed on immunostains). On a patient-by-patient analysis, when including the parametrial nodes, the sensitivity using the size ratio was 0.25 (95% CI, 0.05 to 0.57); using USPIO criteria, sensitivity for reader 1 was 0.90 (95% CI, 0.55 to 1.00; P = .03), and sensitivity for reader 2 was 0.83 (95% CI, 0.52 to 0.98; P = .06). Altogether, there were two patients with false-negative results: one patient with a false-negative result had two side-wall nodes containing micrometastases as well as three positive parametrial nodes (this was a learning curve patient for reader 1); the second patient with a false-negative result had a single, 4-mm paracervical/parametrial node that could not be seen in retrospect. The remainder of the 25 nodes dissected in this patient were correctly predicted to be benign on USPIO-MRI.
False-positive diagnoses on USPIO-MRI (three patients, reader 1; five patients, reader 2) occurred in the following situations. A false-positive result was obtained for a patient in whom extensive local and distant recurrent disease was diagnosed on follow-up CT scan within 3 months of her surgery. This suggests that the positive diagnosis on USPIO-MRI may have been correct but the node suggestive of metastasis was not surgically resected, even though 41 nodes were collected. A false-positive result was obtained in two patients in whom at least one node was histologically positive but in whom USPIO-MRI diagnosed additional nodes. A false-positive result was obtained in four patients in whom no nodal metastasis was present histologically and in whom there was no recurrent nodal disease within 6 months of follow-up. In these latter four patients, misinterpretation was due to a fatty hilum (one patient) heterogeneous signal within a node (two patients); in the fourth patient, all three readers diagnosed metastatic para-aortic nodes with a confidence score of 5, but the patient had concomitant infected pelvic lymphoceles. It is possible that the presence of lymphoceles, secondary to an extensive lymph node dissection in the previous 6 months, may have affected contrast uptake because of disruption of the normal lymph channels. Currently, we do not know what effect previous lymph node dissection of lymphocele has on USPIO uptake.
Interobserver variability (Table 4) was good to very good. The mean coefficient of variability for lymph node size measurements between reader 1 and reader 2 was 0.06 (standard deviation, 0.05).
DISCUSSION
Our results demonstrate a significant increase in the sensitivity, with no loss of specificity, for the noninvasive detection of malignant lymph nodes in patients with cervical and endometrial cancer. This was seen on a node-by-node basis, for which the sensitivity increased from 29% using standard size criteria to 93% using USPIO criteria. This increase in sensitivity was seen to a greater extent on a patient-by-patient basis, for which the sensitivity increased from 27% using standard size criteria to 100% using USPIO criteria. Only one patient with cervical cancer metastatic to lymph nodes along the pelvic side wall had a false-negative result using USPIO criteria by one reader. One patient with a single 4-mm metastatic node within the parametrial tissues also had a false-negative result. All patients with endometrial cancer metastatic to lymph nodes were correctly identified.
Nodal size is the current standard criterion for the prediction of metastases to lymph nodes on cross-sectional imaging. There is some variability in the size threshold used by different groups, but the most widely used threshold value for identifying malignant lymph nodes in the literature is 10 mm short-axis diameter.8,19,20 Our results demonstrate a clear improvement when compared with this threshold level. More recently, a size ratio has been used whereby indeterminate-sized nodes (8 to 10 mm) that are round are considered malignant.14 Our results show a clear improvement when compared with this threshold. Setting a lower short-axis diameter threshold value for diagnosis of malignant nodes improves sensitivity but with a corresponding decrease in the specificity. This is reflected in the ROC curve analysis in which the area under the curve for using short-axis diameter as the diagnostic tool was 0.834 compared with 0.991 using USPIO criteria (although close confidence intervals were observed), reflecting the small number of positive nodes in our group of patients.
Contrast administration was generally well tolerated; flushing or a rash was the most frequent symptom. Contrast administration was stopped in 5% of patients, as a precaution, but no patient had any significant sequelae as a result of the infusion. The MRI sequences used in our study are widely available on most MR systems independent of vendor platforms, and image analysis does not require any additional software.
False-negative diagnoses were predominantly due to our failure to identify nodes in the parametrial tissues, both before and after USPIO administration, as well as in retrospect. This may be due to limitations of our MRI technique, in which we set out specifically to diagnose the pelvic side-wall nodes, which are well seen on MRI and are the traditional focus for radiologists attempting preoperative lymph node diagnosis. Future improvements in technique may help prevent this problem. However, the parametrial tissues frequently contain an extensive plexus of serpiginous veins, and the identification of parametrial nodes is likely to remain challenging. Currently, the parametrial tissues are removed at primary surgery in most patients with cervical carcinoma; thus, preoperative diagnosis at this site may not have an impact on decision making. No positive parametrial nodes were diagnosed at histology in patients with endometrial carcinoma. This may be due to a difference in surgical technique: endometrial carcinoma is not commonly treated by radical hysterectomy; it is usually treated by simple hysterectomy, and thus the parametrial tissues are not available for histologic assessment. Micrometastases are also likely to remain a challenge because of limitations in the spatial resolution of MRI systems.
False-positive diagnoses may reflect incomplete nodal dissection, in which a node may have been inaccessible to or not seen by the surgeon. In these patients, it may be postulated that USPIO-MRI provides a more complete anatomic map of the lymph node involvement. However, false-positive interpretation also occurred where a node seemed heterogeneous or where a fatty hilum was misinterpreted as a deposit. In addition, the presence of bilateral infected pelvic lymphoceles in one patient may have altered the contrast uptake into the nodes. These misinterpretations may be related to learning curves or possibly due to technical factors.
Our findings support earlier published results of USPIO-MRI in other tumors, either in head and neck or breast, or in patients with a tumor in the abdomen or pelvis.21-24 A recently published study restricted to patients with prostate cancer showed the sensitivity for nodal staging using USPIO criteria was 90.5%—significantly greater than that achieved using size criteria, which had a sensitivity of 35.4%.15 On a patient-by-patient basis, the same study had a sensitivity of 100% for detecting patients with one or more positive nodes.
USPIO-MRI has the potential to improve the surgical and nonsurgical management of patients with cervical and endometrial carcinoma in several ways. First, the preoperative localization of malignant nodes may allow tailoring of the surgical lymphadenectomy, allowing rapid collection of any node suggestive of metastasis for frozen section, and by directing the surgeon to a node suggestive of metastasis at a site not usually accessed. This tailored approach may reduce both surgical time and morbidity without a reduction in diagnostic yield. Second, when USPIO-MRI is negative, surgical lymph node sampling could be avoided altogether, with a high degree of confidence. This could be of particular value when surgical risks such as obesity and diabetes are present. Third, USPIO-MRI could accurately map the extent of lymph node involvement in both the pelvis and para-aortic regions to define the radiotherapy fields. This is particularly relevant with the development of intensity-modulated radiotherapy, in which the shape of the field may be closely tailored to the individual patient.
In conclusion, USPIO-MRI improves the preoperative characterization of lymph node metastases compared with standard MRI. This information may provide the clinician with important information in planning the optimal surgical or radiotherapy treatment.
Appendix
Andrea Rockall obtained ethics approval, Medical Control Agency approval, and BUPA Foundation Grant (London, UK); recruited patients; interpreted images, analyzed data, and was the primary author of the manuscript. S. Aslam Sohaib recruited patients at the Royal Marsden Hospital, interpreted images, prepared the manuscript, and analyzed data. M. Harisinghani provided advice on MRI sequences, interpreted images, and provided advice on the manuscript. S. Babar recruited patients and assisted with data collection. N. Singh performed node histopathology. I. Jacobs, D. Oram, A. Jeyarajah, and J. Shepherd sampled surgical lymph nodes and provided advice on the manuscript. R. Reznek acted as the senior supervisor and performed data interpretation.
Authors' Disclosures of Potential Conflicts of Interest
The authors indicated no potential conflicts of interest.
Acknowledgment
We thank Peter A. Armstrong for reading and commenting on the manuscript and Roger A'Hern for statistical advice.
NOTES
Supported by the BUPA Foundation, London, United Kingdom.
Presented in abstract form at the 2nd Annual Meeting of the International Cancer Imaging Society, Paris, France, October 2002, 15th Annual European Congress of Radiology, Vienna, Austria, March 7-11, 2003, and the 89th Scientific Assembly and Annual Meeting of the Radiological Society of North America, Chicago, IL, November 30-December 5, 2003.
Authors' disclosures of potential conflicts of interest are found at the end of this article.
REFERENCES
Stehman FB, Bundy BN, DiSaia PJ, et al: Carcinoma of the cervix treated with radiation therapy: I. A multi-variate analysis of prognostic variables in the Gynecologic Oncology Group. Cancer 67:2776-2785, 1991
Manetta A, Delgado G, Petrilli E, et al: The significance of paraaortic node status in carcinoma of the cervix and endometrium. Gynecol Oncol 23:284-290, 1986
Cohn DE, Peters WA III, Muntz HG, et al: Adenocarcinoma of the uterine cervix metastatic to lymph nodes. Am J Obstet Gynecol 178:1131-1137, 1998
Shepherd JH: Revised FIGO staging for gynaecological cancer. Br J Obstet Gynaecol 96:889-892, 1989
Hertel H, Kohler C, Elhawary T, et al: Laparoscopic staging compared with imaging techniques in the staging of advanced cervical cancer. Gynecol Oncol 87:46-51, 2002
Schlaerth JB, Spirtos NM, Carson LF, et al: Laparoscopic retroperitoneal lymphadenectomy followed by immediate laparotomy in women with cervical cancer: A gynecologic oncology group study. Gynecol Oncol 85:81-88, 2002
Agostini A, Carcopino X, Franchi F, et al: Port site metastasis after laparoscopy for uterine cervical carcinoma. Surg Endosc 17:1663-1665, 2003
Reinhardt MJ, Ehritt-Braun C, Vogelgesang D, et al: Metastatic lymph nodes in patients with cervical cancer: Detection with MR imaging and FDG PET. Radiology 218:776-782, 2001
Yu KK, Hricak H, Subak LL, et al: Preoperative staging of cervical carcinoma: Phased array coil fast spin-echo versus body coil spin-echo T2-weighted MR imaging. AJR Am J Roentgenol 171:707-711, 1998
Bipat S, Glas AS, van der Velden J, et al: Computed tomography and magnetic resonance imaging in staging of uterine cervical carcinoma: A systematic review. Gynecol Oncol 91:59-66, 2003
Heller PB, Maletano JH, Bundy BN, et al: Clinical-pathologic study of stage IIB, III, and IVA carcinoma of the cervix: Extended diagnostic evaluation for paraaortic node metastasis—A Gynecologic Oncology Group study. Gynecol Oncol 38:425-430, 1990
Vassallo P, Matei C, Heston WD, et al: AMI-227-enhanced MR lymphography: Usefulness for differentiating reactive from tumor-bearing lymph nodes. Radiology 193:501-506, 1994
Reinhardt MJ, Technau-Ihling K, Altehoefer C, et al: Lymphangiography causes false-positive findings on 18F-FDG PET imaging. Anticancer Res 23:2941-2944, 2003
Jager GJ, Barentsz JO, Oosterhof GO, et al: Pelvic adenopathy in prostatic and urinary bladder carcinoma: MR imaging with a three-dimensional TI-weighted magnetization-prepared-rapid gradient-echo sequence. AJR Am J Roentgenol 167:1503-1507, 1996
Harisinghani MG, Barentsz J, Hahn PF, et al: Noninvasive detection of clinically occult lymph-node metastases in prostate cancer. N Engl J Med 348:2491-2499, 2003
Vinnicombe SJ, Norman AR, Nicolson V, et al: Normal pelvic lymph nodes: Evaluation with CT after bipedal lymphangiography. Radiology 194:349-355, 1995
Fleiss JL: Statistical methods for rates and proportions, 2nd edition. New York, NY, John Wiley and Sons, 1981
Donner A, Klar N: Methods for comparing event rates in intervention studies when the unit of allocation is a cluster. Am J Epidemiol 140:279-289, 1994
Lin WC, Hung YC, Yeh LS, et al: Usefulness of (18)F-fluorodeoxyglucose positron emission tomography to detect para-aortic lymph nodal metastasis in advanced cervical cancer with negative computed tomography findings. Gynecol Oncol 89:73-76, 2003
Grigsby PW, Siegel BA, Dehdashti F: Lymph node staging by positron emission tomography in patients with carcinoma of the cervix. J Clin Oncol 19:3745-3749, 2001
Anzai Y, Piccoli CW, Outwater EK, et al: Evaluation of neck and body metastases to nodes with ferumoxtran 10-enhanced MR imaging: Phase III safety and efficacy study. Radiology 228:777-788, 2003
Stets C, Brandt S, Wallis F, et al: Axillary lymph node metastases: A statistical analysis of various parameters in MRI with USPIO. J Magn Reson Imaging 16:60-68, 2002
Sigal R, Vogl T, Casselman J, et al: Lymph node metastases from head and neck squamous cell carcinoma: MR imaging with ultrasmall superparamagnetic iron oxide particles (Sinerem MR): Results of a phase-III multicenter clinical trial. Eur Radiol 12:1104-1113, 2002
Pannu HK, Wang KP, Borman TL, et al: MR imaging of mediastinal lymph nodes: Evaluation using a superparamagnetic contrast agent. J Magn Reson Imaging 12:899-904, 2000(Andrea G. Rockall, Syed A)