Fax: (011) 8231920-1171
Comparison of the accuracy of magnetic resonance imaging and positron emission tomography/computed tomography in the presurgical detection of lymph node metastases in patients with uterine cervical carcinoma
A prospective study
Version of Record online: 12 JAN 2006
Copyright © 2006 American Cancer Society
Volume 106, Issue 4, pages 914–922, 15 February 2006
How to Cite
Choi, H. J., Roh, J. W., Seo, S.-S., Lee, S., Kim, J.-Y., Kim, S.-K., Kang, K. W., Lee, J. S., Jeong, J. Y. and Park, S.-Y. (2006), Comparison of the accuracy of magnetic resonance imaging and positron emission tomography/computed tomography in the presurgical detection of lymph node metastases in patients with uterine cervical carcinoma. Cancer, 106: 914–922. doi: 10.1002/cncr.21641
- Issue online: 3 FEB 2006
- Version of Record online: 12 JAN 2006
- Manuscript Accepted: 19 AUG 2005
- Manuscript Revised: 16 AUG 2005
- Manuscript Received: 28 JUN 2005
- National Cancer Center. Grant Number: 0410200
- cervical carcinoma;
- lymph node metastases;
The objective of the current study was to determine the accuracy of magnetic resonance imaging (MRI) and positron emission tomography/computed tomography (PET/CT) for detecting lymph node metastases in patients with uterine cervical carcinoma compared with thin-section histopathologic results from systemic lymphadenectomy.
Twenty-two patients with International Federation of Obstetrics and Gynecology (FIGO) Stage IB–IVA cervical carcinoma who underwent both MRI and PET/CT before lymphadenectomy were included in this study. Lymphadenectomy involved removing all visible lymph nodes in the surgical fields. To enable region-specific comparisons, paraaortic and pelvic lymph nodes were divided into seven regions: the paraaortic area, both common iliac areas, both external iliac areas, and both internal iliac/obturator areas. Histopathologic evaluation of lymph nodes was the diagnostic standard. Chi-square analysis was used to compare the accuracy of MRI and PET/CT for the detection of metastatic lymph nodes. A P value ≤ 0.05 was considered statistically significant.
With MRI, the sensitivity, specificity, and accuracy rates for detecting metastatic lymph nodes in each lymph node group were 30.3% (10 of 33 lymph node groups), 92.6% (112 of 121 lymph node groups), and 72.7% (122 of 154 lymph node groups), respectively; with PET/CT, those rates were 57.6% (19 of 33 lymph node groups), 92.6% (112 of 121 lymph node groups), and 85.1% (131 of 154 lymph node groups), respectively. Statistical analysis showed that PET/CT was more sensitive than MRI (P = 0.026) but that there were no statistical differences noted with regard to specificity (P = 1.000) or accuracy (P = 0.180). Power analysis demonstrated that a sample size of 685 lymph node groups (98 patients) would be necessary to demonstrate that PET/CT was more accurate than MRI (α = 0.05; β = 0.80).
PET/CT was more sensitive than MRI for detecting lymph node metastases in patients with uterine cervical carcinoma. Cancer 2006. © 2006 American Cancer Society.
Carcinoma of the uterine cervix is the second most frequently diagnosed malignancy in women worldwide, and it is the only major gynecologic malignancy that is staged clinically according to International Federation of Obstetrics and Gynecology (FIGO) recommendations.1 Subsequent surgical staging shows that clinical staging of cervical carcinomas is accurate in only approximately 60% of patients, and undiagnosed lymph node metastases are a major problem.2–4 Lymph node metastases in gynecologic malignancies have an adverse impact on survival, especially in patients with cervical carcinoma.5 Although lymph node resection before radiotherapy results in improved survival in patients with macroscopically enlarged pelvic and paraaortic lymph nodes, routine pretreatment surgical staging is not recommended. For this reason, inaccurate assessment of lymph node involvement can lead to suboptimal treatment.6–11
Computed tomography (CT) and magnetic resonance imaging (MRI) have been used to assess paraaortic and pelvic lymph nodes in patients with cervical carcinoma. A meta-analysis of such studies concluded that these methods have only moderate sensitivity and specificity for detecting metastases.12 A major problem with these modalities is that they fail to identify small lymph node metastases. More recently, it has been shown that positron emission tomography (PET) employing the glucose analog [18F]-flouro-2-deoxy-D-glucose (FDG) is more sensitive than CT or MRI for detecting lymph node metastases in patients with cervical carcinoma.13–18 However, PET has lower spatial resolution than CT or MRI. Fused PET/CT, as described by Beyer et al., combines the anatomic detail provided by CT with PET metabolic information.19 Initial studies have shown that this technique improved the anatomic localization of PET abnormalities and reduced the number of equivocal PET interpretations.20–22 However, to our knowledge, no studies to date have compared PET/CT with MRI in terms of detecting lymph node metastases from cervical carcinoma. The objective of the current study was to determine the sensitivity, specificity, and accuracy of MRI and PET/CT in the detection of metastatic lymph nodes in patients with cervical carcinoma compared with thin-section histopathologic results from systemic lymphadenectomy.
MATERIALS AND METHODS
This was prospective study that involved 22 untreated patients with histopathologically confirmed, FIGO Stage IB–IVA invasive cervical carcinoma, as determined by a conventional workup that included MRI and PET/CT scans. These patients were recruited between October 2003 and January 2005, and they ranged in age from 25 years to 65 years (mean age, 50 yrs). The patients had no contraindications to the surgical procedure, had no evidence of distant metastases, and had an Eastern Cooperative Oncology Group (ECOG) performance status of 0–1. Those who did not want to undergo PET/CT or laparoscopic lymphadenectomy (n = 63 patients) and those who had tumors other than squamous cell carcinoma (n = 10 patients) were excluded.
All patients were classified into two groups according to FIGO staging. The first group was comprised of patients who had Stage IB1 disease (tumor size ≤ 4 cm) and Stage IIA disease, and they underwent conventional lymphadenectomy combined with a laparoscopic radical hysterectomy (n = 7 patients). The second group was comprised of patients with Stage IB2 or ≥ Stage IIB disease, and they received chemoradiation preceded by laparoscopic paraaortic and pelvic lymphadenectomy (n = 15 patients). Informed consent was obtained from all patients, and the trial was approved by the Institutional Review Board of the National Cancer Center.
Conventional Staging Workup
After histologic confirmation of invasive cervical carcinoma, FIGO stage was determined by bimanual pelvic examination (S.-Y.P. and J.W.R.), cystoscopy, and sigmoidoscopy.
MRI was performed using a Signa 1.5-T system (General Electric Medical Systems, Milwaukee, WI) with a pelvic array coil for pelvic scans and a Torso phase-array coil for paraaortic scans. MR scans were obtained between 3 and 20 days (mean ± standard deviation, 7 days ± 5 days) before surgery. Scans were obtained by using the following parameters: for the pelvic region, an axial T1-weighted, fast spin-echo sequence (TR/TE, 600/10 msec; slice thickness, 5 mm; interslice gap, 2 mm; field of view, 24 cm × 24 cm; matrix, 256 × 192; echo train length, 4; 3 signals acquired; no fat saturation; bandwidth, 31.25 kilohertz [kHz]); an axial T2-weighted, fast spin-echo sequence (TR/TE, 5000/68 msec; slice thickness, 3 mm; interslice gap, 1 mm; field of view, 24 cm × 24 cm, matrix, 256 × 192; echo train length, 21; 4 signals acquired; no fat saturation; bandwidth, 31.25 kHz); and a sagittal T2-weighted, fast spin-echo sequence (TR/TE, 5000/68 msec; slice thickness, 3 mm; interslice gap, 3 mm; field of view, 24 cm × 24 cm; matrix, 256 × 192; echo train length, 26; 4 signals acquired; no fat saturation; bandwidth, 31.25 kHz); and, for the paraaortic region, an axial T2-weighted, fast spin-echo sequence with 16 seconds of breath holding (TR/TE, 2000/68 msec; slice thickness, 8 mm; interslice gap 2 mm; field of view, 32 cm × 24 cm; matrix, 256 × 160; echo train length, 20; 1 signals acquired; no fat saturation; bandwidth, 31 kHz).
PET/CT Imaging Protocol
Whole-body PET/CT images were obtained by using either one of two combined PET/CT scanners; a Biograph LSO (Siemens Medical Solutions, Hoffman Estates, IL) or a Discovery LS (General Electric Medical Systems). The choice of scanner was arbitrary: The Biograph LSO was used on 12 patients, and the Discovery LS scanner was used on 10 patients. PET/CT scans were obtained between 2 days and 18 days (mean ± standard deviation, 6 days ± 5days) before surgery.
After an 8-hour fast, patients received an intravenous injection of 444–740 megabecquerels (12–20 millicuries [mCi]) FDG and were encouraged to rest. PET/CT scanning was performed from the middle of the skull to the upper thigh 60 minutes after the injection, and additional regional PET/CT scanning was performed from the upper margin of the left kidney to the pelvic floor 120 minutes after the injection. On each PET/CT scan, a spiral CT scan was performed using the following parameters. For the Biograph LSO scanner, a scout view at 30 milliampere (mA) and 130 kilovolt peak (kVp) was followed by a spiral CT scan with a 0.8-second rotation time, 50 mA, 130 kVp, 5-mm section width, 4-mm collimation, and 12-mm table feed per rotation with arms raised. For the Discovery LS scanner, a scout view at 30 mA and 120 kVp was followed by a spiral CT scan with a 0.8-second rotation time, 80 mA, 140 kVp, 5-mm section thickness, and 4.25-mm interval in high-speed mode with arms at the sides of the torso. PET image acquisition followed CT scanning (3 min per bed position of 11.2 cm in 3-dimensional acquisition mode [Biograph LSO] or 4 min per bed position of 14.2 cm in 2-dimensional acquisition mode [Discovery LS]). CT images were reconstructed onto a 512 × 512 matrix and were converted into 511 kiloelectron volt (keV)-equivalent attenuation factors for attenuation correction. PET images were reconstructed onto a 128 × 128 matrix using ordered-subsets expectation maximization and attenuation correction. The standardized uptake value (SUV) was calculated as follows: SUV = (decay-corrected activity [kilobecquerels] per mL of tissue volume)/(injected 18F-FDG activity [kilobecquerels]/body mass [g]). The SUV of the lesion was obtained by placing regions of interest manually around the lesion, and the maximum SUV within a region of interest was used to minimize partial-volume effects. PET, PET/CT, and CT images were analyzed using a dedicated workstation and eNtegra (General Electric Medical Systems) and e.soft (Siemens Medical Solutions) software, which allowed three-dimensional image display (transaxial, coronal, and sagittal) and maximum-intensity projection of PET data.
Classification of Lymph Node Regions
Paraaortic and pelvic lymph nodes were classified into seven groups according to the nearest, largest artery: the paraaortic area (including right and left paraaortic lymph nodes), both common iliac areas, both external iliac areas, and both internal iliac/obturator areas (Fig. 1).
All patients received preoperative bowel preparation for 2 days, received prophylactic antibiotics, and underwent surgery with the same instruments and techniques. Two gynecologic surgeons (S.-Y.P. and J.W.R.) performed all operations. The surgeons were aware of MRI and PET/CT lymph node findings prior to the lymph node dissections. The first group (patients with Stage IB1 and IIA disease) underwent conventional lymphadenectomy combined with radical hysterectomy by laparotomy (n = 7 patients) and the second group (patients with Stage IB2 or ≥ Stage IIB disease) underwent laparoscopic paraaortic and pelvic lymphadenectomy (n = 15 patients).
For both groups, lymphadenectomy was performed systematically with sampling of all lymph node groups. All harvested lymph nodes were grouped according to the name of the adjacent vessel (abdominal aorta, both common iliac arteries, both external iliac arteries, and both internal iliac and obturator arteries). For discrimination of the external iliac lymph node group from the internal/obturator lymph node group, we classified the lymph nodes that were located parallel or anterior to the external iliac artery as the external iliac lymph node group, and the lymph nodes that were located posterior to the external iliac artery and near the internal iliac or obturator artery were classified as the internal iliac/obturator lymph node group. The upper limit of paraaortic lymph node dissection was the duodenum. The margin of lymph node dissection was demarcated by using endoclips, and this margin was confirmed by abdominal X-ray after surgery.
Histopathologic evaluation of lymph nodes was the diagnostic standard. Thin sections were stained with hematoxylin and eosin and were examined microscopically by a pathologist (S.L.). Each lymph node was sliced at 2-mm intervals perpendicular to the greatest dimension to maximize the likelihood of detecting micrometastases.
Analysis and Statistics
MR images were interpreted by a consensus of 2 radiologists (H.J.C. and J.Y.J.) who had 3 years and 4 years of experience in gynecologic imaging, respectively, including MRI. PET/CT images were interpreted by a consensus of 2 physicians (S-.K.K. and K.W.K.) who were trained in nuclear medicine with 4 years and 5 years of experience in PET imaging, respectively. These clinicians were unaware of any results from other imaging studies. The criterion for malignancy on MR images was a pelvic or paraaortic lymph node with a short-axis dimension ≥ 1 cm. MR images were examined from the picture archiving and communication system.
For PET/CT images, interpretation was made visually by using a 5-point grading system as follows: 0, no visible FDG accumulation; 1, less than liver accumulation; 2, similar to liver accumulation; 3, greater than liver accumulation and less than brain cortex accumulation; and 4, similar to brain cortex accumulation.23, 24 A malignant lymphadenopathy was defined as follows: 1, FDG accumulation ≥ Grade 3 and SUV > 2.5 g/mL, or 2) a lesion SUV that did not decrease on the delayed PET image compared with the initial PET image and a corresponding lesion on CT despite FDG accumulation < Grade 3 or SUV < 2.5 g/mL.25 PET/CT images were examined directly from the screen of the computer workstation. Information from MRI and PET/CT scanning was entered into a prospective data base, and histopathology staging information was stored in a separate file.
Independent Student t tests were used to compare the size of metastatic and non-metastatic lymph nodes. The sensitivity, specificity, and accuracy of MRI and PET/CT scanning in detecting lymph node metastases from cervical carcinoma were compared by using chi-square analysis. QuickCalcs software (GraphPad Software Inc., San Diego, CA) was used for analysis. A P value ≤ 0.05 was considered significant.
Metastatic Lymph Node Detection: Region-Specific Analysis
Metastatic lymph nodes were identified in 13 of 22 patients (59.1%) from 33 of the 154 lymph node regions (21.4%) that were examined histopathologically. MRI identified 86 lymph nodes in total, with a mean (± standard deviation) short-axis dimension of metastatic lymph nodes (11.8 ± 3.5 mm; n = 27 lymph nodes) larger compared with that of nonmetastatic lymph nodes (6.5 ± 4.6 mm; n = 59 lymph nodes; P = 0.043).
MRI detected 10 of 33 lymph node groups (30.3%) that were identified as metastatic according to histopathologic examination. These comprised 1 of 4 paraaortic lymph node groups, 0 of 3 right common iliac lymph node groups, 1 of 5 left common iliac lymph node groups, 0 of 0 right external iliac lymph node groups, 1 of 3 left external iliac lymph node groups, 4 of 10 right internal iliac/obturator lymph node groups, and 3 of 8 left internal iliac/obturator lymph node groups (Fig. 2). In comparison, PET/CT images detected 19 of 33 lymph node groups (57.6%) that were identified as metastatic according to pathologic examination, and these comprised 2 of 4 paraaortic lymph node groups, 0 of 3 right common iliac lymph node groups, 4 of 4 left common iliac lymph node groups, 0 of 0 right external iliac lymph node groups, 2 of 3 left external iliac lymph node groups, 6 of 10 right internal iliac/obturator lymph node groups, and 5 of 8 left internal iliac/obturator lymph node groups (P = 0.026) (Table 1; Figs. 2,3).
|Modality||Lymph node groupa|
|No. of positive lymph node groups||4||3||5||0||3||10||8||33|
|No. of lymph node groups evaluated||22||22||22||22||22||22||22||154|
MRI produced 9 false-positive interpretations (1 paraaortic lymph node group, 1 left common iliac lymph node group, 3 right internal iliac/obturator lymph node groups, and 4left internal iliac/obturator lymph node groups), resulting in 92.6% specificity (112 of 121 lymph node groups) (Fig. 4). Similarly, PET/CT produced 9 false-positive interpretations (2 paraaortic lymph node groups, 1 left common iliac lymph node group, 1 right external iliac lymph node group, 2 right internal iliac/obturator lymph node groups, and 3 left internal iliac/obturator lymph node groups), resulting in 92.6% specificity (112 of 121 lymph node groups; P = 1.000) (Table 2) (Fig. 4). Accuracy for the detection of malignant lymph node groups was 72.7% with MRI (122 of 154 lymph node groups) and 85.1% with PET/CT (131 of 154 lymph node groups; P = 0.181; standard power = 0.26). Power analysis demonstrated that a sample size of 685 lymph node groups would be necessary to show that PET/CT was more accurate than MRI (α = 0.05; β = 0.80). Here, the power of the test means the probability of rejecting the null hypothesis given that the alternative hypothesis is true. The result is a decision regarding the sample size at a given α level (0.05) and statistical power (0.80).
|Modality||Lymph node groupa|
|No. of negative lymph node groups||18||19||17||22||19||12||14||121|
|No. of lymph node groups evaluated||22||22||22||22||22||22||22||154|
Metastatic Lymph Node Detection: Patient-by-Patient Comparison
For preoperative lymph node staging, sensitivity was 38.5% with MRI (4 of 13 patients) and 76.9% with PET/CT (10 of 13 patients; P = 0.047). Specificity for the detection of metastatic lymph nodes was 44.4% with MRI (4 of 9 patients) and 55.5% with PET/CT (5 of 9 patients; P = 0.637). Accuracy for the detection of metastatic lymph nodes was 40.9% with MRI (9 of 22 patients) and 68.2% with PET/CT (15 of 22 patients; P = 0.199; standard power = 0.75). Power analysis demonstrated that a sample size of 98 patients would be necessary to show that PET/CT was more accurate than MRI (α = 0.05; β = 0.80).
The presence of metastatic lymph nodes radically modifies the prognosis and treatment of patients with cervical carcinoma.8–11 In patients with surgically staged, locally advanced cervical carcinoma, the 5-year survival rate in those who have negative lymph node status is 57%, and this rate is reduced to 34% when pelvic lymph nodes are involved and to 12% when paraaortic lymph nodes are involved.26 However, the presence of lymph node metastases does not alter the clinical FIGO staging of cervical carcinoma.27
The accuracy and sensitivity rates with MRI for detecting metastases in lymph nodes from cervical carcinoma reportedly are between 76–100% and 36–71%, respectively.28–32 Although there have been several MRI studies in which different relaxation times or special contrast media enhancements were used to differentiate metastatic from hyperplastic lymph nodes, the only MRI criterion that is accepted generally in the evaluation of pelvic lymph node metastases is the size of the lymph node.33–37 Although size criteria of 1–2 cm have been reported in the literature, in the past decade, a greatest dimension of 1 cm has become the preferred criterion as either the maximum or the minimum transverse dimension.14, 29, 37, 38 The main limitation of MRI is that it is impossible to differentiate metastatic lymph nodes from nonmetastatic hyperplastic lymph nodes of similar size. In the current study, the criterion for metastasis was a lymph node short-axis dimension ≥ 1 cm.
The PET accuracy and sensitivity rates for detecting lymph node metastases from cervical carcinoma reportedly are between 67–100% and 92–99%, respectively.13–17 Studies comparing PET with CT and MRI for detecting lymph node metastases from cervical carcinoma have suggested a role for functional imaging.14, 15 In a study of 21 patients with Stage IB–IVA cervical carcinoma, Sugawara et al. reported a sensitivity with PET of 86% for pelvic and paraaortic lymph node metastases compared with a sensitivity with CT of 57%.15 In a study of 24 patients with Stage IB–IIB cervical carcinoma before they underwent surgical staging lymphadenectomy, Reinhardt et al. reported sensitivity, specificity, and accuracy rates of 67%, 97%, and 95%, respectively, with MRI and 81%, 99%, and 98% with PET, respectively, for detecting paraaortic and pelvic lymph node metastases.14
Although it has been demonstrated that PET is superior to CT and MRI for detecting pelvic and paraaortic lymph node metastases from cervical carcinoma, PET is limited in terms of anatomic and spatial resolution compared with either CT or MRI.14, 15 Some reports have stated that PET/CT improved the anatomic localization of PET abnormalities and reduced the number of equivocal PET interpretations.20–22 Cerfolio et al. compared the accuracy of PET/CT with PET alone for the staging of 129 patients with nonsmall cell lung carcinoma and found that PET/CT was more accurate for the detection of lymph node metastases compared with PET alone (96% vs. 93%; P = 0.01).22
To our knowledge, the current study is the first prospective, region-specific, surgical-histopathologic analysis involving PET/CT that covers all cervical carcinoma stages and all lymph node regions. We found relatively low sensitivity rates with both MRI and PET/CT for the detection of metastatic lymph nodes compared with the rates reported by others. One explanation may be that, for the current study, we conducted histopathologic examination of lymph nodes with transverse cuts made at 2-mm intervals, whereas lymph nodes conventionally are examined in only 1 or 2 longitudinal sections. Therefore, it is possible that, with the current histopathologic technique, we detected metastases that conventional techniques may have missed. Another possible explanation is that, in the current study, all visible lymph nodes in the appointed regions were harvested regardless of imaging findings or FIGO staging. For this reason, there may have been more regions without metastatic lymph nodes, possibly explaining why the sensitivity rates were lower than in previous reports.
One limitation of the current study was that the numbers of patients involved were relatively small. This prevented us from comparing the accuracy for each lymph node group, which would have been of particular interest for the paraaortic lymph node group, in which PET scans have shown lower sensitivity for the detection of metastasis.13, 14 We believe that greater patient numbers may have resulted in the observation of a statistically significant difference between PET/CT and MRI in terms of accuracy for detecting metastatic lymph nodes. A second limitation of the study was that surgeons were guided by preoperative MRI and PET/CT findings, which may have resulted in verification bias.
In conclusion, although MRI is useful for detecting enlarged pelvic and paraaortic metastatic lymph nodes, PET/CT is more sensitive than MRI for detecting metastatic lymph nodes from cervical carcinoma.
- 9Is the International Federation of Gynecology and Obstetrics staging system for cervical carcinoma able to predict survival in patients with cervical carcinoma? An assessment of clinimetric properties. Cancer. 2001; 92: 796–804., .
- 17Positron emission tomography (PET) for evaluating para-aortic and pelvic lymph node metastasis in cervical cancer before surgical staging: a surgicopathologic study [abstract]. Proc Am Soc Clin Oncol. 2003; 22: 456a., , , et al.
- 23Clinical molecular anatomical imaging. Philadelphia: J.B. Lippincott Company, 2003..