| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
1 Department of Radiology/MRI UH-B2B311, University of Michigan Health System,
1500 E Medical Center Dr., Ann Arbor, MI 48109-0003.
2 Present address: Department of Radiology, Abdominal Imaging Section, Hacettepe
University Faculty of Medicine, Ankara, Turkey.
3 Department of Biostatistics, School of Public Health, University of Michigan
Health System, Ann Arbor, MI 48109.
4 Department of Pathology, University of Michigan Health System, Ann Arbor, MI
48109.
Received January 8, 2003;
accepted after revision July 9, 2003.
Address correspondence to H. K. Hussain.
Abstract
 Top Abstract IntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
MATERIALS AND METHODS. We conducted a retrospective review of MRI performed in 40 consecutive patients with 42 renal cell carcinomas before radical (n = 35) or partial (n = 4) nephrectomy or exploratory laparotomy (n = 3). The interval between imaging and surgery ranged from 1 to 59 days (mean, 17.9 days). Imaging was performed with T1- and T2-weighted, dynamic gadolinium-enhanced, and time-of-flight sequences. MRI and surgical–pathologic staging was performed using the 1997 TNM staging system. MRI staging was compared with surgical–pathologic staging as the gold standard. Agreement between the two staging methods was assessed using the kappa statistic.
RESULTS. Agreement between MRI and surgical–pathologic staging
was good for T staging (
= 0.72 and 0.78 for reviewers 1 and 2
respectively), poor for N staging ( = 0.13, both reviewers), good for M
staging ( = 0.66, both reviewers), and excellent for the assessment of
venous involvement ( = 0.93, both reviewers). MRI overestimated the T
stage in five patients and the N stage in five and underestimated the T stage
in three, the N stage in four, the M stage in one, and the extent of venous
thrombosis in two patients.
CONCLUSION. MRI is a reliable method for preoperative staging of renal cell carcinoma using the 1997 TNM classification, in particular for assessing venous involvement.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
The commonly used staging classifications for renal cell carcinoma are the TNM [4] and Robson [5–10] classifications. Although the Robson classification is the most widely used in the imaging literature, TNM classification has more subgroups, which allows detailed categorization of the tumor, nodes, and venous involvement. The TNM staging classification underwent a significant revision in 1997 whereby the size limit for T1 tumor was changed from 2.5 to 7 cm [4]. Our aim was to assess the accuracy of MRI for preoperative staging of renal cell carcinoma using the 1997 TNM staging system, with surgical and pathologic staging as the gold standard.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
MRI
All 40 MRI studies were performed on 1.5-T scanners (General Electric
Medical Systems, Milwaukee, WI) using the body (n = 16) or torso
phased array (n = 24) coils. The imaging sequences and their
parameters were as follows:
Transverse T1-weighted spin-echo (n = 20), breath-hold spoiled gradient-echo (SPGR) (n = 15), or both (n = 5) sequences.—The spin-echo sequence parameters were TR range/TE range, 350–600 /14–20; number of excitations, 2; slice thickness, 8–9 mm; intersectional gap, 0–2 mm; matrix, 256 x 224–256 (frequency x phase); respiratory compensation; acquisition time, 4–7 min. The SPGR sequence parameters were 150–200/4.2–5.3; flip angle, 70°; number of excitations,1.0,; slice thickness, 7–8 mm; intersectional gap, 0–2 mm; matrix, 256–320 x 128–160 (frequency x phase); and acquisition time, 22–28 sec.
Transverse T2-weighted fat-suppressed fast spinecho (n = 40) and non–fat-suppressed coronal breath-hold half Fourier single-shot fast spin-echo (n = 22) sequences.—The fast spin-echo sequence parameters were 3650–4600/96–98; number of excitations, 2–3; slice thickness,7–8 mm; intersectional gap, 0–2 mm; matrix, 256 x 224–256 (frequency x phase); echo-train length, 8–12; respiratory triggering; and acquisition time, 2–6 min. The single-shot fast spin-echo sequence parameters were TR/TE, infinite/90; number of excitations, 0.5; slice thickness, 8 mm; intersectional gap, 0–2 mm; matrix, 256 x 128–160 (frequency x phase); and acquisition time, 18–28 sec.
Coronal T1-weighted dynamic breath-hold gadolinium-enhanced 3D SPGR (n = 32) or 2D SPGR (n = 8) imaging.—The 3D SPGR sequence parameters were 6.8–7.5/1.4–2.2; flip angle, 12–60°; number of excitations, 0.5; section thickness, 2.6–3.6 mm; matrix, 256–320 x 160–256 (frequency x phase); and acquisition time, 24–32 sec. The 2D SPGR sequence parameters were TR range/TE, 150–200/1.2; flip angle, 70°; number of excitations,1.0; slice thickness, 7–8 mm; intersectional gap, 0–2 mm; matrix, 256–512 x 128–160 (frequency x phase); fat suppression; and acquisition time, 22–28 sec.
After an unenhanced acquisition, gadolinium-enhanced imaging was performed in the corticomedullary (30 sec), nephrographic (60 sec), and 2-min delayed phases of enhancement. Gadolinium (Magnevist, Berlex, Wayne, NJ; or Omniscan, Amersham Health, Princeton, NJ) was administered at a dose of 0.1 mmol/kg of body weight (maximum, 30 mL) through a 20- to 24-gauge cannula placed in the antecubital fossa, by a hand injection in the earlier studies (n = 21) and at a rate of 2 mL/sec using a power injector (Spectris, Medrad, Pittsburgh, PA) in the later studies (n = 19). In all studies, the arterial phase acquisition was timed using the automated contrast bolus detection technique (Smartprep, General Electric Medical Systems).
Transverse gadolinium-enhanced 2D time-of-flight (n = 33) sequences.—The 2D time-of-flight parameters were 20/9; flip angle, 20°; number of excitations, 2.0; slice thickness, 5–7 mm; intersectional gap, 0–2 mm; and matrix 256 x 192 (frequency x phase).
Image Analysis
All images were independently reviewed on a workstation (Advantage, General
Electric Medical Systems) by two radiologists who had the same level of MRI
experience. The reviewers were aware of the patients' clinical histories but
unaware of the surgical and pathologic findings. The reviewers staged all
tumors according to the 1997 TNM staging system
[4]
(Table 1) and assessed the
extent of vascular thrombosis using the grading system of Gettman et al.
[11] (Fig.
1A,
1B,
1C,
1D,
1E). Two patients had bilateral
tumors, and those were treated as separate lesions.
|
|
|
|
|
|
The tumor diameter was measured in three planes (craniocaudal, anteroposterior, and transverse), and the largest of the three was chosen to represent the tumor size.
The diagnosis of perinephric fat invasion was made when there was loss of capsular integrity, which was indicated by interruption of the low-signal-intensity line around the kidney on T1- and T2-weighted images and thick (> 5 mm) perinephric stranding. Thin perinephric stranding and collateral vessel formation alone were not considered features of perinephric fat invasion. Invasion of Gerota's fascia was diagnosed when continuity of the fascia (i.e., the low-signal-intensity line around the perinephric fat on the T1-weighted images) was disrupted by tumor. The presence in an adjacent organ of a tumor mass in continuity with the renal tumor was considered diagnostic of adjacent organ invasion. Other features, including loss of the tumor–organ free-fat plane, irregular tumor–organ margin, signal changes, and abnormal enhancement in an adjacent organ, were suggestive but not diagnostic of invasion. Renal hilar, paraaortic, and paracaval nodes measuring more than 1 cm in short-axis diameter were considered to be metastatic.
In addition to assessing the extent of vascular thrombosis, the reviewers characterized the nature of the thrombus (tumor or bland) on the MRI. Direct continuity with the renal mass, high signal intensity, and signal heterogeneity compared with skeletal muscle on T2-weighted imaging, and enhancement after administration of gadolinium were considered features of tumor thrombus, whereas diffuse low signal intensity on T1-weighted, T2-weighted, and flow-sensitive gradient-echo sequences, uniform signal intensity, and lack of enhancement were considered features of bland thrombosis.
Statistical Analysis
MRI staging was compared with surgical and pathologic staging as the gold
standard, and agreement between the two staging systems was determined using
the kappa statistic (0.0–0.2, poor; 0.2–0.4, fair; 0.4–0.6,
moderate; 0.6–0.8, good; and 0.8–1.0, excellent). Agreement was
calculated separately for each of the two reviewers. Interobserver agreement
was also assessed using the kappa statistic.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
T Staging
Reviewer 1 correctly staged 36 tumors (86%), and reviewer 2 correctly
staged 34 tumors (81%) (Fig. 2
and Table 2).
|
|
For both reviewers, overstaging of T2 tumors as T3a was attributed to the large tumor sizes (mean, 14.2 cm; range, 8–19.3 cm) and the nodular tumor surface that made it difficult to exclude capsular invasion (Fig. 3). Both reviewers understaged the same T3a tumor because of the presence of a tumor pseudocapsule that mimicked the renal capsule. Both reviewers understaged the same T4 tumors. In one patient, the tumor was invading the outer wall of the inferior vena cava, which made it inoperable. In the other patient, macroscopic colon invasion was not visualized on MRI but was identified and resected at surgery. None of these findings was seen on retrospective review of the MRI studies.
|
N Staging
Both reviewers had the same results for N staging. Thirty-two tumors (76%)
were correctly staged and 10 (24%) were incorrectly staged
(Table 3). In five of six
overstaged tumors, the lymph nodes were larger than 1 cm (range, 1.2–3.5
cm) and were called metastatic nodes (Fig.
4). All these nodes were characterized as reactive on pathology.
In the sixth patient, overstaging was caused by incorrect labeling of tumor
thrombus in the retroaortic component of a circumaortic left renal vein as a
metastatic node. This finding was identified on retrospective review of the
MRI. Of four patients with understaged nodal involvement, one node was smaller
than 1 cm and the other three, although larger than 1 cm (3, 2, and 1.2 cm),
could not be identified separately from the large renal tumors (Fig.
5A,
5B).
|
|
|
|
M Staging
Only two patients had metastatic disease
(Table 3); one had
contralateral adrenal gland metastasis that was correctly staged by both
reviewers and surgically resected. The other patient had a metastatic deposit
in the liver dome that was excluded from the imaging volume. This metastasis
was identified and biopsied during surgery.
Venous Thrombus Assessment
Twenty-two patients had tumor thrombus in the venous system
(Fig. 6). No false-positive
diagnosis of tumor thrombus occurred. In 20 patients, both reviewers correctly
graded the tumor thrombus level. In two patients, both reviewers
underestimated the extent of tumor thrombus. In one, this was a result of the
inability to detect thrombus in the proximal 1 cm of the renal vein adjacent
to the renal hilum; the thrombus was seen only on pathologic examination. In
the other patient, tumor thrombus that appeared to be limited to the renal
vein on MRI was seen at surgery to extend into the lumen of the inferior vena
cava (Table 3). On
retrospective review of the images, the tumor thrombus was again seen to be
limited to the renal vein, with no caval extension. This error was thought to
be due to tumor thrombus progression during the 1-month interval between MRI
and surgery. Six patients had bland thrombus in the contralateral renal vein
(n = 1), the suprarenal inferior vena cava (n = 2), and the
infrarenal inferior vena cava (n = 3). All these patients had tumor
thrombus in the ipsilateral renal vein (n = 3) and renal vein and
inferior vena cava (n = 3). Both reviewers correctly characterized
the bland thrombus in all six patients
(Fig. 7).
|
|
MRI and Pathologic Staging
Excellent agreement ( = 0.93; 95% confidence interval,
0.85–1.00) was seen between reviewers 1 and 2 for T staging. Both
reviewers had the same results for N, M, and venous involvement staging
( = 1). The results for agreement between MRI staging and pathologic
staging are shown in Table
4.
|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
The Robson and TNM systems are the most commonly used staging classifications in the imaging literature for renal cell carcinoma. Unlike the Robson system [5–10], the TNM staging classification has not been widely used in the radiology literature, possibly because of the many subgroups it includes [16]. The TNM classification of renal cell carcinoma was first described in 1978 by the International Union Against Cancer (UICC) [17]. Since then, the system has undergone several revisions in an attempt to improve international communication in terms of epidemiology, survival, and the evaluation of different therapeutic approaches to the treatment of renal cell carcinoma [11, 18].
The T (tumor) component of the TNM staging system is the most important variable in predicting prognosis and survival after surgery. It is determined primarily by the size and extent of tumor and the extent of venous involvement [19, 20]. Studies suggested that the prognostic outcome of patients with T1 and T2 tumors was poorly differentiated using the previously established breakpoint of 2.5 cm [21, 22]. For this reason, the TNM classification was revised in 1997 by the American Joint Committee on Cancer and the UICC, and the breakpoint between T1 and T2 tumors was changed to 7 cm [4]. The rationale behind the 7-cm breakpoint was to create significant survival differences between TNM stage T1 and stage T2 disease. Tsui et al. [23], using multivariate analysis, assessed the prognostic indicators of renal cell carcinoma in 643 patients staged according to the revised 1997 TNM staging classification. Those authors found that the 5-year survival rate is 83% for stage T1 compared with 57% for stage T2 and 42% for stage T3.
Previous studies assessing the accuracy of MRI in staging renal cell carcinoma have used the Robson staging system [5–10]; we found no study assessing the accuracy of MRI for staging using the 1997 TNM classification system. Therefore, it is difficult to compare our results with those of other studies, especially for T staging.
T staging with MRI was highly accurate in our study. We found assessing capsular and Gerota's fascia invasion to be problematic, resulting in overstaging of T2 tumors as T3a. This finding is probably a result of the relatively large T2 tumors in our study (mean size, 14.2 cm). Such tumors compress the perinephric fat and obscure the renal capsule. Difficulty in excluding perinephric fat invasion in large tumors is a well-recognized problem [2, 5, 7]. Narumi et al. [10] reported the accuracy of MRI for assessing local extension of renal cell carcinoma into the perinephric fat and beyond Gerota's fascia to be 62–81%, depending on the sequences used. Although our accuracy for T staging is comparable to that reported in the Narumi study, we used 7 cm instead of 2.5 cm as the cutoff point between T1 and T2 tumors according to the 1997 TNM staging classification. Similarly, Fein et al. [7] reported that the major reason for renal cell carcinoma staging errors is the inability to reliably identify or exclude perinephric fat invasion. Although detection of perinephric fat involvement may not be clinically important when total or radical nephrectomy is contemplated because the kidney and perinephric fat are removed en bloc [1, 23], such involvement may preclude nephron-sparing surgery for smaller tumors. Tumor size is not a good predictor for the presence of perinephric fat invasion. In our study, a major overlap occurred between the sizes of tumors without and with perinephric fat invasion: mean size of T1 tumors, 3.4 cm (range, 0.8–7 cm); of T2 tumors, 14.2 cm (range, 8–19.3 cm); and of T3a tumors, 9.2 cm (range, 7.9–12 cm). The tumors of all four patients in our study who underwent partial nephrectomy were correctly staged as T1 (size range, 0.8–3 cm).
Various criteria have been used in the imaging literature to describe the appearance of perinephric fat invasion. Perinephric stranding and visible perinephric collateral vessels have been described in most Robson stage II tumors (T3A, N0, M0) [24]. However, perinephric fat stranding has also been detected in up to 50% of Robson stage I tumors (T1, N0, M0 and T2, N0, M0) on CT [5]. Thus, this stranding is nonspecific and is attributed to thickening of the perinephric septa from edema, vascular engorgement, or fibrosis [2, 24]. Johnson et al. [5] used the presence of discrete soft-tissue masses of 1 cm or larger as the indicator for perinephric fat invasion on CT. This finding had a high specificity (98%) but low sensitivity (46%), and the authors stated that perinephric fat invasion was difficult to detect on CT.
Difficulty in detecting adjacent organ invasion was another cause of errors in T staging in our study. Two of three T4 tumors were understaged as T3b by both reviewers. In one patient, the tumor was invading the outer wall of the inferior vena cava and could not be resected. Although no tissue sample was taken to prove the presence of caval wall invasion in this patient, the tumor was surgically considered to be a T4 tumor. In the other patient, failure to identify the colonic wall invasion on MRI was believed to be the result of susceptibility artifacts from intraluminal gas and motion artifacts from peristalsis, which obscured the anatomic details on the T2-weighted images. Hricak et al. [9] reported accuracy rates of 97–100% for detecting adjacent organ invasion with MRI. In their series, overstaging was caused by the presence of abnormal signal, indistinct interface, and the absence of a free-fat plane between the tumor and the adjacent organ. The limited data on the accuracy of cross-sectional imaging for detecting adjacent organ invasion are expected. Once extensive adjacent organ invasion or widespread metastasis is detected on imaging, the patient does not undergo surgery as the initial treatment. Instead, other types of palliative therapies, such as interferon and interleukin-2 [25], are used. Thus, the imaging findings cannot be confirmed. No reliable MRI criteria exist for the detection of adjacent organ invasion. Neither the indistinct tumor organ interface nor the abnormal signal intensity in the adjacent organ is a sufficiently accurate or reliable finding. In the absence of a mass in the neighboring organ, the diagnosis of adjacent organ invasion cannot be reliably made on MRI. Similar difficulties have been observed with CT. A sensitivity of 60% and a specificity of 100% have been reported [5] when signs such as adjacent organ enlargement and density change, in addition to loss of the tumor or adjacent organ fat plane, have been used.
Difficulty in detecting metastatic regional lymph nodes is another
well-known problem in staging tumors that applies to CT
[5,
26] and MRI because size (
1 cm) is the only imaging criterion used to determine the presence of
pathologic nodes. This criterion is neither a sensitive nor a specific
feature. Struder et al. [26]
reported that this criterion is associated with a 4% false-negative rate. In
their study of 163 patients with renal cell carcinoma, more than 50% of nodal
enlargements were caused by benign inflammatory changes, probably the result
of extensive tumoral necrosis or venous thrombosis. In our study, most
overstagings (5/6) were caused by the presence of reactive nodes larger than 1
cm. In all these cases, the mean tumor size was 8.3 cm (range, 6.4–11.5
cm), and a tumor thrombus was present in the renal vein (n = 3) or
inferior vena cava (n = 2). One of four underestimated N stages in
our study was due to the presence of microscopic metastasis in a node that was
smaller than 1 cm. The inability to show microscopic metastasis on MRI is
known and has been reported in many studies
[2,
7,
27]. It is possible that with
the introduction of new iron oxide–based contrast agents that are taken
up specifically by lymph nodes, the accuracy of MRI in nodal staging will
improve. Early results from some studies have been encouraging (Harisinghani
MG et al. and Deserno WMLLG et al., presented at International Society for
Magnetic Resonance in Medicine meeting, May 2002). Occasionally, and with
large tumors, the regional nodes and the tumor form a large conglomerate mass,
and the individual elements cannot be distinguished, which was another cause
of underestimating the N stage in two of our patients.
Only two patients in our study had metastatic disease, one of which was missed because of inadequate imaging coverage. Extensive coverage is desirable to exclude all possible metastases, but it is not always practical with MRI because of the limited coverage provided by the available torso phased array coils and the time penalty incurred when imaging large volumes, especially during dynamic imaging. We perform our scanning in the coronal plane to ensure maximal coverage, including the entire liver for complete staging. The introduction of parallel imaging reconstruction algorithms may help overcome this problem. Furthermore, lung metastasis cannot be reliably excluded on MRI and requires evaluation with CT.
Detecting and assessing the extent and nature of vascular thrombosis is highly accurate and reliable with MRI. Our results are similar to those from other studies performed using ECG-gated or nongated unenhanced and gadolinium-enhanced flow-sensitive gradientecho sequences [10, 28–30], 3D gadolinium-enhanced [31], or T1- and T2-weighted sequences [9, 32, 33]. Extension into the renal vein and the inferior vena cava occurs in approximately 25% and 10%, respectively, of cases of renal cell carcinoma [34]. Accurate preoperative assessment for the presence and the extent of renal vein and inferior vena cava tumor thrombus is important for planning subsequent treatment and choosing the appropriate surgical approach. [3, 34]. Although the presence and extent of tumor thrombus in the inferior vena cava has limited influence on the prognosis even when the thrombus extends to the right atrium [3], a significant difference in prognosis exists between patients with tumor invasion of the inferior vena cava wall and those with free-floating tumor thrombus in the inferior vena cava. Hatcher et al. [34] reported a difference in the 5-year survival rate between the two groups of 25% and 69%, respectively. The survival rate can be improved if the involved segment of inferior vena cava is resected [3]. The presence of bland thrombus in the veins does not alter the tumor stage but may influence the surgical approach. Bland thrombus does not adhere to the wall of the vein and can be easily extracted. MRI can reliably differentiate between tumor and bland thrombus; the most useful sequences for this purpose are the gradient-echo sequences, in particular, the flow-sensitive (time-of-flight) sequence. Because of the T2 shortening effect of blood breakdown products, bland thrombus is distinctly low in signal compared with the intermediate-signal-intensity tumor thrombus. On CT, Johnson et al. [5] reported a sensitivity of 78% and a specificity of 100% for detecting tumor extension into the renal vein and inferior vena cava when venous enlargement or identifiable thrombus was used as the imaging criterion for invasion. Welch et al. [35] reported a sensitivity of 85% and a specificity of 98% for detecting renal vein thrombosis on helical and electron-beam CT. Because of its multiplanar capability, MRI is the preferred technique to image tumor extension into the inferior vena cava, but with helical and MDCT scanners, multiplanar reconstruction capabilities, and appropriate timing of the contrast-enhanced acquisitions, CT is also effective in depicting the superior extent of inferior vena cava thrombus [2].
We recognize the limitations of our study. First, our study group was a highly selected patient population, with most of the patients having previously diagnosed renal cell carcinoma and being referred by the physician for preoperative staging; thus, most of our patients had advanced rather than early-stage disease. Second, because of the retrospective nature of our study, heterogeneity in the imaging techniques, parameters, and coil types may have influenced the interpretation and may have underestimated the true performance of state-of-the-art MRI. Third, it was difficult to directly compare our results with those reported in other studies because of limited use of the 1997 TNM staging classification in other studies. Furthermore, many studies report overall accuracy rates rather than accuracy rates for each group within the staging system.
In conclusion, MRI is a reliable method for preoperative staging of renal cell carcinoma, especially for the evaluation of venous involvement. Similar to CT, MRI is limited in evaluating the T stage, especially for large tumors, for which assessment of perinephric fat, Gerota's fascia, and adjacent organ invasion cannot always be reliably made. Suboptimal N staging is the result of reliance on node size for the assessment of metastatic status.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
CiteULike 
Complore 
Connotea 
Del.icio.us 
Digg 
Reddit 
Technorati What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
