Genitourinary Imaging

Tekes et al.
MRI of Bladder Cancer
Downloaded from www.ajronline.org by 87.8.209.140 on 03/11/15 from IP address 87.8.209.140. Copyright ARRS. For personal use only; all rights reserved

Dynamic MRI of Bladder Cancer:

Evaluation of Staging Accuracy
Aylin Tekes1 OBJECTIVE. The purpose of this study was to evaluate the accuracy of gadolinium-
Ihab Kamel1 enhanced MRI in staging bladder cancer in a series of patients with surgically proven blad-
Khursheed Imam1 der cancer.
Gilberto Szarf1 MATERIALS AND METHODS. Seventy-one patients with biopsy-proven bladder can-
cer underwent MRI on a 1.5-T scanner with a phased-array pelvic coil. Conventional T1-
Mark Schoenberg2
weighted spin-echo, T2-weighted spin-echo, and unenhanced and enhanced (0.1 mmol/kg gad-
Khurram Nasir3
olinium) fast spoiled gradient-echo images with fat suppression were obtained. Two blinded re-
Richard Thompson4 viewers evaluated the MR images and assigned a stage that was compared with the pathologic
David Bluemke1 stage (n = 67) or with clinical follow-up for at least 2 years after MRI (n = 4).
Tekes A, Kamel I, Imam K, et al.

RESULTS. Agreement among the reviewers was good in assigning a radiologic stage for
bladder cancer (kappa = 0.80). On a stage-by-stage basis, MRI accuracy was 62%, and over-
staging was the most common error (32%). Staging accuracy improved to 85% and 82% in dif-
ferentiating superficial from invasive tumors and organ-confined from non-organ-confined
tumors, respectively. The time interval between MRI and transurethral resection (≤ 60 days and
≥ 61 days) was not a statistically significant factor in differentiating superficial from invasive
and organ-confined from non-organ-confined tumors (p > 0.05). MRI accuracy in staging tran-
sitional cell carcinoma was not significantly different from that obtained in staging non–tran-
sitional cell carcinoma (p > 0.05).
CONCLUSION. MRI shows good reproducibility between reviewers for staging blad-
der cancer. Although overall staging accuracy was only moderate, the accuracy for differen-
tiating superficial versus invasive disease and organ-confined versus non-organ-confined
disease was high.

arcinoma of the urinary bladder is travesical tumors show significantly higher

C one of the more common malig-

Received February 24, 2004; accepted after revision
June 14, 2004.
1Russell H. Morgan Department of Radiology, Johns
Hopkins School of Medicine, 601 N Caroline St.,
Ste. 3235A, Baltimore, MD 21287. Address correspondence to
I. Kamel (ikamel@jhmi.edu).
2Department of Urology, Brady Urological Institute, Johns
Hopkins School of Medicine, Baltimore, MD 21287.
3Johns Hopkins School of Public Health, Baltimore, MD
21287.
4Department of Cardiology, Johns Hopkins School of
Medicine, Baltimore, MD 21287.
AJR 2005;184:121–127
0361–803X/05/1841–121
© American Roentgen Ray Society
nant tumors of the urinary tract in
both male and female patients [1].
Accurate preoperative staging is the most im-
portant factor in determining the appropriate
management of bladder carcinoma because
the therapeutic method chosen and prognosis
depend on the clinical and radiologic stage at
presentation [2]. Superficial tumors can be
treated by local endoscopic resection with or
without adjuvant installation of chemothera-
peutic agents, whereas invasive tumors are
treated by curative cystectomy or by palliative
chemotherapy or radiation therapy. Clinical
staging can differentiate superficial tumors
from invasive tumors [3]. However, clinical
staging is not reliable for determining tumor
extension beyond the bladder wall; therefore,
imaging of the urinary bladder and the extra-
vesical pelvis is needed [3]. Patients with ex-
recurrence rates and worse survival than those
with organ-confined tumors [4]. Therefore,
distinguishing between organ-confined and
non-organ-confined tumors is essential.
Prior studies have reported that CT was a
valuable tool in staging bladder carcinoma [5,
6]. MRI with dynamic contrast administration
has been shown to be superior to CT, particu-
larly in detecting superficial and multiple tu-
mors and in detecting extravesical tumor
extension and surrounding organ invasion [1,
5, 7–12].
The purposes of this study were to evaluate
the overall accuracy of state-of-the-art dy-
namic gadolinium-enhanced MRI in staging
bladder cancer on a stage-by-stage basis and
to determine the usefulness of MRI in deter-
mining organ-confined versus non-organ-
confined disease, which is the main objective
of imaging these patients.

AJR:184, January 2005 121


Materials and Methods
Patient Selection
Inclusion criteria for this study were that pa-
tients had undergone state-of-the art dynamic en-
hanced MRI examination, with bladder carcinoma
pathologically documented within 6 months of imag-
ing. A search of our database identified MRI exami-
Downloaded from www.ajronline.org by 87.8.209.140 on 03/11/15 from IP address 87.8.209.140. Copyright ARRS. For personal use only; all rights reserved
nations of 71 consecutive patients with histologically
proven bladder cancer who underwent imaging be-
tween January 1998 and June 2001. Permission to
review these clinical MRI examinations and medi-
cal records for scientific research was granted be-
fore the study.
MRI Technique
Patients were imaged using a 1.5-T MR scanner
(Signa, GE Healthcare) with a phased-array pelvic
coil. Conventional T1-weighted spin-echo images
(TR/TE, 550/9; 512 × 192 matrix; 20-cm field of
view; 6-mm section thickness; 2-mm intersection
gap; 4 signals acquired) and T2-weighted fast spin-
echo images (TR range/TE range, 4,000–5,500/80–
120; 256 × 256 matrix; 24-cm field of view; 6-mm
section thickness; 2-mm intersection gap; 4 signals
acquired) were obtained. Subsequently, fast multi-
planar spoiled gradient-echo images with fat sup-
pression (180–300/1.7–4.2; 70° flip angle; 512 × 92
matrix; 20-cm field of view; 6-mm slice thickness;
2-mm intersection gap; 2 signals acquired) were
obtained in the axial plane before and after gado-
pentetate dimeglumine (Magnevist, Berlex) injec-
tion (0.1 mmol/kg). Enhanced images were
acquired during the arterial phase (20 sec), which
was immediately followed by the venous phase.
The acquisition time was 52–86 sec for each phase.
Sagittal and coronal gadolinium-enhanced images
were added if the tumor was located in the base or
the dome of the bladder. These additional images
were acquired in nine patients.
Diagnostic MRI Criteria
MR images were interpreted independently by
two MR radiologists with special interest in uro-
logic imaging without prior knowledge of the final
staging obtained at transurethral resection, cystec-
tomy, or clinical follow-up. Each reviewer assigned
a radiologic stage using criteria similar to those pre-
viously described in the literature.
On T2-weighted images, the normal bladder
wall was identified as a hypointense line outlining
the bladder lumen [5–7, 10, 13, 14]. On dynamic
contrast-enhanced MR images, bladder tumors,
mucosa, and submucosa (lamina propria) enhanced
early, but the muscle layer maintained its hy-
pointensity [9, 15].
An intact, hypointense line (muscle layer) at the
base of the tumor was classified as stage T1; an ir-
regular inner margin of hypointense line, stage T2a;
122
Tekes et al.
a disrupted hypointense line without perivesical fat
infiltration, stage T2b; a lesion with an irregular,
shaggy outer border and streaky areas of the same
signal intensity of the tumor in perivesical fat, stage
T3b; and a lesion extending into an adjacent organ or
abdominal and pelvic side walls with the same signal
intensity of the primary tumor, stage T4a or T4b, re-
spectively [8]. Lymph nodes were considered abnor-
mal if the long axis was 10 mm or more [8].
All patients included in our study were found to be
free of distant metastasis before they were referred for
MRI. Their workup for distant metastasis included
chest CT, abdominal MRI, and bone scanning.
Data Analysis
Data were analyzed using STATA software (ver-
sion 7, Stata). Continuous variables were expressed
as means ± SD. Interobserver agreement between
the two MRI reviewers was calculated using kappa
statistics. Agreement between observers was char-
acterized by weighted kappa values and correlation
coefficients. Kappa scores between 0.41 and 0.6
were considered moderate agreement; 0.61–0.80,
good agreement; and greater than 0.80, excellent
agreement [16].
Sensitivity, specificity, and accuracy of MRI
were assessed on a stage-by-stage basis, and the
gold standard was pathologic confirmation in all
cases. Pathologic staging conformed to the updated
TNM system of the International Union Against
Cancer [17] (Table 1). In addition, the data were re-
grouped to evaluate the accuracy of MRI staging in
distinguishing superficial (≤ T1) from invasive
(≥ T2) tumors and organ-confined (≤ T2b) from
non-organ-confined (≥ T3) tumors.
To analyze the effect of the time interval be-
tween MRI and prior transurethral resection on
staging accuracy, we classified patients into two
groups: patients who had transurethral resection 60
or fewer days before MRI and patients who had
transurethral resection 61 or more days after MRI.
Staging accuracy was evaluated separately for tran-
sitional cell carcinomas versus non–transitional
cell carcinomas. Significant differences were de-
clared for p values less than 0.05.
Results
A total of 71 patients were included in the
study: 62 men and nine women ranging in age
between 38 and 88 years (mean, 64 years).
Before MRI, all patients underwent clinical
staging including cystoscopy and bimanual
examination. In addition, 62 patients under-
went transurethral resection 7–165 days
(mean, 61 days) before MRI. After MRI,
treatment was radical cystectomy (n = 39),
partial cystectomy (n = 2), transurethral tu-
mor resection (n = 26), or palliative radiation
therapy (n = 3). The remaining patient had a
history of in situ tumor and received intraves-
ical chemotherapy before MRI. After MRI,
the patient had urine analysis and cystoscopy
and was found to be disease-free. Therefore,
no additional treatment was necessary.
All patients received treatment within 150
days (mean, 31 days) after MRI. Forty-one
(58%) of the 71 patients had pelvic lym-
phadenectomy. For the remaining 30 patients
(42%), the absence of lymph node involvement
was established by clinical follow-up and MRI
studies at 6-month intervals for at least 2 years.
Histologic diagnoses were transitional cell car-
cinoma (n = 60), squamous cell carcinoma (n =
1), adenocarcinoma (n = 6), small cell carci-
noma (n = 1), and carcinosarcoma (n = 3).
Tumor Appearance
The final pathologic staging revealed 24 pa-
tients with stage Ta–T1 disease, 10 with stage
T2b, 21 with stage T3a–b, five with T4a, and two
with stage T4b. None of the patients had stage
T2a tumor. Nine patients were stage T0. These
patients initially had in situ or lamina propria in-
TNM Classification for
TABLE 1 Bladder Cancer (1997)
[17]
Stage Characteristics
Ta Noninvasive papillary carcinoma
Tis Carcinoma in situ: “flat” tumor
T1 Tumor invades subepithelial connective
tissue
T2a Tumor invades superficial muscle
(inner half)
T2b Tumor invades deep muscle (outer half)
T3a Tumor invades perivesical tissue
microscopically
T3b Tumor invades perivesical tissue
macroscopically
T4a Tumor invades prostate or uterus or
vagina
T4b Tumor invades pelvic wall or abdominal
wall
N0 No regional lymph node metastasis
N1 Metastasis in a single lymph node ≤ 2 cm
in greatest dimension
N2 Metastasis in a single lymph node > 2 cm
but ≤ 5 cm in greatest dimension, or
multiple lymph nodes
N3 Metastasis in lymph node > 5 cm in
greatest dimension
M0 No distant metastasis
M1 Distant metastasis
AJR:184, January 2005
 MRI of Bladder Cancer

vasive tumors, and they received intravesical che- TABLE 2 Staging Results in 71 Patients on Stage-by-Stage Basis (Reviewer 1)
motherapy. MRI was indicated to stage their
disease. After MRI, biopsy confirmed the ab- Histopathology Stage
MRI Stage
sence of disease. These patients were followed up T0 Ta–T1 T2a T2b T3a–T3b T4a T4b Total
every 6 months for 2 years, and none of the fol-
low-up examinations revealed tumor recurrence. T0 5 0 0 0 0 0 0 5
Tumors were detected in 62 (87%) of the 71
Downloaded from www.ajronline.org by 87.8.209.140 on 03/11/15 from IP address 87.8.209.140. Copyright ARRS. For personal use only; all rights reserved

Ta–T1 1 16 0 0 0 1 0 18
patients on pathologic confirmation. Of these T2a 1 1 0 0 1 0 0 3
patients, 45 patients (63%) had mass lesions, T2b 2 4 0 6 2 0 0 14
whereas 17 (24%) had diffuse wall thickening.
T3a–T3b 0 2 0 4 11 0 0 17
Of the mass lesions, 20 were papillary and 25
were sessile. Tumor size ranged from 0.5 to 7.3 T4a 0 1 0 0 6 4 0 11
cm (mean, 2.5 cm). Twelve patients had multi- T4b 0 0 0 0 1 0 2 3
ple tumors; in such cases, the highest tumor Total 9 24 0 10 21 5 2 71
stage present was used for the analysis.
Note.—Numbers in boldface indicate number of correctly staged patients in each row. Sum of numbers under
All 62 detected tumors were isointense rel- each bold number represents overstaged cases; sum of numbers above each bold number represents
ative to bladder wall muscle on T1-weighted understaged cases.
images. On T2-weighted images, 50 tumors
(81%) were isointense and 12 (19%) were
slightly hyperintense relative to muscle. On Sixty-two tumors that were present at the time scores of both reviewers to perform addi-
dynamic contrast-enhanced MR images, all of imaging were detected correctly (sensitiv- tional statistical analysis. All T0 tumors (n =
tumors had increased enhancement compared ity of 100%). On a stage-by-stage basis, tu- 9) were excluded from further analysis be-
with uninvolved bladder. Fifty-three tumors mors were staged correctly in 44 (62%) of 71 cause they were found to be tumor-free af-
(85%) showed early, intense enhancement on patients (Figs. 1–3), overstaged in 23 patients ter stage-by-stage analysis. We evaluated
images obtained beginning 20 sec after gado- (32%) (Fig. 4), and understaged in four pa- the ability of MRI to distinguish between
linium administration. Eight tumors showed tients (6%) for reviewer 1 (Table 2). Reviewer superficial (those without muscle invasion)
heterogeneous enhancement at 20 sec, 2 correctly staged on a stage-by-stage basis and invasive tumors (Tables 4 and 5). Of
whereas one of the tumors showed superficial 37 (52%) of 71 tumors, overstaged tumors in the 62 tumors, 53 were staged correctly,
enhancement on the delayed images. 26 patients (37%), and understaged tumors in eight were overstaged, and one was under-
eight patients (11%) (Table 3). staged, yielding an overall accuracy of
Tumor Staging Staging accuracy was evaluated several 85%. We also evaluated the accuracy of
Interobserver agreement was good in as- ways to reflect clinical utility. Despite the MRI in classifying organ-confined (those
signing a radiologic stage (kappa = 0.80). good interobserver agreement, we used the within the bladder) versus non-organ-con-

A B
Fig. 1.—Images in 58-year-old man with correctly staged papillary (Ta) transitional cell carcinoma of bladder.
A, Axial T2-weighted image (TR/TE, 4,000/80) shows polypoid mass (arrow) arising from right posterolateral wall with homogeneous low signal intensity.
Note that low-signal muscular layer is intact.
B, Axial arterial phase gadolinium-enhanced image (200/2.9) shows bright, homogeneous enhancement of mass (arrow).

AJR:184, January 2005 123

 Tekes et al.
Downloaded from www.ajronline.org by 87.8.209.140 on 03/11/15 from IP address 87.8.209.140. Copyright ARRS. For personal use only; all rights reserved

A B
Fig. 2.—Images in 70-year-old man with correctly staged T2b transitional cell carcinoma of bladder.
A, Axial T2-weighted image (TR/TE, 4,000/80) shows that sessile mass arising from right lateral wall (arrow) disrupts low-signal-intensity muscle layer.
B, Axial arterial phase gadolinium-enhanced image (200/2.9) shows early enhancement of sessile mass (arrow).

fined tumors. Of the 62 cases, 51 were cor-
rectly classified, seven were overstaged,
and four were understaged, yielding an
overall accuracy of 82%.
The effect of the time interval between
MRI and biopsy on staging accuracy was as-
sessed. One group included patients who un-
derwent transurethral resection 60 or fewer
days (mean, 33 days) before MRI (n = 34),
and the second group (n = 28) included pa-
tients who had transurethral resection 61 or
more days (mean, 95 days) before MRI.
Staging accuracy between the two groups
was not statistically different in classifying
superficial tumors from invasive tumors or in
differentiating organ-confined tumors from
non-organ-confined ones (p > 0.05) (Tables
6 and 7).
Eleven patients (15%) had non–transi-
tional cell carcinoma, and 60 patients
(85%) had transitional cell carcinoma.
Seven (64%) of the 11 cases of non–transi-
tional cell carcinoma were staged correctly,
and 37 (62%) of the 60 cases with transi-

A B
Fig. 3.—Images in 68-year-old man with correctly staged T4a transitional cell carcinoma of bladder.
A, Axial T2-weighted image (TR/TE, 4,000/80) obtained above level of mass shows two diverticula arising from ureterovesical junction bilat-
erally (white arrows) and left hydroureter (arrowhead). Note enlarged left external iliac lymph nodes (black arrows).
B, Axial arterial phase gadolinium-enhanced image (200/2.9) obtained slightly lower than A shows polypoid tumor with homogeneous
enhancement arising from left base of bladder (small short arrow) and extending into perivesical fat (arrowheads). Note that asymmetric
enhancement in left seminal vesicle (large short arrow) correlates with organ invasion at radical cystectomy. Enlarged right inguinal lymph
node (long arrow) also can be seen.

124 AJR:184, January 2005

 MRI of Bladder Cancer
Downloaded from www.ajronline.org by 87.8.209.140 on 03/11/15 from IP address 87.8.209.140. Copyright ARRS. For personal use only; all rights reserved

A B
Fig. 4.—Images in 56-year-old woman with stage T3b transitional cell carcinoma of bladder. Tumor was overstaged by both reviewers.
A, Axial T2-weighted image (TR/TE, 4,000/80) shows tumor with heterogeneous signal intensity disrupting bladder wall and causing left hydroureter
(large arrow). Note lymph node in left obturator chain (arrowhead). No clear fat plane (small arrows) is visible between uterus and tumor, suggest-
ing uterine invasion. No uterine invasion was detected at pathology.
B, Axial venous phase fast spoiled gradient-echo image (200/1.9) obtained after administration of gadolinium shows intense enhancement of
sessile mass arising from left lateral wall and filling most of bladder lumen. Note loss of fat plane (arrows) between uterus (arrowhead) and mass.

tional cell carcinoma were staged correctly.

TABLE 3 Staging Results in 71 Patients on Stage-by-Stage Basis (Reviewer 2)
Staging accuracy was not statistically dif-
ferent between transitional and non–transi- Histopathology Stage
tional cell carcinoma (p > 0.05). MRI Stage
T0 Ta–T1 T2a T2b T3a–T3b T4a T4b Total
Of the 71 patients, 10 had pathologic
lymph node involvement. Among the find- T0 6 0 0 0 0 0 0 6
ings for the 61 patients who were free of Ta–T1 1 11 0 1 0 1 0 14
lymph node involvement, there was one T2a 1 7 0 1 2 0 0 11
false-positive MR interpretation by both
T2b 1 2 0 3 3 0 0 9
reviewers. The false-positive lymph node
was benign at pathology, but on MRI it ex- T3a–T3b 0 3 0 4 11 0 0 18
ceeded 10 mm in the long axis (14 mm). T4a 0 1 0 1 5 4 0 11
Both reviewers correctly detected lymph T4b 0 0 0 0 0 0 2 2
node involvement in seven of 10 patients
Total 9 24 0 10 21 5 2 71
on MRI, resulting in an accuracy of 96%,
Note.—Numbers in boldface indicate number of correctly staged patients in each row. Sum of numbers under
sensitivity of 78%, and specificity of 98%. each bold number represents overstaged cases; sum of numbers above each bold number represents
understaged cases.
Discussion
Our study was performed on a large group
of patients with pathologically proven blad-
der carcinoma who underwent state-of-the-
art MRI for preoperative staging. We have
shown that MRI staging of bladder cancer is Accuracy of MRI in Differentiating Superficial (≤ T1)
highly reproducible among experienced re- TABLE 4 from Invasive (≥ T2) and Organ-Confined (≤ T2b)
viewers with pathologic confirmation of the from Non–Organ-Confined (≥ T3) Disease (Reviewer 1)
disease stage in all cases (kappa = 0.80). Our
Predictive Value
overall accuracy of dynamic MRI on a stage- Tumor Stage Sensitivity (%) Specificity (%) Accuracy (%)
by-stage basis was 62%. Overstaging was the Positive (%) Negative (%)
most common error (32%) in the current
study (Tables 2 and 3), contrary to the find- ≤ T1 vs ≥ T2 97 67 77 96 83 (74–92)
ings of a prior study by Buy et al. [10] who ≤ T2b vs ≥ T3 86 84 77 90 85 (77–94)
used a 0.5-T MR scanner without contrast ad- Note.—Nine T0 cases were excluded. Numbers in parentheses are 95% confidence interval.

AJR:184, January 2005 125

 Tekes et al.

Accuracy of MRI in Differentiating Superficial (≤ T1) by curative cystectomy and palliative chemo-
TABLE 5 from Invasive (≥ T2) and Organ-Confined (≤ T2b) therapy, radiation therapy, or both. One of the
from Non–Organ-Confined (≥ T3) Disease (Reviewer 2) most important reasons for performing preop-
erative imaging is distinction of organ-con-
Predictive Value
Tumor Stage Sensitivity (%) Specificity (%) Accuracy (%) fined disease from tumor that has spread
Positive (%) Negative (%) outside the bladder (Figs. 3 and 4). Although
clinical staging including transurethral resec-
Downloaded from www.ajronline.org by 87.8.209.140 on 03/11/15 from IP address 87.8.209.140. Copyright ARRS. For personal use only; all rights reserved

≤ T1 vs ≥ T2 95 55 71 90 76 (66–86) tion or biopsy can distinguish superficial

≤ T2b vs ≥ T3 79 79 71 55 79 (69–89) from invasive tumors, it is not capable of de-
Note.—Nine T0 cases were excluded. Numbers in parentheses are 95% confidence interval. tecting extravesical disease. This distinction
is important because patients with non-organ-
confined disease have higher recurrence rates
Effect of Time Interval Between MRI and Transurethral Resection in and worse survival rates than patients with or-
TABLE 6
Staging Accuracy (Reviewer 1) gan-confined disease [4].
Predictive Value Our staging accuracy was 85% for assess-
Tumor Stage Sensitivity (%) Specificity (%) Accuracy (%) ing superficial versus invasive disease and
Positive (%) Negative (%) was slightly lower than that reported in the lit-
≤ 60 days (n = 39) erature by Scattoni et al. [18], who reported
≤ T1 vs ≥ T2 96 50 81 86 82 (69–95)a
an accuracy of 92% using a 0.5-T MR scanner
with contrast administration. Our accuracy in
≤ T2b vs ≥ T3 86 71 79 80 79 (66–93)b
differentiating organ-confined from non-or-
≥ 61 days (n = 32) gan-confined tumors was 82% and also was
≤ T1 vs ≥ T2 100 76 69 100 84 (71–98)a lower than that reported in a 1988 study by
≤ T2b vs ≥ T3 83 92 71 96 91 (80–100)b Buy et al. [10]; those researchers reported an
Note.—Nine T0 cases were excluded. Numbers in parentheses are 95% confidence interval. accuracy of 95% using a 0.5-T MR scanner
aFor differentiating ≤ T1 vs ≥ T2 between ≤ 60 days and ≥ 61 days, p > 0.05. without any contrast administration. How-
bFor differentiating ≤ T2b vs ≥ T3 between ≤ 60 days and ≥ 61 days, p > 0.05.
ever, our accuracy in differentiating organ-
confined from non-organ-confined tumors
was higher than the 73% accuracy reported in
Effect of Time Interval Between MRI and Transurethral Resection in a previous study that included patients who
TABLE 7
Staging Accuracy (Reviewer 2) underwent transurethral resection 7–16 days
Predictive Value before MRI [11].
Tumor Stage Sensitivity (%) Specificity (%) Accuracy (%) Management of bladder carcinoma starts
Positive (%) Negative (%) with cystoscopy, bimanual examination, and
≤ 60 days (n = 39) transurethral resection. The value of transure-
≤ T1 vs ≥ T2 96 42 79 83 79 (66–93)a
thral resection is to confirm the histology and
to stage the tumor. In addition, transurethral
≤ T2b vs ≥ T3 82 71 78 75 77 (63–91)b
resection alone or combined with intravesical
≥ 61 days (n = 32) chemotherapy may provide definitive therapy
≤ T1 vs ≥ T2 91 62 56 93 72 (55–88)a for superficial bladder tumors, which account
≤ T2b vs ≥ T3 67 85 50 92 81 (67–96)b for approximately two thirds of all bladder
Note.—Nine T0 cases were excluded. Numbers in parentheses are 95% confidence interval. carcinomas. Transurethral resection before
aFor differentiating ≤ T1 vs ≥ T2 between ≤ 60 days and ≥ 61 days, p > 0.05. MRI has been suggested as a possible cause
bFor differentiating ≤ T2b vs ≥ T3 between ≤ 60 days and ≥ 61 days, p > 0.05.
of overstaging bladder carcinoma [19], be-
cause the differentiation between acute
ministration and reported an accuracy of field-strength scanners without dynamic con- edema or hyperemia due to transurethral re-
60%. The most common staging error re- trast administration and a small sample size section and tumor is stated to be difficult, es-
ported in that study was underestimation [5, 7, 13]. pecially immediately after transurethral
(33%) of tumor extent. The improvement of Clinical management primarily is based on resection [12, 15, 20–23]. The differentiation
image resolution in our study likely led to im- distinguishing superficial (Fig. 1) from mus- is difficult on T2-weighted images and gado-
proved detection of perivesical fat stranding, cle-invasive (Fig. 2) disease. Approximately linium-enhanced images [8, 15, 23, 24]. All
which was frequently reactive or inflamma- two thirds of all bladder cancers are superfi- patients referred for MRI at our institution un-
tory rather than metastatic but was thought by cial, and treatment options differ dramatically dergo transurethral resection or biopsy before
our reviewers, on occasion, to represent extra- between superficial and invasive disease. Su- MRI. Therefore, excluding these patients is
vesical spread of tumor. Our staging accuracy perficial tumors are treated with local endo- not a practical approach before radiologic
was lower than those previously reported, scopic resection with or without adjuvant staging of bladder carcinoma.
which ranged between 72% and 95%. How- intravesical installations of chemotherapeutic To evaluate the effect of the time interval
ever, those studies were performed using low- agents, whereas invasive tumors are treated between transurethral resection and MRI on

126 AJR:184, January 2005


staging accuracy, we classified the patients
into two groups. Our analyses show that there
is no statistically significant difference in
staging accuracy between groups with a short
(≤ 60 days) versus long (≥ 61 days) duration
between MRI and transurethral resection or in
differentiating superficial from invasive dis-
Downloaded from www.ajronline.org by 87.8.209.140 on 03/11/15 from IP address 87.8.209.140. Copyright ARRS. For personal use only; all rights reserved
ease and organ-confined from non-organ-
confined disease. These findings are in agree-
ment with a previous study that showed there
was no statistically significant difference in
staging 1 week (early) versus 4 weeks (late)
after transurethral resection [1]. However, in
that study, both time intervals between trans-
urethral resection and MRI were within the
short-duration group that we defined in the
current study.
Our study group consisted of 11 patients with
non–transitional cell carcinomas, including
squamous cell carcinoma, adenocarcinoma,
small cell carcinoma, and carcinosarcoma.
Non–transitional cell carcinoma is reported to
differ from transitional cell carcinoma. These
tumors are less common and generally are
larger than transitional cell carcinomas. In addi-
tion, they are aggressive and usually extend be-
yond the bladder wall at the initial time of
diagnosis [25–27]. Our results show that there is
no statistically significant difference in the accu-
racy of MRI for staging transitional and non–
transitional cell carcinomas on a stage-by-stage
basis and for differentiating superficial from in-
vasive and organ-confined from non-organ-con-
fined tumors.
We conclude that MRI staging of bladder
tumors is reproducible among reviewers. On
a stage-by-stage basis, the accuracy of MRI is
62%, with overstaging being the most com-
mon error. However, when certain specific
features of bladder cancers are assessed, MRI
is considerably more accurate. Overall, stag-
ing accuracy of MRI is 85% and 82% in dif-
ferentiating superficial tumors from invasive
disease and organ-confined tumors from non-
organ-confined disease, respectively. Contin-
ued improvements in MR hardware, such as
the advent of higher-magnetic-field-strength
AJR:184, January 2005
MRI of Bladder Cancer
MR scanners and higher-resolution imaging
techniques may aid in improving staging ac-
curacy further.
References
1. Barentsz JO, Jager GJ, Witjes JA, Ruijs JH. Primary
staging of urinary bladder carcinoma: the role of MRI
and a comparison with CT. Eur Radiol 1996;6:129–133
2. MacVicar AD. Bladder cancer staging. BJU Int
2000;86[suppl 1]:111–122
3. Barentsz JO, Witjes JA, Ruijs JH. What is new in
bladder cancer imaging. Urol Clin North Am
1997;24:583–602
4. Stein JP, Lieskovsky G, Cote R, et al. Radical cys-
tectomy in the treatment of invasive bladder can-
cer: long-term results in 1,054 patients. J Clin
Oncol 2001;19:666–675
5. Amendola MA, Glazer GM, Grossman HB, Aisen
AM, Francis IR. Staging of bladder carcinoma: MRI–
CT–surgical correlation. AJR 1986;146:1179–1183
6. Bryan PJ, Butler HE, LiPuma JP, Resnick MI, Kursh
ED. CT and MR imaging in staging bladder neo-
plasms. J Comput Assist Tomogr 1987;11:96–101
7. Fisher MR, Hricak H, Tanagho EA. Urinary blad-
der MR imaging. Part II. Neoplasm. Radiology
1985;157:471–477
8. Kim B, Semelka RC, Ascher SM, Chalpin DB,
Carroll PR, Hricak H. Bladder tumor staging:
comparison of contrast-enhanced CT, T1- and T2-
weighted MR imaging, dynamic gadolinium-en-
hanced imaging, and late gadolinium-enhanced
imaging. Radiology 1994;193:239–245
9. Tanimoto A, Yuasa Y, Imai Y, et al. Bladder tumor
staging: comparison of conventional and gadolin-
ium-enhanced dynamic MR imaging and CT. Ra-
diology 1992;185:741–747
10. Buy JN, Moss AA, Guinet C, et al. MR staging of
bladder carcinoma: correlation with pathologic
findings. Radiology 1988;169:695–700
11. 11. Husband JE, Olliff JF, Williams MP, Heron CW,
Cherryman GR. Bladder cancer: staging with CT
and MR imaging. Radiology 1989;173:435–440
12. Tachibana M, Baba S, Deguchi N, et al. Efficacy of
gadolinium-diethylenetriaminepentaacetic acid-en-
hanced magnetic resonance imaging for differentia-
tion between superficial and muscle-invasive tumor
of the bladder: a comparative study with computer-
ized tomography and transurethral ultrasonography.
J Urol 1991;145:1169–1173
13. Rholl KS, Lee JK, Heiken JP, Ling D, Glazer HS.
Primary bladder carcinoma: evaluation with MR
imaging. Radiology 1987;163:117–121
14. Wood DP Jr, Lorig R, Pontes JE, Montie JE.
The role of magnetic resonance imaging in the
staging of bladder carcinoma. J Urol 1988;140:
741–744
15. Neuerburg JM, Bohndorf K, Sohn M, Teufl F,
Guenther RW, Daus HJ. Urinary bladder neo-
plasms: evaluation with contrast-enhanced MR
imaging. Radiology 1989;172:739–743
16. Landis JR, Koch GG. The measurement of ob-
server agreement for categorical data. Biometrics
1977;33:159–174
17. Sobin LH, Fleming ID. TNM classification of malig-
nant tumors, fifth edition (1997): Union Internationale
Contre le Cancer and the American Joint Committee
on Cancer. Cancer 1997;80:1803–1804
18. Scattoni V, Da Pozzo LF, Colombo R, et al. Dy-
namic gadolinium-enhanced magnetic resonance
imaging in staging of superficial bladder cancer. J
Urol 1996;155:1594–1599
19. Barentsz JO, Jager GJ, van Vierzen PB, et al.
Staging urinary bladder cancer after transurethral
biopsy: value of fast dynamic contrast-enhanced
MR imaging. Radiology 1996;201:185–193
20. Barentsz JO, Ruijs SH, van Erning LJ. Magnetic
resonance imaging of urinary bladder cancer: an
overview and new developments. Magn Reson Q
1993;9:235–258
21. Doringer E, Joos H, Forstner R, Schmoller H. MRT of
bladder carcinoma: tumor staging and gadolinium
contrast behavior [in German]. Rofo Fortschr Geb
Rontgenstr Neuen Bildgeb Verfahr 1991;154:357–363
22. Neuerburg JM, Bohndorf K, Sohn M, Teufl F,
Gunther RW. Staging of urinary bladder neo-
plasms with MR imaging: is Gd-DTPA helpful? J
Comput Assist Tomogr 1991;15:780–786
23. Sparenberg A, Hamm B, Hammerer P, Samberger
V, Wolf KJ. The diagnosis of bladder carcinomas
by NMR tomography: an improvement with Gd-
DTPA [in German]? Rofo Fortschr Geb Rontgen-
str Neuen Bildgeb Verfahr 1991;155:117–122
24. Sohn M, Neuerburg J, Teufl F, Bohndorf K. Gado-
linium-enhanced magnetic resonance imaging in the
staging of urinary bladder neoplasms. Urol Int
1990;45:142–147
25. Tekes A, Kamel IR, Chan TY, Schoenberg MP,
Bluemke DA. MR imaging features of non-transi-
tional cell carcinoma of the urinary bladder with
pathologic correlation. AJR 2003;180:779–784
26. Tekes A, Kamel IR, Imam K, Chan TY, Schoen-
berg MP, Bluemke DA. MR imaging features of
transitional cell carcinoma of the urinary bladder.
AJR 2003;180:771–777
27. Tekes A, Kamel IR, Szarf G, Chan TY, Schoen-
berg MP, Bluemke DA. Carcinosarcoma of the
urinary bladder: dynamic contrast-enhanced MR
imaging with clinical and pathologic correlation.
AJR 2003;181:139–142
127