NMIBC vs MIBC
High-grade NMIBC is frequently under-staged at TUR [9,10]. In fact, guideline recommendations call for a repeat TUR if a T1 lesion is diagnosed even if muscularis propria is present in the specimen . These procedures require anaesthesia and carry with them a significant cost, and a low but present complication rate . As Kim et al. showed an overall staging accuracy of 23% for CT with tumours ≤pT3a, imaging with CT is unable to accurately distinguish between pTa and pT3a disease [21,22]. Could the improved discrete visualization of MRI have potential benefits in staging, possibly obviating the need for a repeat resection?
T1W and T2W images in multiple planes are required for staging bladder tumours using MRI . On T1W imaging, the bladder tumour has an intermediate SI that is similar to that of muscle, making differentiation of these two structures potentially difficult on this sequence. But, the T1W sequence is helpful in delineating the luminal component of a tumour and additionally, the extent of gross perivesical fat infiltration . More importantly for defining NMIBC, on T2W imaging, the bladder wall is low in SI . Tumours have intermediate to high SI and the muscularis propria has relatively lower SI. Therefore, a normal underlying muscle wall near the tumour suggests an absence of muscle infiltration. With contrast enhancement, the tumour and submucosal SI increase, whereas uninvolved muscle SI is unchanged. On the other hand, an irregular, heterogeneously enhancing muscle layer represents muscular involvement. Applying these MRI criteria, Tekes et al.  (n= 71) studied patients with a known diagnosis of urothelial carcinoma of the bladder using T1W, T2W and DCE-MRI. Pathological confirmation shows that these MRI sequences together can identify a ≤pT1 lesion vs a ≥pT2 lesion with a sensitivity, specificity and accuracy of 95–97%, 55–67% and 85%, respectively. Table 1 summarizes the findings of this study, as well as those of other selected studies that examined MRI as a bladder cancer staging modality.
Table 1. Selected studies of magnetic resonance imaging for bladder cancer T-staging
|Study (n) [Ref.]||MRI technique||Pathological confirmation with RC, n (%)||Overall staging accuracy (%)||Over-staging error rate (%)||NMIBC vs MIBC (%)||Organ-confined vs Non-organ-confined bladder cancer (%)|
|Tekes et al. (71) ||T1W + T2W + DCE-MRI||39 (58)||62||32||85||95–97*||55–67||82||79–86||79–84|
|Hayashi et al. (71) ||T1W + CE-EMRI||19 (27)||83||16||87||91||87||NR|
|Takeuchi et al. (40) ||T2W||17 (43)||67||NR||79||88||74||85||50||95|
|T2W + DCE-MRI||79**||88||94||86||90||80||92|
|T2W + DW-MRI||88**||96a||88||100a||92||70||97|
|T2W + DCE-MRI + DW-MRI||92**||98a||94||100a||94||80||97|
|El-Assmy et al. (106) ||T2W||72 (68)||40||76c||6||NR||NR||15||NR||NR|
|Kobayashi et al.f (117) ||T2W||0 (0)g||79–82||NR||79–83||68–76||80–91||NR|
|Kim et al. (36) ||T1W||22 (61)||75h||25h||NR||78i||78||78|
The differentiation of MIBC from NMIBC can be hindered by non-malignant changes associated with recent TUR, resulting in over-staging of the cancer . Although contrast enhancement can partially ameliorate over-staging, it remains a confounding factor in post-TUR imaging . In the early phases of DCE-MRI sequences, the tumour, mucosa and submucosa enhance and the bladder wall remains hypo-intense . Taking advantage of this observation, investigators have used this submucosal linear enhancement as the basis for staging criteria. Hayashi et al.  studied 71 patients with bladder tumours, using gadolinium-enhanced T1W-MRI with an endorectal coil. Their diagnostic algorithm defined NMIBC if the submucosal linear enhancement was intact and MIBC if the line was interrupted. Patients were pathologically staged with deep TUR or RC. The sensitivity, specificity and accuracy for detecting muscle invasion was 91%, 87% and 87%, respectively. Visualization of submucosal linear enhancement could not always differentiate between inflammatory tissue and tumour extension through the submucosal layer, resulting in over-staging error in 16% of cases. These results were promising but required both gadolinium administration as well as an endorectal coil.
Similarly, using DCE-MRI, Takeuchi et al.  noted that 60% of lesions (n= 40) showed similar enhancement in tumour and submucosal tissue, and submucosal linear enhancement was difficult to identify in these cases. Although this group did not report their over-staging error with DCE-MRI, the authors concluded that contrast-enhanced images may be limited in terms of differentiating pT1 tumours with intact submucosal linear enhancement from pT2 tumours with submucosal linear enhancement disruptions.
In a cohort of putative bladder cancer patients (n= 106) who had not yet undergone TUR, DW-MRI was prospectively compared with T2W-MRI for staging purposes . Standard T2W-MRI proved inaccurate at diagnosing pT1 and pT2 disease, with accuracies of 6% and 24%, respectively. DW-MRI improved this dramatically, to 64% and 75%, respectively, and interestingly, as pathological stage increased, both modalities provided increasingly accurate staging. Of note, DW-MRI had a 24.3% over-staging error for pT2 tumours, but did not under-stage any of these tumours. The overall staging accuracy for T2W in this study was only 39.6%, significantly lower than other published studies, a fact that the authors attributed to a high incidence of perivesical inflammation in the cohort and a high rate of over-staging . In fact, Takeuchi et al.  also found that DW-MRI had superior accuracy to T2W-MRI for differentiating NMIBC from MIBC, 96% and 79%, respectively. However, the differences in accuracy for both T2W and DW-MRI between these two series are striking and according to a recent review from Giannarini et al. , those differences were probably attributable to the superior image quality achieved in the former study through varying diffusion sequences.
Kobayashi et al.  (n= 117) and Takeuchi et al.  used DW-MRI to predict T-stage and ADC to predict tumour grade. Important for interpretation, tumour, submucosal tissue and muscle show high, low and intermediate SI (Fig. 1A), respectively, on DW-MRI . In both studies, ADC values were lower in high-grade tumours and decreased as tumour stage increased. At a threshold ADC value, clinically aggressive phenotypes such as MIBC and high-grade T1 tumours could be differentiated from more indolent phenotypes with sensitivity, specificity and accuracy of 88, 85 and 87%, respectively. The area under the receiver operating curve for predicting clinically aggressive tumours based on ADC value was 0.921 . The mean ADC were (1.29 ± 0.21) × 10–3 mm2/s, (1.13 ± 0.24) × 10–3 mm2/s and (0.81 ± 0.11) × 10–3 mm2/s, for G1, G2 and G3 tumours, respectively . A statistically significant difference was observed between G1 and G3 tumours and between G2 and G3 tumours, but not between G1 and G2 lesions . Interobserver variability calculations favoured the use of DW-MRI over T2W-MRI in both studies with κ= 0.88 vs κ= 0.67, and κ= 0.88 vs κ= 0.70, respectively [16,28].
Figure 1. Stage pT1 papillary urothelial carcinoma in a 70-year-old man. (A) Transverse diffusion-weighted magnetic resonance image shows C-shaped high signal intensity (SI) area with a low SI stalk connecting to left side of bladder wall. (B) An inchworm creeping along a branch: diffusion-weighted magnetic resonance image finding resembles the arch-like shape of an inchworm. Reprinted from Takeuchi M, Sasaki S, Ito M, et al. Radiology 2009; 251: 112–21 with permission of the authors and the publisher, Radiological Society of North America ©2009.
Download figure to PowerPoint
A notable finding on DW-MRI, potentially pathognomonic for pT1 disease, is the ‘inchworm’ sign (Fig. 1A,B). This sign refers to the low SI stalk seen to invaginate into the high SI tumour on DW-MRI. In fact, the positive predictive value of this sign for pT1 disease was as high as 100% (28/28) in the series from Takeuchi et al., although potentially lacking in specificity [16,28]. The two radiologists reviewing images for the purpose of the study under-staged six histologically proven MIBC patients as NMIBC based on this imaging finding, for an under-staging rate of 24–27%. Interestingly, when an ADC threshold value differentiating MIBC from NMIBC was applied to the images with positive ‘inchworm’ sign, the under-staging rates dropped to 4–4.5% . Seemingly, as has been shown in other malignancies, ADC reflects histological tumour grade and stage .
Organ-confined vs non-organ confined
Detecting extravesical extension of bladder cancer is the most important feature of the bladder wall evaluation before RC, as its presence negatively affects prognosis and may indicate the need for multimodal therapy with neo-adjuvant cisplatin-based combination chemotherapy when possible [31,32]. Contrast-enhanced CT is reasonably accurate at this assessment and perivesical fat invasion on multi-detector row helical CT (MDCT) has been shown to have sensitivity and specificity of 89% and 95%, respectively . The updated European Association of Urology (EAU) Guidelines recommend either DCE-MRI or contrast-enhanced CT in patients being evaluated for radical treatment citing the increased sensitivity and decreased specificity for MRI versus CT .
Head-to-head studies comparing CT and MRI for staging of urothelial carcinoma of the bladder are rare. Kim et al. , in 1994, compared CT with several techniques of MRI for patients diagnosed with urothelial carcinoma of the bladder and reported the different modalities' ability to differentiate pTa–pT3a from pT3b–pT4a disease. The sensitivity, specificity and accuracy for CT and DCE-MRI were 93%, 71% and 83%, and 100%, 72% and 86%, respectively. Despite lack of significance, the authors concluded that DCE-MRI was a superior staging tool to CT and T1W or T2W MRI.
More recently, MDCT was used to predict perivesical invasion. Sixty-seven patients with urothelial carcinoma of the bladder were imaged before RC, and 16 bladder cancers with perivesical invasion were identified by pathology . Identification of perivesical invasion with MDCT showed a sensitivity, specificity and accuracy of 89%, 95% and 93%, respectively. In comparing the MRI results reported by Kim et al.  with their results, these authors concluded that MRI provided greater sensitivity through its superior visualization of perivesical fat. However, this came at the cost of reduced specificity because of the misinterpretation of perivesical inflammation as a malignant process [22,33].
In a recent study from El-Assmy et al.  (n= 106) staging accuracy for ≤pT2 disease was 70% and 15%, for DW-MRI and T2W imaging, respectively (P < 0.001). Accuracy and sensitivity for diagnosing non-organ-confined bladder cancer (>pT2) was 92% and 80%, for DW-MRI and T2W imaging, respectively (P > 0.05). Specificity of MRI in diagnosing >T2 disease was not reported, but based on the published data, was limited by over-staging errors for pT1–pT2 lesions. DW-MRI and T2W-MRI over-staged 19/66(29%) and 50/66(76%) of pT1–T2 lesions, respectively.
Tekes et al.  (n= 71) found the sensitivity, specificity and accuracy to be 79–86%, 79–84% and 82% for differentiating ≤pT2b from ≥pT3 or greater disease using T2W and DCE-MRI. As in other studies, employing DCE-MRI resulted in over-staging error, 32% over all cases .
Takeuchi et al.  (n= 40) studied patients with T2W, DW-MRI and DCE-MRI. Diagnostic accuracy for differentiating ≤pT2 from ≥pT3 was tested for all combinations of these MRI sequences. Accuracies ranged from 85 to 94% and in terms of accuracy, no combination of sequences showed diagnostic superiority in differentiating organ-confined from non-organ-confined bladder cancer. However, the addition of DCE-MRI or DW-MRI sequences, or both, to T2W images, improved sensitivity for detecting non-organ-confined disease from 50% to 70–80%, although not reaching significance . Specificity of the MRI finding of non-organ-confined disease was high for all sequences, 92–97%, with no inter-sequence statistical differences.
Lymph node evaluation
Bladder cancer metastatic to regional LN is strongly associated with poor prognosis despite effective local and systemic therapy [4–6]. Although patients with enlarged, biopsy-proven LN metastases are typically given definitive chemotherapy , micro-metastatic LN deposits escape detection. MRI and CT diagnoses of positive LN are based on size criteria where nodes >8 mm or >10 mm are considered positive for round and oval nodes, respectively [26,36]. Size criteria alone lack both sensitivity and specificity, because small metastatic nodes may be missed and enlarged benign nodes may be misclassified . The accuracy of MRI for nodal staging has been reported in the range of 73–90%, which is similar to that reported for CT [11,26]. However, reports such as these should be considered in the context of the great variability in the LN dissections, the details of which are rarely reported in the radiological literature, that serve as the reference standard against which the imaging studies are measured. For example, while Kim et al.  reported on 67 RC patients imaged with MDCT before surgery, only five patients with a total of six LN metastases were found.
There is some evidence that MRI is superior to CT at identifying sub-centimetre pelvic LN but conventional MRI cannot distinguish between small benign nodes and those with micrometastatic deposits [38,39]. In a comparison study of the ability of CT and MRI to detect LN, MRI detected 271 vs 189 detected by CT (P < 0.001), and most of the difference was accounted for by nodes 1–5 mm in size . However, tumour involvement in the identified nodes was not assessed so although more nodes overall were visualized, the study did not tell us whether more malignant nodes were seen. Moreover, in another study comparing positron emission tomography-CT with T2W/DCE-MRI for LN staging before RC, all three small metastatic nodes were missed by MRI and two of three were missed by positron emission tomography-CT .
A recent study investigated the use of DW-MRI for nodal staging in patients scheduled for RC . One of the strengths of this study from Papalia et al.  is that all patients had an extended pelvic LN dissection with a mean nodal yield of 29. Values of ADC were obtainable for nodes in a total of 72 nodal basins. Histopathological evaluation of those nodal basins was then correlated with the recorded ADC values and a receiver operating characteristic curve analysis was performed. From the receiver operating characteristic curve, the authors determined that a threshold ADC of 0.86 × 10–3 mm3/s showed the clearest separation between metastatic and non-metastatic nodes. Using that value, the sensitivity, specificity and positive and negative predicted values were 76.4%, 89.4%, 86.6% and 71.4% . The authors noted the variability of ADC readings from centre to centre and the operator dependency of reading these studies as a limitation of the technique.
Several trials have investigated the utility of MRL for bladder cancer LN staging, using USPIO particles as a contrast material . Deserno et al. (n= 58), in a prospective study of patients scheduled for RC, showed that the use of MRL substantially improved MRI sensitivity and negative predictive value for nodal involvement. Comparing pre-contrast with post-contrast images, sensitivity went from 76% to 96% and negative predictive value from 91% to 98% (P < 0.01). Remarkably, 10/12 metastatic nodes measuring 6–9 mm at pathological review were identified preoperatively. An obvious criticism of the technique is the length of time required for study completion (24–36 h between pre- and post-contrast). Additionally, the images obtained may be difficult to interpret requiring highly specialized expertise . Also, as previously noted, another limitation of this study seems to be the lack of a clearly defined standard template for lymphadenectomy, possibly creating a detection bias.
A recent study from Thoeny et al.  (n= 21), which served as a pilot for a larger clinical trial (NCT00622973) improved upon this technique by incorporating the benefits of DW-MRI with those of USPIO administration. The restricted diffusion capability of water molecules in malignant nodes leads to increased SI on DW-MRI, as does the impaired USPIO uptake in these nodes. The authors hypothesized that the hyperintense signal in malignant nodes resulting from the use of this combined modality would improve the ability to differentiate from benign nodes .
The cohort, who had either a diagnosis of prostate or bladder cancer, was studied with conventional MRI sequences and DW-MRI sequences both pre-contrast and post-contrast infusion with Ferumoxtran-10. Images were interpreted in two ways: using the technique of pre- and post-contrast comparison, as in Deserno et al. ; and using the technique of the USPIO-enhanced MRI and DW-MRI (USPIO-DW-MRI) with subsequent morphological correlation. All patients underwent an extended pelvic LN dissection identifying a total of 26 malignant nodes. Overall, 24/26 were correctly identified with USPIO-DW-MRI and although diagnostic accuracies were of similarly high quality in the first method, and both methods require a second set of MRI images 24–36 h after Ferumoxtran is infused, the time spent interpreting the images was significantly less using the new method (13 min/patient vs 80 min; P < 0.001) . Additionally, the USPIO-DW-MRI images seem relatively easy to interpret in that the radiologists who served as readers in this study were not previously familiar with the technique . Taken together, these studies show that USPIO can significantly improve the accuracy of detecting LN metastasis in pelvic malignancies. However, image acquisition time is long, but the addition of DW-MRI techniques may increase the ease and the speed with which these images are interpreted. Despite some encouraging findings, further trials are necessary before FDA approval and widespread inclusion into clinical practice is considered.