The long-term outcome of treated high-risk nonmuscle-invasive bladder cancer

Time to change treatment paradigm?




The treatment of high-risk nonmuscle-invasive bladder cancer (NMIBC) is difficult given its unpredictable natural history and patient comorbidities. Because current case series are mostly limited in size, the authors report the outcomes from a large, single-center series.


The authors reviewed all patients with primary, high-risk NMIBC at their institution from 1994 to 2010. Outcomes were matched with clinicopathologic data. Patients who had muscle invasion within 6 months or had insufficient follow-up (<6 months) were excluded. Correlations were analyzed using multivariable Cox regression and log-rank analysis (2-sided; P < .05).


In total, 712 patients (median age, 73.7 years) were included. Progression to muscle invasion occurred in 110 patients (15.8%; 95% confidence interval [CI], 13%-18.3%) at a median of 17.2 months (interquartile range, 8.9-35.8 months), including 26.5% (95% CI, 22.2%-31.3%) of the 366 patients who had >5 years follow-up. Progression was associated with age (hazard ratio [HR], 1.04; P = .007), dysplastic urothelium (HR, 1.6; P = .003), urothelial cell carcinoma variants (HR, 3.2; P = .001), and recurrence (HR, 18.3; P < .001). Disease-specific mortality occurred in 134 patients (18.8%; 95% CI, 16.1%-21.9%) at a median of 28 months (interquartile range, 15-45 months), including 28.7% (95% CI, 24.5%-33.3%) of those who had 5 years of follow-up. Disease-specific mortality was associated with age (HR, 1.1; P < .001), stage (HR, 1.7; P = .003), dysplasia (HR, 1.3; P = .05), and progression (HR, 5.2; P < .001). Neither progression nor disease-specific mortality were associated with the receipt of bacillus Calmette-Guerin (P > .6).


Within a program of conservative treatment, progression of high-risk NMIBC was associated with a poor prognosis. Surveillance and bacillus Calmette-Guerin were ineffective in altering the natural history of this disease. The authors concluded that the time has come to rethink the paradigm of management of this disease. Cancer 2012. © 2012 American Cancer Society.


Bladder cancer is the fourth commonest male malignancy worldwide, with 297,300 new cases and 112,300 deaths in 2008.1 The majority of tumors are urothelial cell carcinoma (UCC) in histologic subtype. Clinical and molecular data suggest 2 discrete disease phenotypes of UCC characterized by low-grade and high-grade differentiation.2, 3 Low-grade UCCs are managed by endoscopic resection, intravesical chemotherapy, and life-long surveillance. High-grade UCCs account for approximately 33% of cases at first diagnosis and can present with or without muscle invasion. Nonmuscle-invasive bladder cancer (NMIBC), including carcinoma in situ (CIS), represents half of these high-grade cases. A variable proportion (range, 25%-75%) of these tumors will progress to muscle invasion within 5 years of diagnosis.4 This event marks a dramatic increase in the risk of metastasis and disease-specific mortality (DSM).5 Consequently, the clinical care of patients with high-grade nonmuscle-invasive UCC usually is directed at preventing or monitoring progression to muscle invasion. There are few definitive data on the progression rate and DSM in patients with conservatively managed high-grade nonmuscle-invasive UCC. Most reports are limited in size, contain heterogeneous populations, or have short follow-up (for a review, see Kulkarni et al6). The objective of the current study was to examine the outcome of such tumors in a large series managed at a single institution with a view to expanding the literature base and reveal trends in outcome.


Bladder Cancer Patients and Tumors

The Royal Hallamshire Hospital is the only urologic service for the city of Sheffield, United Kingdom (population, approximately 800,000). Thus, patients with urologic diseases within this community attend and remain under the care of this department for the duration of their treatment and surveillance.

To determine the outcomes of patients with UCC, we retrieved all histology reports that included bladder tissue from the Department of Pathology between January 1, 1994 to December 31, 2009. We identified patients with UCC and matched those patients with their hospital billing (date and type of visit), pathologic, pharmacy (intravesical chemotherapy), and death certification records (from the Trent Cancer Registry). In addition, the medical records from 2% of all patients (chosen at random) and those from patients who had missing data or conflicting identity codes were reviewed. To ensure integrity, 2 workers separately merged data streams to create parallel databases (F.T. and N.R.), which were then compared, and any differences were resolved.

Selection Criteria

Histologic grading and staging of each UCC was performed by dedicated uropathologists using the 1973 World Health Organization and TNM classifications. All patients whose primary tumor was nonmuscle-invasive and high-risk, defined as either poorly differentiated (grade 3), lamina propria invasion (T1), or the presence of CIS (alone or with a bladder tumor), were included. Patients who had muscle-invasive disease at reresection or within 6 months of diagnosis and those who had <6 months of follow-up were excluded. Patients who were referred for a second opinion or from treatment from out of the region also were excluded.


Three months after patients underwent initial resection, surveillance was started with cystoscopy according to evidence-based regimes7 and regional guidelines (available at: [accessed October 10, 2010]). In general, patients with high-risk tumors underwent rigid cystoscopy for their first check and during bacillus Calmette-Guerin (BCG) therapy. During the evaluation period, the benefits from early reresection for these patients became apparent. Consequently, reresection was introduced as standard practice in the later years of recruitment. Patients were offered induction or maintenance BCG as first-line treatment. Patients underwent radical cystectomy (RC) for muscle-invasive disease, for BCG-refractory cancers at the surgeon's discretion, for severe urinary symptoms, or as primary treatment in selected patients.

Statistical Analysis

The primary outcomes were progression-free, disease-specific, and overall survival. Secondary outcomes were RC and recurrence-free survival. Tumor recurrence was defined as a second bladder cancer with similar or better stage or grade compared with the original. Progression was defined as a new tumor with pathologically proven muscle invasion or radiologically proven metastases. DSM was identified from death certification and was defined when bladder cancer was the main attributable cause of death (parts 1a or 1b of the United Kingdom death certificate) or was suggested from case notes (eg, pathologic or radiologic evidence of metastases/tumor progression, palliative treatment before death). Nonbladder cancer-specific mortality (non-DSM) was defined as death certification that attributed another cause of death, and this was supported by pathologic or case note review. Patients with an equivocal cause of death without postmortem proof (eg, renal failure with a prior history of UCC) were excluded from analysis.

Associations between clinicopathologic features and tumor behavior were examined using univariable and stepwise multivariable Cox proportional hazards regression models with the SPSS statistical software package (version 14.0; SPSS Inc., Chicago Ill). Outcomes with respect to time were plotted using the Kaplan-Meier and competing event methods, and compared using the log-rank test. Competing mortality was plotted using a smoothed nonlinear fit to the events in GraphPad Prism (version 5.0, GraphPad Software, San Diego, Calif). All statistical tests were 2-tailed, and significance defined as P < .05.


Patients and Tumors

In total, 951 patients whose primary tumors fulfilled the inclusion criteria were identified. We excluded 72 patients who developed invasion within 6 months of diagnosis (including at reresection), 104 patients with <6 months follow-up, 54 patients who were referred from other hospitals, and 9 patients with non-UCC histology (Fig. 1). This left 712 patients (median age, 73.7 years) with a median follow-up of 62 months (interquartile range, 27.6-119.3 months). The patients were typical for UCC with men in the majority (78%). The cohort characteristics are summarized in Table 1. Tumors were poorly differentiated (including 305 staged as pathologic T1 [pT1] tumors) except for 19 grade 1 tumors (5 with CIS and 14 pT1 tumors) and 181 grade 2 tumors (28 with CIS and 159 pT1 tumors). One hundred seventy-three patients (24.3%) underwent reresection within 3 months, and 366 patients (51.4%) underwent reresection within 6 months. The initial endoscopic resection and accurate staging of a cancer is critical to appropriate treatment.8 In this cohort, muscle was present in 482 patients (67.7%), and flat urothelium was present in 392 patients (55.1%) at first resection. Although the incidence of RC increased from 8% (no muscle or flat urothelium) to 12.4% (either tissue) and 14.4% (both tissues) with complete transurethral resection staging, neither progression nor DSM varied with this measure (P > .05; log-rank test).

Figure 1.

This is a flow chart of the patients and events within the current study. UCC indicates urothelial cell carcinoma; pT, pathologic tumor classification.

Table 1. Clinicopathologic Features of the Patients and Tumors Reported in This Series of High-Risk Nonmuscle-Invasive Bladder Cancers
FeatureNo. of Patients (%)
  1. Abbreviations: Cis, carcinoma in situ; IQ, interquartile; TUR, transurethral resection; UCC, urothelial cell carcinoma.

Total712 (100)
Age at diagnosis: Median [range], y73.7 [41-97]
Follow-up: Median [IQ range], mo62 [27.6-119.3]
 Men557 (78.2)
 Women155 (21.8)
Tumor differentiation 
 Grade119 (2.7)
 Grade 2181 (25.4)
 Grade 3512 (71.9)
Pathologic tumor classification 
 pTis68 (9.6)
 pTa160 (22.5)
 pT1481 (67.6)
 pTx3 (0.4)
Growth pattern 
 Papillary387 (54.4)
 Both201 (28.2)
 Solid37 (5.2)
 Missing87 (12.2)
UCC variant histology 
 Sarcomatoid6 (0.8)
 Neuroendocrine2 (0.3)
 Plasmacytoid3 (0.4)
 Small cell2 (0.3)
 Micropapillary7 (1)
 Glandular component28 (3.9)
Vascular invasion 
 Present31 (4.4)
Background urothelium 
 Absent320 (44.9)
 Cis223 (31.3)
 Low grade dysplasia14 (2)
 Normal155 (21.8)
Muscle in TUR sample, % 
 Absent150 (21.1)
 Present482 (67.7)
 Not known80 (11.2)
 Reresection by 3 mo173 (24.3)
 Reresection by 6 mo366 (51.4)
 No reresection173 (24.3)

Bacillus Calmette-Guerin Administration

The case notes for all patients diagnosed since 2000 were reviewed specifically for BCG treatment. Immunotherapy was received by 249 of 398 patients (63%). Receipt of BCG was more common among younger patients (mean age BCG was received, 72.2 years vs 74.2 years; P = .05; t test), higher grade tumors (43% of grade 1 tumors vs 68% of grade 3 tumors; P < .001; chi-square test), lower stage cancers (71% pTa tumors vs 63% pT1 tumors; P = .07; chi-square test), and tumors with dysplastic urothelium (57% with normal urothelium vs 73% with CIS; P = .05; chi-square test).

Recurrence and Progression

A recurrence developed in 402 patients (56.5%; 95% confidence interval [CI], 52.8%-60.1%) at a median of 4.8 months (interquartile range, 2.8-7.6 months) (Fig. 2a). In multivariable analysis, recurrence was associated with tumor size (using specimen weight as a surrogate marker; Cox hazard ratio [HR], 1.02; 95% CI, 1.0-1.03; P < .03) and receipt of BCG administration (HR, 0.62; 95% CI, 0.39-0.99; P = .04) but not with patient age, tumor grade, stage, growth pattern, or complete transurethral resection staging.

Figure 2.

The natural history of high-risk nonmuscle-invasive bladder cancer is illustrated. (a-d) The risk of 4 clinical outcomes were calculated using Kaplan-Meier survival analysis in 712 patients with high-risk nonmuscle-invasive bladder cancer: (a) recurrence, (b) progression-free survival, (c) disease-specific survival, and (d) overall survival.

Progression to muscle-invasive disease occurred in 110 patients (15.8%; 95% CI, 13%-18.3%) at a median of 17.2 months (interquartile range, 8.9-35.8 months) (Fig. 2b). Progression occurred in 83 of 462 patients (18%; 95% CI, 14.7%-21.7%) and 97 of 366 patients (26.5%; 95% CI, 22.2%-31.3%) with a minimum 3 years and 5 years of follow-up, respectively. In univariable analysis, progression to muscle invasion was associated with advanced patient age, tumor grade, growth pattern, the status of the background urothelium, previous recurrent tumors, and histologic variants of UCC (Cox model; P < .05) (Table 2, Fig. 3a,b). In multivariable analysis, progression was associated with age (HR, 1.04; 95% CI, 1.0-1.1; P = .007), dysplasia in the flat urothelium (HR, 1.6; 95% CI, 1.2-2.17; P = .003), variants of UCC (HR, 3.2; 95% CI, 1.6-6.3; P = .001), and previous recurrence (HR, 18.3; 95% CI, 4.5-75.6; P < .001) (Table 2) but with BCG (Fig. 3b). The combination of stage and grade improved progression stratification. For example, 62 of 305 patients (20.3%; 95% CI, 16.7%-25.8%) who had grade 3 pT1 tumors progressed to muscle-invasive disease compared with 12 of 66 patients (18.2%; 95% CI, 11.1%-30%) who had CIS and 13 of 137 patients (9.5%; 95% CI, 6.1%-16.3%) who had grade 3 pTa tumors.

Table 2. Associations Between Clinicopathologic Features and Progression to Muscle Invasion
     Univariable AnalysisMultivariable Analysis 
 No. ProgressionProgression95% CI95% CI 
  • Abbreviations: BCG, bacillus Calmette-Guerin; CI, confidence interval; Cis, carcinoma in situ; HR, hazard ratio; TUR, transurethral resection; UCC, urothelial cell carcinoma.

  • a

    Only patients from 2000 to 2009 were analyzed (n = 369).

Specimen weight, g            
Age, y            
Pathologic tumor classification            
Growth pattern            
Vascular invasion            
Background urothelium            
Muscle in TUR samples            
Reresection within 3 mo            
Reresection within 6 mo            
Figure 3.

These charts illustrate (a,b) progression to muscle invasion and (c,d) disease-specific mortality (DSM) in 712 patients with high-risk nonmuscle-invasive bladder cancer according to the background urothelium, the receipt of intravesical bacillus-Calmette-Guerin (BCG), progression to invasion after initial resection, and primary tumor classification plotted in Kaplan-Meier survival analyses. Comparisons were made using the log-rank test (P values are provided in the lower left corner of each chart).

Radical Cystectomy and Radiotherapy

Ninety-one patients (12.8%) underwent RC for disease progression at a median of 14.2 months (n = 30; 33%), at a median of 10.8 months for BCG-refractory NMI disease (n = 54; 59%), and as primary treatment at a median of 2.1 months in selected patients (n = 7; 8%). Of those with BCG-refractory NMI disease, RC histology revealed upstaging to muscle-invasive tumors in 12 of 54 patients (22%). RC was more common in younger patients (mean age at cystectomy, 65.6 years vs 73.9 years; P < .001; t test) and those with dysplastic urothelium (11.1% of those with normal urothelium vs 20% with CIS; P = .04; chi-square test). DSM occurred in 21 of 91 patients (23.1%) who underwent RC and in 3 of 7 patients (42.9%) who received radical radiotherapy. In the RC cohort, 55 patients (60.4%) remained alive without evidence of disease. The remaining 15 patients (16.4%) died of other causes.

Disease-Specific Mortality

In total, 134 patients (18.8%; 95% CI, 16.1%-21.9%) died of bladder cancer at a median of 28 months (interquartile range, 15-45 months) (Fig. 2c). DSM occurred in 85 of 495 patients (17.2%; 95% CI, 14.1%-20.7%) and in 115 of 401 patients (28.7%; 95% CI, 24.5%-33.3%) at 3 years and 5 years postdiagnosis, respectively. In multivariable analysis, DSM was associated with advanced age (Cox HR, 1.1; 95% CI, 1.0-1.1; P < .001), progression to muscle invasion (HR, 5.2; 95% CI, 3.1-8.8; P < .001) (Table 3, Fig. 3c), disease stage (HR, 1.7; 95% CI, 1.2-2.5; P = .003) (Fig. 3d), and dysplastic urothelium (HR, 1.3; 95% CI, 1.0-1.7; P = .05).

Table 3. Associations Between Clinicopathologic Features and Disease-Specific Mortality
     Univariable AnalysisMultivariable Analysis
 No DSMDSM 95% CI  95% CI 
  • Abbreviations: BCG, bacillus Calmette-Guerin; Cis, carcinoma in situ; DSM, disease-specific mortality; TUR, transurethral resection;

  • UCC, urothelial cell carcinoma.

  • a

    Only patients from 2000 to 2009 were analyzed (n = 369).

Specimen weight, g            
Age, y            
Pathologic tumor classification
Growth pattern            
Vascular invasion            
Background urothelium            
Muscle in TUR samples            
Reresection within 3 mo            

Overall and Competing Mortality

In total, 376 patients (52.8%; 95% CI, 49.1%-56.5%) died in this series, including 134 deaths from UCC and 49 deaths (6.9%) from other tumors. The median time to non-DSM was 45.8 months (Fig. 2d). Competing mortality plots reveal the risk of DSM and non-DSM for the 10 years after diagnosis (Fig. 4). The risk of non-DSM was associated significantly with age (mean age, 75.7 years for non-DSM vs 71.3 years for DSM; P < .0001; t test) but not with UCC-related factors. Younger patients with pT1 disease had the lowest risk of competing mortality (37.7% for non-DSM vs 23.3% for DSM; ratio, 1.62), whereas older patients with less advanced disease had the highest risk (22.3% vs 40.1%, respectively; ratio, 1.83). Non-DSM was lower in patients who underwent cystectomy than in the entire population (15.7% after cystectomy vs 36.3%; P < .001; chi-square test), supporting anesthetic selection criteria.

Figure 4.

Competing mortality is illustrated among patients with high-risk nonmuscle-invasive bladder cancer. The risk of disease-specific or nonurothelial cell carcinoma-related death is plotted for the 10 years after initial diagnosis using stage and age stratification. pT indicates pathologic tumor classification; UCC, urothelial cell carcinoma.


Bladder cancer is one of the most expensive human malignancies to manage. Estimates indicate that, in the United States alone, it costs up to $200,000 per patient lifetime (in 2001) to treat the disease9; and, in 2002 the National Health Service spent £55.4 M on bladder cancer.10 The majority of this cost is spent on surveillance of NMI tumors in an attempt to identify and treat worsening disease before the onset of metastasis.10 The frequency of surveillance and the intensity of subsequent treatments is greatest in patients with high-risk disease, like those reported in this series. Thus, current treatment regimens should detect worsening disease in time to treat affected patients. The data presented within our report question the efficacy of these regimens, support previous case series (for review, see Kulkarni et al6), and partially explain why there is little relation between surveillance intensity and disease-specific survival.11

Our data represent the largest series of high-risk cases reported to date, and the findings are consistent with previous reports (for review, see Kulkarni et al6). Despite disparities in the initial treatment (eg, induction vs maintenance BCG), approximately 1 in 4 patients with high-risk NMI tumors progress to muscle invasion within 5 years of diagnosis. This event markedly reduces disease-specific survival, such that many of these patients die from their tumors (approximately50%). Although it is appealing to consider a regimen of initial bladder sparing and radical treatment for patients with disease progression, these outcomes do not support the universal safety of this approach. For example, in our cohort, DSM was related most directly to the presence of disease progression. This single event increased DSM greater than 5-fold compared with patients who did not have disease progression. Consequently, the long-term cure rate from radical treatment dropped from 90% or 91% down to 49% with muscle invasion. These outcomes match those reported by other groups for BCG failure in patients who underwent RC and appear no better or worse than cure rates for patients who had primary muscle-invasive cancer who underwent RC.12, 13 For example, Schrier et al observed that the cancer-specific survival rate after RC was better for patients with de novo muscle-invasive cancer (67% at 3 years) than for patients who failed on BCG (37% at 3 years).12, 13 The reasons for the poor outcomes in patients with this disease are multiple and include key factors like the inability of clinicians to accurately stage the disease (eg, in our series, 22% of presumed NMI tumors had invasion), the lack of pathologic reproducibility when reporting the grade and stage of this disease,14 and patient/physician compliance with surveillance and treatment regimens.

The risk of progression to invasion and DSM from high-risk NMI UCC has been known for many years.15 Given the age of typical patients with UCC, their comorbidities, and the perceived disadvantages of radical treatment, bladder-sparing approaches have been developed. Of these, maintenance BCG became the first-line treatment for high-risk NMI UCC when randomized trials and meta-analyses revealed its use reduced progression by 5% at 2.5 years (eg, Sylvester et al16 reported a relative reduction of 27%). However, the composition of these meta-analyses and recent data now question the surety of this finding. First, only 8% patients within their analysis had high-risk NMI UCC.16 The majority of their patients had intermediate-risk disease with a recurrent (not progressive) tumor phenotype. Second, their median follow-up was short (2.5 years). Our data illustrated in Figure 2 indicate that progression and mortality often occur beyond this time point (which actually reflects only the median mortality time in our series). Consequently, many BCG failures could be missed by this short follow-up. Finally, a more recent individual case-note meta-analysis failed to support an effect of BCG on disease progression.17 That later analysis had a larger proportion of high-risk patients (23%) and a longer follow-up (median, 4.4 years).

Although, previously, it was reasonable to use BCG for its reported reduction in progression, the other perceived benefit of a bladder-sparing approach is that of presumed improved quality of life. However, there are few data to support this presumption.18 Maintenance BCG is a laborious regimen that involves many intravesical instillations and cystoscopies. Most patients (74%) report toxicity, including local symptoms (cystitis, 25%-43%; bladder pain and frequency, 31%; hematuria, 35%) and systemic symptoms (fever >39°C, 15%; malaise, 23%; skin rash, 3%).19, 20 These can be precursors of serious toxic events (BCG sepsis or visceral infection). Only 30% of patients complete maintenance BCG.19, 20

Alternate treatment approaches for high-risk UCC include newer intravesical agents/delivery mechanisms (for review, see Lammers et al21) or lowering the threshold for radical treatment. Contemporary series reveal that modern RC has low complication (20%) and mortality (<2%) rates, reflecting changes in surgical and anesthetic practice.22, 23 Although most morbidity is transient (including wound problems, paralytic ileus, and infection), the effects of cystectomy on a patient's lifestyle must be recognized. It is noteworthy that there are few reliable data describing quality of life after RC. One prospective study using a validated questionnaire indicated that recovery takes 3 to 6 months, and most parameters returned to preoperative levels by 1 year.24 Cancer outcomes from primary RC for high-risk NMI UCC reveal promising disease-specific survival rates (eg, 79% in 10 years25). Despite the lack of evidence from randomized controlled trials, these factors have encouraged a change of practice toward early RC for patients with high-risk UCC.26

Although, to our knowledge, the current series represents the largest reported to date, it is retrospective and contains limitations. Perhaps two of the greatest concerns are regarding reresection and the use of BCG. For the former, only 54% of our population underwent reresection. To counter any potential bias from this (ie, the inclusion under-staged T2 disease), we excluded all patients who developed muscle invasion within 6 months of diagnosis. Consequently, the median time to progression was approximately around 1.5 years after diagnosis, which was in keeping with the progression times and rates reported in others series (for review, see Kulkarni et al6). With regard to BCG, the use of this agent was poorly documented before 2000. To compensate for this, we excluded all patients before the year 2000 from any analyses that included BCG (the BCG data set included nearly 400 patients). Although approximately 66% of patients received BCG, many received only an induction regimen or failed to complete the full maintenance course. Perhaps these limitations suggest that our BCG findings should be considered with caution, but it is important to recognize that they support the outcomes reported in randomized controlled trial meta-analyses.17

In conclusion, our data reveal that approximately 1 in 4 patients with high-risk NMI UCC progress to muscle invasion when they are followed for a minimum of 5 years. This event dramatically worsens the outcome from this disease and is difficult to identify beforehand. Although surveillance is tailored to the risk of progression, our data suggest that it often fails to alter the natural history of progressive tumors; therefore, we believe it is time to review the treatment paradigm for this disease. Specifically, the receipt of BCG did not reduce progression. In patients who have an anticipated life expectancy ≥5 years, RC is likely to be more effective and more cost effective at preventing disease progression and DSM than bladder-sparing treatment.27


We thank Drs. Anderson, Chapple, Hastie, Hall, Haynes, Inman, Oakley, Reid, and Smith for allowing us to study their patients and the staff and patients of the Royal Hallamshire Hospital.


J.W.F.C. was supported by a GlaxoSmithKline Clinician Scientist fellowship and by project grants from Yorkshire Cancer Research, Sheffield Hospitals Charitable Trust, Astellas Educational Foundation, and the European Union (the European Community's Seventh Framework Program; grants: FP7/2007-2013 and HEALTH-F2-2007-201438).


The authors made no disclosures.