To assess the prognostic factors that could be used to predict tumour recurrence and progression, and to construct and validate a predictive index.
To assess the prognostic factors that could be used to predict tumour recurrence and progression, and to construct and validate a predictive index.
Between June 1991 and December 2000, 533 patients (418 men and 115 women; mean age 55.4 years) underwent complete transurethral resection of histologically confirmed pTa and pT1 transitional cell carcinoma of the bladder, after which 377 (test series) were randomized into two subsequent studies, of six groups, to receive adjuvant intravesical sequential bacillus Calmette-Guérin (BCG) and epirubicin, BCG alone, epirubicin (50 or 80 mg), adriamycin 50 mg or no adjuvant therapy. Factors potentially affecting tumour recurrence or progression were assessed using univariate and multivariate analysis, i.e. tumour stage, histological grade, DNA ploidy, history of recurrence, multiplicity, size, tumour configuration, associated carcinoma in situ, recurrence at the first 3-month check cystoscopy and the use of adjuvant therapy. The regression coefficients determined by Cox regression analysis were used to construct a predictive index (PI). The algebraic sum of the regression coefficients of the factors with independent and significant association with disease-free survival for each case represented a proportional hazard score (PHS). The PI was validated in another series of 156 patients (validation series) in whom the same regression coefficients for the same significant factors as the test series were used to categorize it into three risk groups. Kaplan-Meier survival curves were plotted to compare the different risk categories in both test and validation series.
The mean (sd, range) follow-up in the test and validation series were 58 (19, 5–96) and 28.3 (14.9, 2–94) months, respectively. In the test series, tumour stage, DNA ploidy, multiplicity, history of recurrence, tumour configuration, cystoscopy result and the type of adjuvant therapy had independent significance for recurrence on multivariate analysis. For progression, the cystoscopy result, DNA ploidy and grade were the only independent and significant predictors. The ranges of PHS for the factors affecting recurrence-free and progression-free survival were 0.0–7.14 and 0.0–5.84, respectively, which were divided equally into three risk categories with significant differences on Kaplan-Meier curves and a log-rank test (P < 0.001). The three categories in the validation series were significantly different from each other and each was comparable with that in the test series.
Tumour stage, DNA ploidy, multiplicity, history of recurrence, tumour configuration and type of adjuvant therapy affected independently the rate of recurrence after resecting superficial bladder tumour. Recurrence at the 3-month cystoscopy, histological grade and DNA ploidy were the only predictors of progression to muscle-invasion. The PI dividing the patients into three risk groups with different treatment and follow-up strategies for recurrence and progression was reproducible in a validation series.
Superficial bladder tumours are characterized by recurrence in 50–70% of cases [1,2]; when adjuvant intravesical therapy was given after transurethral resection of bladder tumour (TURBT), the incidence of recurrence was reduced . Although most recurrences are still superficial, progression to muscle-invasive disease occurs in 10–30% of patients [2,3]. Therefore, when superficial bladder tumour is diagnosed it is important to identify patients who are at risk of disease recurrence and progression. If it were possible to clearly define which subset of superficial bladder tumours is likely to progress pre-emptive therapy could be given, possibly cystectomy . Alternatively, an intense regimen of adjuvant intravesical therapy could be given, after TURBT, to patients whose tumours are likely to recur. Similarly, if tumour subsets which are unlikely to recur or progress could be identified then a less aggressive therapy could be advocated. Thus life-threatening progression with insufficient treatment, or complications arising from unnecessarily intense therapy, could both be avoided.
Identifying the prognostic factors that determine the risk in each patient for recurrence and progression remains a subject of extensive research [2–7]. Most of the studies assessing these prognostic factors are criticised for disparities in the number and characteristics of patients and their tumours. In addition, some of the studies included patients treated with TURBT alone and others with TURBT and adjuvant therapy; some analysed these factors only in a univariate model.
Therefore, the need for a properly conducted multivariate analysis of the prognostic factors for recurrence and progression in superficial bladder tumour prompted the present single-centre prospective study. From this multivariate analysis and the objectively calculated risk for patients in three categories, a predictive index (PI) was defined in a test series and validated in another series of patients.
Between June 1991 and December 2000 the study included 533 patients with histologically confirmed superficial bladder tumours (stages pTa and pT1 TCC), comprising 418 men and 115 women (mean age 55.4 years, sd 9.4, range 37–79), who were evaluated and followed regularly until July 2002. Patients who had received previous pelvic radiotherapy or systemic chemotherapy were excluded. All patients underwent TURBT. In the first 3 years of the study patients were randomly allocated after resection into one of four groups, receiving adjuvant intravesical chemotherapy as epirubicin 50 (group 3), or 80 mg (group 4), adriamycin (doxorubicin) 50 mg (group 5) and no adjuvant therapy (control/group 6). In the next 3 years, subsequent patients were assigned to receive either sequential BCG + epirubicin, alternating with each other (group 1) or BCG alone (group 2). The doses and schedules of therapy were the subject of two previous reports [3,6]. There were 64, 68, 60, 61, 66 and 58 patients in groups 1–6, respectively; these 377 patients were designated the test series and prognostic factors affecting recurrence and progression in this group were analysed, with the construction of the PI. To validate the PI this predictive model was applied to a series of 156 patients (the validation series) who had received adjuvant intravesical BCG + epirubicin or BCG alone after resection; Table 1 shows the patient and tumour characteristics of the whole series.
|Variable||Series, N (%)|
|pTa||38 (10.1)||12 (7.7)|
|pT1||339 (89.9)||144 (92.3)|
|Yes||20 (5.3)||9 (5.8)|
|No||357 (94.7)||147 (94.2)|
|G1||54 (14.3)||26 (16.7)|
|G2||241 (63.9)||101 (64.7)|
|G3||82 (21.8)||29 (18.6)|
|Diploid/tetraploid||313 (83)||123 (79)|
|Aneuploid||64 (17)||21 (21)|
|Single||134 (35.5)||60 (38.5)|
|Multiple||243 (64.5)||96 (61.5)|
|Tumour size, cm|
|< 3||247 (65.5)||61 (39)|
|≥ 3||130 (34.5)||95 (61)|
|Papillary||346 (91.8)||138 (88.5)|
|Solid||31 (8.2)||18 (11.5)|
|Primary||222 (58.9)||125 (80)|
|Recurrent||155 (41.1)||31 (20)|
|BCG + epirubicin||66 (17.5)||83 (53)|
|BCG alone||58 (15.4)||73 (47)|
|Epirubicin 50 mg||64 (17)||–|
|80 mg||68 (18)||–|
|Doxorubicin 50 mg||60 (15.9)||–|
|TURBT alone||61 (16.2)||–|
|Recurrence at 3 months|
|No||347 (92)||145 (93)|
|Yes||30 (8)||11 (7)|
All patients were evaluated on entry and at each follow-up visit; the initial evaluation included urine analysis, urine culture, serum creatinine level, fasting blood glucose level, a complete blood count, urinary tract plain X-ray, chest X-ray, IVU and a bladder wash for cytology and DNA studies. The evaluation at each visit included a clinical examination to detect disease recurrence and/or progression, urine analysis, urine culture, serum creatinine level, a complete blood count and electrocardiography. Cysto-urethroscopy, urine cytology and flow cytometry were performed every 3 months during the first 2 years and every 6 months thereafter, and IVU annually. Abdominal ultrasonography and liver function tests were carried out when necessary.
Two events were considered for evaluation during the follow-up period, i.e. recurrence of superficial disease and progression to muscle-invasive disease. The recurrence-free survival and progression-free survival were independently correlated with several clinical and pathological variables, including pathological stage, histological grade, tumour ploidy, associated carcinoma in situ (CIS), tumour multiplicity (single vs multiple), tumour size (< 3 vs ≥ 3 cm), tumour configuration (papillary vs solid), history of recurrence (primary vs recurrent) and type of adjuvant therapy. When these prognostic variables were correlated with recurrence, those who developed muscle-invasive progression were excluded from the evaluation. The period of recurrence or progression-free survival was defined as the time between the last TURBT before inclusion into the study and the development of recurrence or muscle-invasive progression.
For the univariate analysis, Kaplan-Meier survival curves  were constructed to estimate recurrence-free and progression-free survival in relation to different clinical and pathological variables, and the statistical differences computed using the log-rank test , with P < 0.05 considered to indicate significance.
For the multivariate analysis, those factors with significant associations with recurrence-free or progression-free survival on univariate analysis were assessed, using a Cox stepwise proportional-hazard regression analysis  to identify variables which had an independent and significant association with recurrence-free or progression-free survival.
The regression coefficients determined by multivariate analysis were used to construct the PI. The algebraic sum of the regression coefficients of the factors with independent and significant association with disease-free survival for each case represented a proportional hazard score (PHS), calculated as RC1 + RC2 + RC3 + RC4 + . . . where RC1 is the regression coefficient of the first significant factor, and so on. The possible range of the PHS of the patients was divided equally into three risk groups and Kaplan-Meier curves then constructed for each group.
The values of the regression coefficients of the significant and independent factors, in relation to recurrence and progression, obtained from the test series were applied as such to the same variables in the validation series. The PHS was then calculated for each case, from the formula. The range of PHS of the test series was again applied to the validation series, to divide the latter into three risk categories of similar PHS range to those of the test series. Kaplan-Meier curves were then reconstructed for these risk groups of the validation series in relation to recurrence and progression, and the results compared with those of the test series.
The mean (sd, range) follow-up of the test and validation series were 58 (19, 5–96) and 28.3 (14.9, 2–94) months, respectively. All groups were comparable in patient and tumour characteristics (Fisher's exact test).
In the test series, factors significantly affecting recurrence-free survival on univariate analysis were tumour stage (pTa vs pT1; P < 0.001), DNA ploidy (P < 0.001), multiplicity of the tumour (P = 0.006), history of recurrence (primary vs recurrent tumour; P = 0.003), recurrence at the first 3-month check cystoscopy (recurrence vs no recurrence; P = 0.002), tumour configuration (papillary vs solid; P < 0.001) and adjuvant intravesical therapy (P < 0.001). Histological grade, associated CIS and tumour size were not significant factors.
Table 2 shows the Cox proportional-hazard regression analysis of factors affecting recurrence-free survival in 334 patients of the test series. Factors maintaining independent significance were tumour stage, DNA ploidy, tumour multiplicity, history of recurrence, tumour configuration, recurrence and type of adjuvant therapy.
|N||Regression (sem)||RR (95% CI)||P||N||Regression (sem)||RR (95% CI)||P|
|pT1||303||2.67 (1.01)||14.39 (1.99–103.7)|
|G2||–||–||–||241||1.41 (1.03)||4.09 (0.55–30.7)|
|G3||–||–||–||82||2.08 (1.04)||7.99 (1.04–61.2)|
|Aneuploid||51||0.71 (0.24)||2.04 (1.29–3.23)||64||0.72 (0.34)||2.06 (1.06–4.0)|
|Multiple||215||0.77 (0.21)||2.15 (1.42–3.26)|
|Tumour configuration||< 0.001|
|Solid||25||1.35 (0.29)||3.86 (2.21–6.75)|
|Recurrent||136||0.42 (0.19)||1.52 (1.05–2.20)|
|Therapeutic arm||< 0.001|
|BCG + epirubicin||62||–||1.00|
|BCG alone||48||0.76 (0.48)||2.13 (0.84–5.42)||0.112|
|Epirubicin 50 mg||56||1.28 (0.44)||3.58 (1.51–8.51)||0.004|
|80 mg||62||1.08 (0.45)||2.94 (1.21–7.14)||0.018|
|Doxorubicin 50 mg||51||1.56 (0.43)||4.75 (2.04–11.1)||< 0.001|
|TURBT alone||55||2.78 (0.42)||16.1 (7.13–36.3)||< 0.001|
|Recurrence at 3 months||0.015||< 0.001|
|Yes||5||1.23 (0.51)||3.40 (1.27–9.16)||30||3.0 (0.32)||20.9 (11.12–39.3)|
Factors significantly affecting progression-free survival on univariate analysis were histological grade (P = 0.001), DNA ploidy (P < 0.001), tumour configuration (P = 0.007) and recurrence (P < 0.001). However, factors maintaining independent significance on multivariate analysis for progression-free survival were recurrence, histological grade and DNA ploidy (Table 2).
The range of PHS for significant factors in recurrence-free survival was 0.0–7.14, and this was divided equally into three risk groups (Table 3) . Kaplan-Meier survival curves were constructed for these risk groups (Fig. 1a); the difference among the groups in recurrence-free survival, as estimated by the log-rank test, was significant (P < 0.001).
|Risk category||N||PI range||Mean (sem) 5-year |
disease-free survival, %
The range of PHS for significant factors in progression-free survival was 0.0–5.84; again, it was divided equally into three groups (Table 3). Kaplan-Meier curves for these risk groups (Fig. 1b) showed a statistically significant difference among the groups in progression-free survival (log-rank test P < 0.001).
The PHS for the test series for significant factors in recurrence-free survival was applied to the validation series, e.g. if a patient had multiple, recurrent T1, grade 2, aneuploid tumours, the PHS was calculated as the sum of the regression coefficients (from the test series) for multiple tumour (0.766), recurrent tumour (0.419), papillary tumour (0), stage T1 (2.667) and aneuploid tumour (0.713) giving 4.565. With this total PHS this patient would be in the medium-risk category of the validation series. The range of PHS was divided into the three risk categories (Table 3) and Kaplan-Meier survival curves constructed for them; there was a statistically significant difference among them for recurrence-free survival (log-rank test, P < 0.001; Fig 1a). The PI of the test series was reproducible in the validation series and a separate comparison of Kaplan-Meier curves for each risk group in each series gave insignificant differences between the test and validation series (log-rank P of 0.81, 0.08 and 0.11, respectively).
The range of PHS for the significant factors in progression-free survival for the test group was also used to give three risk groups (Table 3) and Kaplan-Meier curves for these risk groups showed statistically significant differences among then for progression-free survival (P < 0.001; Fig. 1b). The PI of the test series was reproducible in the validation series and a separate comparison of Kaplan-Meier survival curves of each risk group in each series gave insignificant differences between test and validation series (log-rank P of 0.34, 0.54 and 0.17, respectively).
TCC of the bladder is characterized by many recurrences with time and in the whole urothelium, making it probably the most frequently diagnosed tumour in man . Adjuvant intravesical therapy can be given after TURBT to control existing tumours, avert new tumour occurrence and possibly prevent progression to muscle-invasive disease [12,13]. However, superficial bladder tumours are not uniform and the response to intravesical adjuvant therapy is heterogeneous. Also, some low risk tumours may not recur for years and need no adjuvant therapy after TURBT.
Extensive studies have assessed the prognostic value of different clinicopathological tumour variables, e.g. stage, grade, size, multiplicity, location, configuration, rate and pattern of recurrence at study entry, association with CIS and other dysplastic urothelial changes, and recurrence at the check cystoscopy [5,7,14–22]. Quantitative morphometric methods, flow cytometry and cell image analysis also provide further prognostic information [23,24]. Furthermore, several cellular and nuclear markers have been identified as having potential prognostic value . The prognostic significance of most of these factors is well documented but the relative value of each remains uncertain . Most studies assessing these factors used univariate analysis, i.e. each prognostic variable was examined individually for any relationship with recurrence, progression, disease-free survival or duration of remission ; this approach has the advantages of simplicity and transparency. However, it does not consider possible interactions among the predictive variables . In addition, most previous studies have been criticised as either being retrospective or including heterogeneous groups of patients, e.g. some assessed primary tumours, others primary and recurrent tumours, some included patients treated with TURBT alone, others only those receiving adjuvant therapy. As yet there is no full consensus about the arrangement of different prognostic variables in hierarchical order, according to their significance in predicting the risk of recurrence and/or progression.
In the present study multivariate analysis, using a Cox proportional-hazard model, was used to circumvent these limitations and to identify the relative importance of several correlated prognostic factors. Seven factors influenced the recurrence-free survival independently and significantly, i.e. tumour stage, DNA ploidy, multiplicity, history of recurrence, tumour configuration, recurrence at cystoscopy and the use and type of adjuvant intravesical therapy. The recurrence rate was higher in T1, aneuploid, multiple, recurrent, solid tumours, in patients with recurrence at cystoscopy and/or in control patients receiving no adjuvant therapy compared with Ta, diploid or tetraploid, single, primary, papillary tumours, in patients with no recurrence and/or patients receiving adjuvant therapy. Also, sequential therapy with BCG and epirubicin (group 1) and BCG alone (group 2) gave a better recurrence-free survival than in groups receiving cytotoxic chemotherapy (adriamycin or epirubicin). The present results cover those in most studies assessing prognostic factors in superficial bladder tumours [5,7,14–22,28–30]. Tumour stage is a major prognostic factor for recurrence in most studies [28–30] and in the present, but it was not significant in some [7,16,17]. In groups of patients with superficial bladder cancer treated with adjuvant intravesical therapy (chemo- or immunotherapy) like most of the present patients, the biological behaviour of the tumour may be altered , and thus the influence of stage and grade as prognostic factors may be less marked. Grade was not a significant factor in recurrence-free survival in the present and earlier studies [7,16,17,30]. Tumour multiplicity significantly affected recurrence in three studies [14,16,17] and in the present.
A positive history of recurrence at study entry was a significant factor for a higher probability of future recurrences in some studies [14,16]. Dalesio et al. reported that the recurrence rate before study entry was a good predictor of prognosis. Parmar et al. used recurrence at the 3-month cystoscopy as an alternative to recurrence rate before study entry, and found a strong correlation between it and future recurrences. Many studies have shown that measuring DNA ploidy can provide additional prognostic information to the conventional factors of stage and grade [23,31,32]. Hemstreet et al. stated that the most important predictors of death and recurrence were stem-line aneuploidy and the presence of cells with > 5C DNA, respectively. In the present and a previous report there was an independent and significant correlation between DNA ploidy and disease-free survival ; DNA ploidy could also predict the response to intravesical therapy .
A positive recurrence at the first check cystoscopy significantly affected recurrence- and progression-free survival, a conclusion also reported by others [5,16,33]. Fitzpatrick et al. noted the importance of recurrence pattern in 414 new patients with stages pTa, grade 1 and 2 disease treated by TURBT. In that study, patients free of tumour at 3 months had an 80% chance of having no further recurrences, while those with a recurrence at 3 months were much less likely to remain free of tumour, and had a higher chance of recurrence at every future visit. In addition, Parmar et al., using multivariate Cox regression analysis, found that the 3-month cystoscopy finding and the number of tumours at presentation were independent significant prognostic variables for recurrence. Furthermore, Solsona et al. reported that recurrence at the 3-month cystoscopy was a significant predictor of muscle-invasive progression, and advocated considering early cystectomy when stage T1 grade 3 tumour, bladder CIS or prostatic mucosal or duct involvement was present at the 3-month clinical evaluation.
In addition to recurrence at the first check cystoscopy, DNA ploidy and histological grade were significant prognostic factors for muscle-invasive progression in the present multivariate analysis. As most prognostic tumour biomarkers have not been confirmed to be of significant clinical utility, and because of the great discrepancy between the results of relevant studies, we elected not to include such tumour markers in the present multivariate analysis.
There have been few attempts to construct a PI and use it to categorize superficial bladder tumours into different risk groups. The aim of such categorization is to standardize different therapeutic and follow-up approaches for the different risk groups. Herr et al. defined four risk groups based on two variables, i.e. the presence or absence of T1 disease and positive or negative urinary cytology. In that study the lowest probability of tumour progression was in the group with a negative biopsy and negative cytology. Parmar et al. used two variables (the result of cystoscopy at 3 months and the number of tumours) to construct a PI for recurrence in primary superficial bladder tumour. As the regression coefficients of the two selected variables were similar, the authors reported three prognostic groups: negative cystoscopy at 3 months and single tumour, positive cystoscopy or multiple tumours, and positive cystoscopy and multiple tumours at presentation. Although the three-category PI proposed by this group is simple, it does not include many variables that are considered by most urologists to be important determinants of recurrence in superficial bladder tumours, e.g. stage, grade and size of the tumour.
In a retrospective study  the prognostic factor analysis of Parmar et al. was applied to a group of 232 newly diagnosed consecutive patients with Ta/T1 TCC of the bladder. These authors concluded that this prognostic factor analysis provided a reliable guide to the likelihood of tumour recurrence but stated that modifying follow-up cystoscopy schedules from this analysis required a prospective study. Three risk groups were defined by another study  from a prognostic analysis and the objectively calculated risk for the patients; the regression coefficient of each of the statistically significant factors was used to calculate the relative risk (Z-score) for certain events (recurrence, progression or patient death). The sum of the corresponding Z-score values of the significant factors was used to calculate the individual PIs and to categorize the patients into different risk groups.
Allard et al. described a PI based on the number of adverse primary tumour characteristics, i.e. primary tumour multiplicity, diameter > 3 cm, stage T1 and grade 2 or 3. They found that their simple proposed PI was a strong indicator of the clinical course of superficial bladder cancer within 3 years of the first TURBT, but the relative value of each of these prognostic factors was not considered.
The PI calculated in the present study was used to advocate different treatment and follow-up strategies, and was validated by applying the model to another series of 156 patients. The model was reproducible but further validation is needed by applying the PI in future studies at different centres.
transurethral resection of bladder tumour
carcinoma in situ
proportional hazard score.