© 1991 S. Karger AG, Basel 0302-2838/91/0192-0097Î2.75/0
Eur Urol 1991;19:97-100
Local BCG Failures in Superficial Bladder Cancer A Multivariate Analysis of Risk Factors Influencing Survival
1616844
Harry W. Herr, Eric A. Klein, André Rogatko Urologie Service, Department of Surgery, and Department of Biostatistics, Memorial Sloan-Kettering Cancer Center, New York, N.Y., USA
Key Words. Superficial bladder tumor • BCG ■ Local failure Abstract. Of 221 patients with superficial bladder cancers treated with intravesical BCG, BCG therapy failed locally within the bladder or prostate in 41 patients. A multivariate analysis was done using data on these local failures to identify clinical variables influencing survival. Patients presenting with T i tumors which relapsed initially in the prostate were at highest risk of dying of bladder cancer. Cystectomy is justified in such patients. Others with noninvasive local refractory disease can be managed by continued conservative therapy.
BCG is an effective therapy for superficial bladder tumors, but its growing use has generated a current prob lematic question - how to manage the patient with per sistent or recurrent noninvasive (Ta, TiS) or superficially invasive (Ti) transitional carcinoma confined to the bladder or prostate? Such patients are considered ‘local’ failures and some, but not all, are at risk for progression to muscle invasion or métastasés. Clearly, alternative treatment strategies are needed, but few data exist indi cating optimal therapy for individual cases. For some, immediate cystectomy is required. For others, continued conservative therapy may suffice. To address this ques tion, we performed a multivariate analysis of risk factors potentially impacting on deaths due to bladder cancer in patients locally failing initial BCG therapy.
Materials and Methods Of 221 patients with superficial bladder tumor (Ta, Tls, Ti ) receiv ing intravesical BCG (Frappier) between March 1978 and July 1984, BCG therapy failed locally within the bladder or prostate in 41 patients. After 6 weekly instillations, interval evaluations were done every 3 months with cystoscopy, biopsy and urinary cytology.
Local failure was defined as a need for a change in therapy after BCG because of (1) persistent tumor at the same stage within 6 months (no response to BCG), (2) recurrence of the same T stage following an initial complete response to BCG, (3) progression of T a> Tls to T,, (4) progression to muscle invasion (stage T2 or T3) or (5) involvement of the prostatic urethra, ducts or stroma (T4). The mean interval to local failure was 14.7 (range: 3-55) months. New treatment instituted at the time of local failure included a second course of intravesical therapy in 27 (66%) patients (BCG in 17 and chemotherapy in 10), partial cystectomy in 2 (5%), immediate radical cystectomy in 11 (27%) and systemic chemotherapy in 1 (2%). Ten patients subsequently underwent delayed cystectomy at a mean of 35 months after initial BCG fail ure. The indications for immediate and delayed cystectomy are shown in table 1. Follow-up was complete on all patients through January 1988 or until death. Statistical Methods Our main objective was to assess the effect of various clinical variables on the rate of cancer deaths. Survival was measured from the time of local failure until death from disease or last follow-up (minimum > 5 years). Survival functions were estimated for each of the variables by the product-limit method. The dependence of sur vival time on each variable taken one at a time (univariate analysis) was determined by the logrank test. Proportional hazard regression [1] was used to incorporate all of the variables in the same model (multivariate analysis). Forward stepwise procedure and likelihood ratio tests were used to select the variables with the greatest prog nostic value. For calculation of survival, patients were considered Downloaded by: University of Exeter 144.173.6.94 - 5/8/2020 3:53:34 AM
Introduction
Herr/Klein/Rogatko
98
1.00
I
XU—u
Uli I III 111
delayed (10); (7) upper tract tumors - yes (10) versus no (31); (8) prostatic involvement - yes (10) versus no (31), and (9) time of prostatic failure - early (5) versus late (5). Early prostatic relapse was defined as the first indication of BCG failure and late relapse as any prostatic involvement detected at any time after other indica tions of BCG failure.
JU XX
0 80-
0.60-
Results p 0.40 -
Univariate Analysis 3 c 0.20 o
0 00
20 00
T
T
T
40 00
60 00
80.00
1
100 00
120.00
Change in treatment, months
Multivariate Analysis
Fig. 1. Survival as function of combining initial T i tumor stage and early prostatic failure by multivariate analysis. A = Patients with noninvasive tumors (Ta or TIS) and no prostatic relapse; □ = patients with superficially invasive (T i) tumors and early prostatic failure. Graduations indicate last follow-up.
Table 1. Indications for cystectomy
Discussion Cystectomy, n
Initial nonresponse to BCG Endoscopically uncontrolled T, Progression in stage to T, T2 or T3 T4a
After neoadjuvant systemic chemotherapy Severe irritative symptoms (T0)
Table 3 shows results of the multivariate analysis. Combining the significant variables of early prostatic failure plus T i tumor stage identified a group of patients at significant (p = 0.008) risk of dying of bladder cancer. Figure 1 displays estimated survival curves for this highrisk group.
immediate (n = 11)
delayed (n = 10)
5 1
3
2
0 3
1
2
0 0
1 1
2
Includes 2 patients with T1S + T1; 1 with T is alone and 1 with T i alone.
‘censored’ if they were still alive at the time of the last follow-up. Clinical variables studied were: (1) initial stage - Ta, Tls (32) versus Ti (9); (2) reason for change in treatment - progression (13) versus nonprogression (28) in T stage; (3) stage at the time of local failure - Ta, Tls (22) versus Ti (12) versus T2t (7); (4) intravesical therapy after a second course - yes (12) versus no (8); (5) bladder preservation - cystectomy (21) versus no cystectomy (20); (6) time of cystectomy - immediate (11) versus
Few studies address the issue of whether conservative management is safe in patients failing BCG with super ficial tumors confined to the bladder or prostate. Some have suggested that a second course of BCG can cure some of these patients [2] whereas others feel cystectomy is appropriate [3]. Not all tumors which recur or persist after BCG are destined to proceed (at least within 5 years) to muscle invasion or métastasés. Many patients may not need cys tectomy to control refractory tumors if they can be man aged by repeated transurethral resection and further intravesical therapy. The challenge is to identify predic tive factors, either at initial diagnosis or at the time of failure, which distinguish between patients needing early cystectomy and those safely manageable by conservative therapies preserving bladder and sexual function. Using univariate analysis, we identified two risk fac tors which had an adverse impact on survival. These were involvement of the prostate at the time of BCG failure and the presence of T i tumor when BCG therapy was begun. Early prostatic failure included 5 patients (2 with urethral or ductal disease and 3 with stromal inva sion). Late prostatic involvement also included 5 pa tients (all with urethral disease only) and did not adDownloaded by: University of Exeter 144.173.6.94 - 5/8/2020 3:53:34 AM
oa o a 0 00
Table 2 shows results of the univariate analysis. Only early post-BCG prostatic involvement (p = 0.016) and initial Ti tumor stage (p = 0.25) adversely influenced survival.
Local BCG Failures
99
Table 2. Statistical description of the explanatory variables Variables
Survival, months
n
Censored %
Initial stage Noninvasive Superficially invasive Reason for change in treatment Nonprogression Progression Stage at time of local failure Noninvasive Superficially invasive Muscle invasive Subsequent treatment No Yes Bladder preservation Cystectomy No cystectomy Time of cystectomy Immediate Delayed Upper tract tumors No Yes Early prostatic involvement No Yes Late prostatic involvement No Yes
Dead of disease n
mean
SEM
105.57 59.17
7.831 10.458
32 9
90.62 66.67
3 3
99.41 75.93
9.593 8.263
28 13
85.71 84.62
3 3
101.09 79.00 61.00
10.359 5.612 11.475
22 12 7
86.36 91.67 71.93
3 2
69.00 90.35
15.802
8 12
100.00 75.00
79.48 97.37
4.935 12.287
21 20
71.18 87.00
8.354
11
77.11 109.00
Logrank test ---------------d.f. statistics
p value
5.207
0.025
1
0.006
0.9374
2
1.734
0.4202
0 3
1
1.582
0.2084
85.71 85.00
3 3
1
0.071
0.7897
72.73 100.00
3 0
1
3.000
0.833
10 4.487 7.589
31 10
83.33 40.00
5 1
1
0.290
0.5900
104.39 51.00
6.951 12.816
36 5
88.89 60.00
4 2
1
5.839
0.0157
103.79 75.00
6.080
36 5
86.11 80.00
5 1
0.064
0.8006
Table 3. Survival according to multivariate analysis: Statistical description of the levels of the new variable Clinical variables
Initial Ta, Tls without early prostatic involvement Initial T1 stage and early prostatic failure
Survival, months mean
SEM
107.79 60.56
8.793 8.439
n
Censored, %
Dead of disease, n
28 13
92.86 69.23
2 4
Logrank test: d.f. = 1, statistics = 7.006, p value = 0.0081.
have died of disease compared with 2 (7%) among the other 28 with other tumor diathesis. Several other variables had no impact on survival (ta ble 2). These included the reason for change in therapy at the time of BCG failure as defined by progression versus nonprogression in T stage, actual T stage at the time of Downloaded by: University of Exeter 144.173.6.94 - 5/8/2020 3:53:34 AM
versely affect survival. Nine patients presented with T i tumor and 3 died of disease. A multivariate statistical model which combined both early prostatic involvement and T i tumor at initial diag nosis resulted in definition of a group of patients at high est risk for worse survival. Of 13 such patients, 4 (31 %)
Herr/Klein/Rogatko
100
Table 4. P categories of patients undergoing cystectomy Pathologic stage
Cystectomy, n immediate (n = 11)
PoNoMo P1S or PA N0M0 Pis. PiNoMo Pis, PtaNoMo
P4aNoMo Any N+Mo
1 3 3 0 1 3
delayed (n = 10) 1
2 2
2
In summary, this study shows that patients receiving BCG for T1 tumors who relapse initially within the pros tate are at highest risk of dying from bladder cancer. Immediate cystectomy in such patients is justified. Other patients locally failing BCG therapy can be further managed with intravesical measures. Cystectomy is re served for persistent nonresponders without fear of re ducing overall survival.
2 1
References 1
Harry W. Herr, MD Memorial Hospital 1275 York Avenue New York, NY 10021 (USA)
Downloaded by: University of Exeter 144.173.6.94 - 5/8/2020 3:53:34 AM
BCG failure and the occurrence of upper tract tumors. Cystectomy had no impact on survival whether or not it was performed early (at the time of first BCG failure) or late (at the time of subsequent failure). Table 4 shows that there were no significant differences in the P stages between these two cystectomy groups. Patients undergo ing delayed cystectomy had bladder function preserved for a mean of 35 months from diagnosis without jeopar dizing survival.
Cox DR: Regression models and life-tables. J R Stat Soc 1972; 34:187. 2 Catalona WJ, Hudson MA, Gillen DP: Risks and benefits of repeated courses of intravesical BCG therapy for superficial bladder cancer. J Urol 1987; 137:220. 3 Stöckle M, Alken P, Engelmann E: Radical cystectomy - Often too late? Eur Urol 1987;13:361.
Eur Urol 1991;19:97-100
Local BCG Failures in Superficial Bladder Cancer A Multivariate Analysis of Risk Factors Influencing Survival
1616844
Harry W. Herr, Eric A. Klein, André Rogatko Urologie Service, Department of Surgery, and Department of Biostatistics, Memorial Sloan-Kettering Cancer Center, New York, N.Y., USA
Key Words. Superficial bladder tumor • BCG ■ Local failure Abstract. Of 221 patients with superficial bladder cancers treated with intravesical BCG, BCG therapy failed locally within the bladder or prostate in 41 patients. A multivariate analysis was done using data on these local failures to identify clinical variables influencing survival. Patients presenting with T i tumors which relapsed initially in the prostate were at highest risk of dying of bladder cancer. Cystectomy is justified in such patients. Others with noninvasive local refractory disease can be managed by continued conservative therapy.
BCG is an effective therapy for superficial bladder tumors, but its growing use has generated a current prob lematic question - how to manage the patient with per sistent or recurrent noninvasive (Ta, TiS) or superficially invasive (Ti) transitional carcinoma confined to the bladder or prostate? Such patients are considered ‘local’ failures and some, but not all, are at risk for progression to muscle invasion or métastasés. Clearly, alternative treatment strategies are needed, but few data exist indi cating optimal therapy for individual cases. For some, immediate cystectomy is required. For others, continued conservative therapy may suffice. To address this ques tion, we performed a multivariate analysis of risk factors potentially impacting on deaths due to bladder cancer in patients locally failing initial BCG therapy.
Materials and Methods Of 221 patients with superficial bladder tumor (Ta, Tls, Ti ) receiv ing intravesical BCG (Frappier) between March 1978 and July 1984, BCG therapy failed locally within the bladder or prostate in 41 patients. After 6 weekly instillations, interval evaluations were done every 3 months with cystoscopy, biopsy and urinary cytology.
Local failure was defined as a need for a change in therapy after BCG because of (1) persistent tumor at the same stage within 6 months (no response to BCG), (2) recurrence of the same T stage following an initial complete response to BCG, (3) progression of T a> Tls to T,, (4) progression to muscle invasion (stage T2 or T3) or (5) involvement of the prostatic urethra, ducts or stroma (T4). The mean interval to local failure was 14.7 (range: 3-55) months. New treatment instituted at the time of local failure included a second course of intravesical therapy in 27 (66%) patients (BCG in 17 and chemotherapy in 10), partial cystectomy in 2 (5%), immediate radical cystectomy in 11 (27%) and systemic chemotherapy in 1 (2%). Ten patients subsequently underwent delayed cystectomy at a mean of 35 months after initial BCG fail ure. The indications for immediate and delayed cystectomy are shown in table 1. Follow-up was complete on all patients through January 1988 or until death. Statistical Methods Our main objective was to assess the effect of various clinical variables on the rate of cancer deaths. Survival was measured from the time of local failure until death from disease or last follow-up (minimum > 5 years). Survival functions were estimated for each of the variables by the product-limit method. The dependence of sur vival time on each variable taken one at a time (univariate analysis) was determined by the logrank test. Proportional hazard regression [1] was used to incorporate all of the variables in the same model (multivariate analysis). Forward stepwise procedure and likelihood ratio tests were used to select the variables with the greatest prog nostic value. For calculation of survival, patients were considered Downloaded by: University of Exeter 144.173.6.94 - 5/8/2020 3:53:34 AM
Introduction
Herr/Klein/Rogatko
98
1.00
I
XU—u
Uli I III 111
delayed (10); (7) upper tract tumors - yes (10) versus no (31); (8) prostatic involvement - yes (10) versus no (31), and (9) time of prostatic failure - early (5) versus late (5). Early prostatic relapse was defined as the first indication of BCG failure and late relapse as any prostatic involvement detected at any time after other indica tions of BCG failure.
JU XX
0 80-
0.60-
Results p 0.40 -
Univariate Analysis 3 c 0.20 o
0 00
20 00
T
T
T
40 00
60 00
80.00
1
100 00
120.00
Change in treatment, months
Multivariate Analysis
Fig. 1. Survival as function of combining initial T i tumor stage and early prostatic failure by multivariate analysis. A = Patients with noninvasive tumors (Ta or TIS) and no prostatic relapse; □ = patients with superficially invasive (T i) tumors and early prostatic failure. Graduations indicate last follow-up.
Table 1. Indications for cystectomy
Discussion Cystectomy, n
Initial nonresponse to BCG Endoscopically uncontrolled T, Progression in stage to T, T2 or T3 T4a
After neoadjuvant systemic chemotherapy Severe irritative symptoms (T0)
Table 3 shows results of the multivariate analysis. Combining the significant variables of early prostatic failure plus T i tumor stage identified a group of patients at significant (p = 0.008) risk of dying of bladder cancer. Figure 1 displays estimated survival curves for this highrisk group.
immediate (n = 11)
delayed (n = 10)
5 1
3
2
0 3
1
2
0 0
1 1
2
Includes 2 patients with T1S + T1; 1 with T is alone and 1 with T i alone.
‘censored’ if they were still alive at the time of the last follow-up. Clinical variables studied were: (1) initial stage - Ta, Tls (32) versus Ti (9); (2) reason for change in treatment - progression (13) versus nonprogression (28) in T stage; (3) stage at the time of local failure - Ta, Tls (22) versus Ti (12) versus T2t (7); (4) intravesical therapy after a second course - yes (12) versus no (8); (5) bladder preservation - cystectomy (21) versus no cystectomy (20); (6) time of cystectomy - immediate (11) versus
Few studies address the issue of whether conservative management is safe in patients failing BCG with super ficial tumors confined to the bladder or prostate. Some have suggested that a second course of BCG can cure some of these patients [2] whereas others feel cystectomy is appropriate [3]. Not all tumors which recur or persist after BCG are destined to proceed (at least within 5 years) to muscle invasion or métastasés. Many patients may not need cys tectomy to control refractory tumors if they can be man aged by repeated transurethral resection and further intravesical therapy. The challenge is to identify predic tive factors, either at initial diagnosis or at the time of failure, which distinguish between patients needing early cystectomy and those safely manageable by conservative therapies preserving bladder and sexual function. Using univariate analysis, we identified two risk fac tors which had an adverse impact on survival. These were involvement of the prostate at the time of BCG failure and the presence of T i tumor when BCG therapy was begun. Early prostatic failure included 5 patients (2 with urethral or ductal disease and 3 with stromal inva sion). Late prostatic involvement also included 5 pa tients (all with urethral disease only) and did not adDownloaded by: University of Exeter 144.173.6.94 - 5/8/2020 3:53:34 AM
oa o a 0 00
Table 2 shows results of the univariate analysis. Only early post-BCG prostatic involvement (p = 0.016) and initial Ti tumor stage (p = 0.25) adversely influenced survival.
Local BCG Failures
99
Table 2. Statistical description of the explanatory variables Variables
Survival, months
n
Censored %
Initial stage Noninvasive Superficially invasive Reason for change in treatment Nonprogression Progression Stage at time of local failure Noninvasive Superficially invasive Muscle invasive Subsequent treatment No Yes Bladder preservation Cystectomy No cystectomy Time of cystectomy Immediate Delayed Upper tract tumors No Yes Early prostatic involvement No Yes Late prostatic involvement No Yes
Dead of disease n
mean
SEM
105.57 59.17
7.831 10.458
32 9
90.62 66.67
3 3
99.41 75.93
9.593 8.263
28 13
85.71 84.62
3 3
101.09 79.00 61.00
10.359 5.612 11.475
22 12 7
86.36 91.67 71.93
3 2
69.00 90.35
15.802
8 12
100.00 75.00
79.48 97.37
4.935 12.287
21 20
71.18 87.00
8.354
11
77.11 109.00
Logrank test ---------------d.f. statistics
p value
5.207
0.025
1
0.006
0.9374
2
1.734
0.4202
0 3
1
1.582
0.2084
85.71 85.00
3 3
1
0.071
0.7897
72.73 100.00
3 0
1
3.000
0.833
10 4.487 7.589
31 10
83.33 40.00
5 1
1
0.290
0.5900
104.39 51.00
6.951 12.816
36 5
88.89 60.00
4 2
1
5.839
0.0157
103.79 75.00
6.080
36 5
86.11 80.00
5 1
0.064
0.8006
Table 3. Survival according to multivariate analysis: Statistical description of the levels of the new variable Clinical variables
Initial Ta, Tls without early prostatic involvement Initial T1 stage and early prostatic failure
Survival, months mean
SEM
107.79 60.56
8.793 8.439
n
Censored, %
Dead of disease, n
28 13
92.86 69.23
2 4
Logrank test: d.f. = 1, statistics = 7.006, p value = 0.0081.
have died of disease compared with 2 (7%) among the other 28 with other tumor diathesis. Several other variables had no impact on survival (ta ble 2). These included the reason for change in therapy at the time of BCG failure as defined by progression versus nonprogression in T stage, actual T stage at the time of Downloaded by: University of Exeter 144.173.6.94 - 5/8/2020 3:53:34 AM
versely affect survival. Nine patients presented with T i tumor and 3 died of disease. A multivariate statistical model which combined both early prostatic involvement and T i tumor at initial diag nosis resulted in definition of a group of patients at high est risk for worse survival. Of 13 such patients, 4 (31 %)
Herr/Klein/Rogatko
100
Table 4. P categories of patients undergoing cystectomy Pathologic stage
Cystectomy, n immediate (n = 11)
PoNoMo P1S or PA N0M0 Pis. PiNoMo Pis, PtaNoMo
P4aNoMo Any N+Mo
1 3 3 0 1 3
delayed (n = 10) 1
2 2
2
In summary, this study shows that patients receiving BCG for T1 tumors who relapse initially within the pros tate are at highest risk of dying from bladder cancer. Immediate cystectomy in such patients is justified. Other patients locally failing BCG therapy can be further managed with intravesical measures. Cystectomy is re served for persistent nonresponders without fear of re ducing overall survival.
2 1
References 1
Harry W. Herr, MD Memorial Hospital 1275 York Avenue New York, NY 10021 (USA)
Downloaded by: University of Exeter 144.173.6.94 - 5/8/2020 3:53:34 AM
BCG failure and the occurrence of upper tract tumors. Cystectomy had no impact on survival whether or not it was performed early (at the time of first BCG failure) or late (at the time of subsequent failure). Table 4 shows that there were no significant differences in the P stages between these two cystectomy groups. Patients undergo ing delayed cystectomy had bladder function preserved for a mean of 35 months from diagnosis without jeopar dizing survival.
Cox DR: Regression models and life-tables. J R Stat Soc 1972; 34:187. 2 Catalona WJ, Hudson MA, Gillen DP: Risks and benefits of repeated courses of intravesical BCG therapy for superficial bladder cancer. J Urol 1987; 137:220. 3 Stöckle M, Alken P, Engelmann E: Radical cystectomy - Often too late? Eur Urol 1987;13:361.
