Pretreatment prognostic factors for survival in patients with advanced urothelial tumors treated in a phase I/II trial with paclitaxel, cisplatin, and gemcitabine


  • The authors write on behalf of the Spanish Oncology Genitourinary Group (SOGUG).



New chemotherapeutic agents, including paclitaxel and gemcitabine, are active in advanced bladder carcinoma, and combination regimens with these agents have shown promising results. Unlike conventional chemotherapy regimens, such as methotrexate, vinblastine, doxorubicin, and cisplatin, there are no data available on key predictive factors for response and survival with these novel agents. Since this information is needed for selection of patients for these new combinations and for stratification purposes in ongoing randomized trials, the authors aimed to study the predictive factors for response and survival to the current regimen containing cisplatin, paclitaxel, and gemcitabine.


The authors studied 56 patients with advanced urothelial tumors treated on a Phase I/II trial of paclitaxel, cisplatin, and gemcitabine (TCG) to identify pretreatment characteristics that were prognostic for survival using this novel combination. The pretreatment characteristics analyzed were age, gender, Eastern Cooperative Oncology Group performance status, histopathology (pure transitional versus other), visceral (liver, lung, or bone) metastasis, number of sites of disease, lactate dehydrogenase, and hemoglobin.


The factors that were associated with a worse survival in univariate analysis were performance status > 0, presence of visceral metastasis, and more than one site of malignant disease. In a multivariate model, performance status (P = 0.044) and visceral disease (P = 0.008) showed independent statistical significance for decreased survival. Patients were then grouped based on these two independent prognostic factors. Median survival times in the groups of patients with zero, one, or two of these risk factors were 32.8 months, 17 months, and 9.6 months, respectively (P = 0.0005).


A pretreatment performance status > 0 and the presence of visceral metastasis have a profound impact on survival when using the TCG regimen. These two variables will be used to stratify patients in the upcoming Phase III randomized trial comparing this TGC regimen with a gemcitabine/cisplatin regimen in advanced urothelial tumors. Cancer 2002;95:751–7. © 2002 American Cancer Society.

DOI 10.1002/cncr.10762

Cisplatin based combination chemotherapy regimens like MVAC (methotrexate, vinblastine, doxorubicin, and cisplatin) or CMV (cisplatin, methotrexate, and vinblastine) have been considered among the standard chemotherapeutic regimens for patients with advanced carcinoma of the urothelium.1, 2 The overall response rate (complete plus partial) with these cisplatin based combination regimens ranges from 50 to 70%, with complete responses seen in 15 to 25% of cases.1, 3 Nonetheless, almost all responding patients ultimately relapse within the first year, with a median survival of approximately 12-14 months4 and a > 6 years disease free survival in 3.7% of patients.5 Several pretreatment disease related prognostic factors that can affect outcome in patients with advanced urothelial tumors treated with MVAC have been identified. A series of studies in patients treated within MVAC clinical trials convincingly showed that pretreatment performance status and visceral metastatic sites were independent prognostic factors that predicted patient survival.6, 7

The dismal long term outcome with currently available regimens has led to the search for new treatment approaches. Among the recently available new chemotherapy agents, paclitaxel and gemcitabine have shown to have a promising activity profile in bladder carcinoma,8, 9, 10 and efforts have been aimed at integrating them into combination regimens, in particular with cisplatin.11–19 Taking in consideration the significant activity of paclitaxel and gemcitabine, either alone or in combination with cisplatin, their different mechanism of action, and their nonoverlapping toxicities, we decided to study the feasibility and activity of a regimen combining paclitaxel, cisplatin, and gemcitabine (TCG) in patients with advanced urothelial transitional cell carcinoma (TCC). This combination regimen was highly active in patients with advanced urothelial tumors and had a tolerable toxicity profile; survival data were encouraging.20

Based on these promising Phase I/II data with our three-agent TCG combination,20 and on the results of a randomized trial showing similar efficacy of MVAC compared with a less toxic cisplatin/gemcitabine regimen,21 an Intergroup Trial (jointly coordinated by EORTC and SWOG [European Organization for Research and Treatment of Cancer and Southwest Oncology Group, respectively]) randomized study comparing TCG with a regimen of cisplatin and gemcitabine in advanced urothelial tumors is being conducted. In the ongoing randomized trial EORTC #30987, survival will be the major endpoint of the study. To guarantee an even distribution among the two treatment groups, it was necessary to identify pretreatment factors that might have an influence on survival. Studies with the MVAC regimen have showed convincingly that pretreatment factors such as performance status and visceral metastasis predict survival likelihood.7 However, whether the same prognostic factors were applicable to bladder carcinoma patients treated with newer regimens including gemcitabine or paclitaxel was yet unknown. In particular, taxanes have been shown to be equally active in visceral and nonvisceral sites in other tumor types,22 and we could not anticipate whether the same activity profile might apply to bladder carcinoma as well. Therefore, we analyzed potentially relevant clinicopathologic pretreatment factors and their relationship with survival in the cohort of patients treated in the Phase I/II trial of the TCG regimen.20 We found that performance status and visceral metastasis were the two pretreatment independent prognostic factors for survival, and, accordingly, they have been incorporated as the stratifying variables in the randomized clinical trial comparing TCG versus a regimen of gemcitabine and cisplatin.


Patient Population

From March 1997 to December 1998 a total of 61 patients from 11 participating institutions were included in the trial.20 Fifteen patients were entered in the Phase I component of the study (three at dose level 1, three at dose level 2, three at dose level 3, and six at dose level 4; Table 1). Once dose level 3 was chosen for the Phase II study, 46 additional patients were entered at this dose level, for a total of 49 patients registered at dose level 3. Of these 46 additional patients, 3 patients were evaluable for toxicity but not for response; in one patient the pathology review failed to confirm the presence of TCC after having received Day 1 of therapy of the first cycle. The second patient was excluded due to an early death secondary to a massive bowel infarction that was diagnosed 10 days after Day 1 of therapy of the first cycle. A third patient at dose level 3 received treatment on Day 1 only, refused further therapy, and was lost to followup. Overall, 58 patients on study were eligible and evaluable for response and toxicity, as previously reported.20 Two of these 58 patients had insufficient on-study information, which gave finally 56 patients assessable for the current prognostic analysis.

Table 1. Dose Escalation Scheme
Dose Levelcisplatin Day 1 (mg/m2)gemcitabine Days 1 and 8 (mg/m2)paclitaxel Days 1 and 8 (mg/m2)No. of evaluable patients

The details of the Phase I/II trial have been previously published.20 Briefly, the investigational chemotherapy regimen consisted of paclitaxel (1 hour infusion), followed by cisplatin at a fixed dose of 70 mg/m2, and subsequently gemcitabine in a 30 minute infusion on Day 1. Paclitaxel and gemcitabine were administered again at the same doses on Day 8. The studied dose levels for paclitaxel and gemcitabine are shown in Table 1. Patients continued treatment up to a maximum of six cycles if stable or responding disease was shown and toxicity was acceptable. Response was assessed by re-evaluation of known sites of disease by physical examination, cystoscopy, and radiography after every two cycles of therapy or as clinically indicated. After achieving a maximum response to chemotherapy, selected patients were referred for postchemotherapy surgical resection of residual carcinoma and assessment of pathologic response.

Survival Analysis

Survival was measured from time of initiation of chemotherapy until death or last followup. The following categorical pretreatment characteristics were examined in univariate analysis for their possible association with survival: lactate dehydrogenase (LDH normal vs. abnormal [greater than the normal upper limit]), hemoglobin (< 10 g/dL vs. ≥ 10 g/L), age (≤ 60 years vs. > 60 years), gender (male vs. female), Eastern Cooperative Oncology Group (ECOG) performance status (0 vs. 1 and 2, 0 vs. 1 vs. 2), presence or absence of visceral (liver, lung, or bone) metastasis, number of disease sites, and histopathology (pure transitional cell vs. mixed histopathologies). Survival distributions were estimated using the Kaplan-Meier method.23 The relationship between survival and each of the variables was analyzed, using the log-rank test for the chosen categorical variables. A multivariate regression analysis was planned to assess the independent value of the association between all the significant variables in the univariate analysis. All statistical tests were conducted at the two-sided 0.05 level of significance.


Patient Characteristics

The distribution of pretreatment clinical characteristics for the 56 patients is shown in Table 2. The median age was 65 years (range, 36 to 79 years), and 71.4% of patients were male. The performance status was 0 in 17 patients (30.3%), 1 in 35 patients (62.5 %), and 2 in 4 patients (7.1%). Visceral metastases (lung, liver, or bone disease) were present in 20 (35.7%) of the 56 patients in the cohort, and 19 patients (33.9%) had multiples sites of disease. The median survival time for the entire cohort of 56 patients was 15.8 months (95% confidence interval [CI], 11.9-19.5 months) at a median followup of 18 months (Fig 1). At the time of this analysis, 17 (30.3%) of the 56 patients were alive.

Table 2. Pretreatment Patient Characteristics: Univariate Analysis for Survival
FactorNo.No. of deathsMedian survival (months; 95% CI)ORP
  • a

    Liver, lung, or bone. CI: confidence interval; OR: odds ratio; PS: performance status; LDH: lactate dehydrogenase; TCC: transitional cell carcinoma.

Age, years     
 ≤ 60151116.7 (15.2-18.2)1.040.9
 > 60412812.5 (5-19.9)(0.5-2.1) 
 Male402816.6 (10.4-22.7)0.630.23
 Female161117.7 (1.2-35.8)(0.3-1.3) 
 017925.6 (22.6-27.6)2.3 (1.1-4.9)0.024
 1352613.4 (9.8-16.9)4.7 (1.4-15.7) 
 24410.0 (1.8-16.5)  
 Normal362517.4 (15.5-19.3)1.260.55
 Abnormal12912.2 (4.7-19.8)(0.6-2.7) 
 ≥ 10493417.4 (11.8-23.0)0.840.84
 < 107515.5 (1.8-28.1)(0.3-2.1) 
Visceral diseasea     
 Present191811.4 (7.4-15.4)0.350.002
 Absent372121.7 (12.2-29.3)(0.2-0.7) 
Number of sites of disease     
 1412518.2 (6.8-29.6)1.970.04
 > 1151415.5 (11.4-18.5)(1.01-3.8) 
 Pure TCC362317.9 (11.8-22.1)1.740.1
 Mixed TCC201613.3 (11.2-14.4)(0.9-3.3) 
Figure 1.

Overall survival for the entire cohort of patients (n = 56).

Univariate Survival Analysis

In these series,20 six patients were treated at dose levels 1 or 2, while the vast majority were treated at dose level 3 (n = 46) or dose level 4 (n = 6; Table 1). However, no significant differences in survival were found between patients treated with lower doses (dose levels 1 or 2) compared to those treated with higher doses (dose levels 3 —the recommended levels for further Phase II/III studies— or dose level 4; P = 0.083). Therefore, all the patients were grouped together for additional analysis.

All the patients included in this analysis had been staged following the 1997 American Joint Committee on Cancer classification and had advanced disease as was considered incurable by surgical treatment. Since the objective of the current study was solely to analyze whether the classic MVAC prognostic factors for survival applied to our current new regimen, we decided to divide the patients in two groups in an analogous fashion to the MVAC studies: 1) nonvisceral disease, which included locally advanced disease (T4bN1-3M0) as well as metastatic lymph nodes/masses; and 2) visceral disease, which included liver, lung, bone, or other distant sites of disease. The inclusion of bone metastasis in the group of visceral disease was also due to the fact that this criterion was also used in the MVAC studies.7 By univariate analysis, there were significant differences in survival when comparing locally advanced disease (T4bN1-3M0) with visceral disease (P = 0.001) or when comparing metastatic lymph nodes/masses with visceral disease (P = 0.001). However, there were no significant differences in survival in locally advanced disease versus nonvisceral metastasis (P = 0.23). This finding provided further justification to the grouping of locally advanced and metastatic lymph nodes/masses as nonvisceral disease for further analysis.

Eight pretreatment variables were subjected to univariate analysis for prognostic significance (Table 2). Of these variables, the following were associated with an adverse prognosis: performance status greater than 0, number of sites with disease greater than 1, and the presence of visceral metastasis. Histopathology, gender, age, LDH, and hemoglobin did not have an impact on survival.

Multivariate Survival Analysis

The three pretreatment variables with univariate significance (ECOG performance status [0 vs. 1-2], number of sites, visceral metastasis) were entered into the multivariate regression analysis using the Cox proportional hazards model. These variables were selected because of their statistical significance in the univariate analysis and their potential clinical interaction. Two of these factors were determined to have independent significance for survival in the Cox model: visceral metastasis (presence or absence) and ECOG performance status as a discrete variable (0 vs. 1-2 or 0 vs. 1 vs. 2; Table 3). An ECOG performance status of 0 and the presence of visceral metastasis were associated with odds ratio of 2.19 and 0.4, respectively.

Table 3. Multivariate Analysis for Survival
FactorOR (95% CI)P
  1. OR: odds ratio; CI: confidence interval.

Performance status  
 02.19 (1.02-4.68)0.044
Visceral disease  
present0.40 (0.20-0.79)0.008
Sites of disease  
11.49 (0.74-3.00)0.26

Risk Groups

The ECOG performance status and the presence of visceral metastases were the two pretreatment risk factors that predicted survival independently of the other factors analyzed. Based on these two factors, the number of patients with zero-risk, one-risk, and two-risk factors was 14 (25%), 26 (46.4%), and 16 (28.6%), respectively (Table 4). Patients with no risk factors (ECOG performance status 0 and no visceral metastasis) had a median survival time of 32.8 months (95% CI, 25.3 to 40.3), patients with one risk factor had a median survival time of 18.0 months (95% CI, 11.5 to 24.9), and patients with the two risk factors (performance status > 0 and visceral metastasis) had a median survival time of 10.6 months (95% CI, 6.5 to 14.6). The survival impact of an ECOG performance status of 0 and a lack of visceral metastasis was similar, as suggested from the observed median survival in patients with performance status > 0 and no visceral metastasis (18 months) and in patients with performance status 0 and visceral metastasis (21.1 months) (Table 4).

Table 4. Classification of Patients Based on Performance and Visceral Disease
FactorNo. of patientsNo. of deathsMedian survival (months) (95% CI)OR (95% CI)P
  1. CI: confidence interval; OR: odds ratio; PS: performance status.

PS 0 & no visceral14632.8 (25.3-40.3)3.1 (1.27-7.34)0.0086
PS 0 & visceral231518.0 (9.9-25.9)1.1 (0.58-2.10)0.771
PS 1-2 & no visceral3321.1 (7.1-35.1)0.8 (0.26-2.76)0.79
PS 1-2 & visceral161510.6 (6.5-14.6)0.3 (0.16-0.62)0.001
Number of risk factors    0.0005
 014632.8 (25.3-40.3)2.3 
 1261818.0 (11.5-24.9)(0.91-5.81) 5.6 
 2161510.6 (6.5-14.6)(2.11-19.91) 


In the current study we have characterized pretreatment prognostic factors in patients with metastatic or unresectable urothelial carcinoma treated with a novel regimen of paclitaxel, cisplatin, and gemcitabine (TCG). The current analysis indicates that among eight potential pretreatment prognostic factors analyzed, two of them, performance status > 0 and visceral metastatic disease, are independent adverse prognostic factors for predicting survival in patients treated with this TCG regimen. These two factors have been taken into account for the design of the ongoing Phase III randomized trial (EORTC#30987) with TCG as the experimental arm.

To date, several studies have confirmed the antitumor activity of regimens including taxanes or gemcitabine in bladder carcinoma. Paclitaxel given in combination with cisplatin has been evaluated at least in three studies11–13 that have included a total of 82 patients, with an overall response rate ranging from 63% to 72%. A study including 25 patients has a reported response rate of 60% with docetaxel, another taxane, in combination with cisplatin.24 Median survival was 13 months in one of the cisplatin/paclitaxel studies and 13.6 months in the cisplatin/docetaxel study. The combination of gemcitabine and cisplatin has been evaluated in bladder carcinoma in three studies using different schedules of administration.14–16 In total, 115 patients have been treated with this combination, with an overall response rate ranging from 41% to 71% and with complete response rate ranging from 25 to 35%. The median survival time was consistently reported to be at least 13 months in two of the studies.15, 16 In recent Phase I/II studies using three-drug combinations (triplets), median survival was 18 months with a regimen of ifosfamide, paclitaxel, and cisplatin,25 15 months with a regimen of paclitaxel, carboplatin, and methotrexate,26 and 15.8 months with the current paclitaxel, cisplatin, and gemcitabine regimen. Collectively, it is apparent that there is a wide variation in reported survival rates in bladder carcinoma patients treated in different studies with novel drugs. Although these survival differences may reflect different levels of activity of these individual regimens, taking in consideration their similarity, the likeliest explanation for the observed outcomes may be differences in the patient populations studied.

The impact of pretreatment factors (i.e., patient selection) on outcome has already been shown in series of patients treated with the conventional MVAC regimen. In particular, a low performance status and the presence of visceral disease (lung, liver, or bone metastasis) are the main adverse prognostic factors in patients treated with MVAC.6, 7 Therefore, it could be possible that selection variations among trials with novel regimens might be the underlying reason for the differences in survival reported, as have been discussed above. However, the role of prognostic factors for survival with these novel regimens was yet unknown and a simple extrapolation of prognostic factors from MVAC series might be misleading, since new agents or regimens potentially more active than MVAC might counteract the negative impact of poor performance status and visceral metastasis or change the critical prognostic factors.

In the current series of patients treated with paclitaxel, cisplatin, and gemcitabine, performance status and visceral disease were the two patient characteristics with independent prognostic significance for survival in patients with locally advanced or metastatic TCC. These identified prognostic factors are the same reported for the MVAC series, indicating their clinical importance across different regimens. When the patients were categorized on the basis of having zero, one, or two risk factors, the median overall survival was remarkably similar for each category compared to that reported in the initial MVAC series from investigators at Memorial Sloan-Kettering Cancer Center (MSKCC) in New York;7 i.e., zero risk factors groups had a median overall survival of 32.8 months for TCG and 33.0 for MVAC, single risk factor groups had a median overall survival of 18 months for TCG and 13.4 for MVAC, and double risk factors groups had a median overall survival of 10.6 months for TCG and 9.3 for MVAC. While this historical comparison suggests similar survival among TCG and MVAC by risk category, a well recognized potential disadvantage of the TCG results is its multi-institutional nature (11 centers participated) compared to the MVAC MSKCC series. To explore whether this difference might affect adversely TCG data compared to MVAC data, it would be of interest to compare the survival for risk category of the current TCG trial with survival by risk category in multi-institutional MVAC studies.

There was a possibility that in the current series postchemotherapy surgery and the use of second line chemotherapy might have influenced final patient outcome. For this reason, we analyzed these two variables for possible relationships with patient survival and found that, in univariate analysis, posttherapy surgery was associated with survival while second line chemotherapy was not. However, postchemotherapy surgery lost its statistical significance when included in a multivariate analysis with the pretreatment characteristics that were significantly associated with survival in the univariate analysis; i.e. performance status, visceral disease, and number of disease sites (data not shown).

The current study had a relatively small number of patients, a lack of central pathology review, and a retrospective nature. Since the prognostic factors found, performance status and visceral disease, have been incorporated prospectively in a large Phase III trial, we face a unique opportunity to confirm or refute the validity of the current findings. The current study has been restricted to the analysis of pretreament prognostic factors that are routinely available at most centers treating bladder carcinoma, since our goal was solely to provide stratifying factors for the Phase III trial. However, we plan to study in molecular and pathologic markers of prognosis, such as p53 or mdm2,26, 27 in the current series to provide additional insights into the interplay between biology and chemoresponse of bladder carcinoma.

In summary, the current study has implications for stratification of patients based on performance status and visceral metastasis for Phase III randomized trials, since both MVAC7 and novel regimens share the same independent pretreatment prognostic factors. As a result of the current findings, the upcoming Intergroup Trial (EORTC#30987) randomized study comparing TCG with a regimen of cisplatin and gemcitabine in advanced urothelial tumors has incorporated performance status and visceral metastasis as stratifying variables. Further, the current findings provide useful information for physicians and patients when considering the possible activity of this combination in the treatment of advanced urothelial tumors.