Adjuvant chemotherapy in muscle-invasive bladder carcinoma

A pooled analysis from phase III studies




The treatment of muscle-invasive bladder carcinoma should include both the eradication of local disease and the elimination of potential micrometastases. To date, the ‘gold standard’ treatment for muscle invasive bladder carcinoma has been recognized to be radical cystectomy. Adjuvant chemotherapy (AC) has the advantage of being administered to patients with known prognostic factors of recurrence. A pooled analysis was used to verify whether AC is able to increase the disease-free survival (DFS) and overall survival (OS) of patients with muscle-invasive bladder carcinoma who had undergone radical cystectomy.


All published randomized Phase III studies were considered eligible and a literature-based pooled analysis was accomplished. DFS and OS were the endpoints. Event-based relative risk ratio (RR) and 95% confidence intervals (95% CI) were derived. Combined effect estimation was computed with a fixed- and a random-effect model. A heterogeneity test was applied as well.


Five trials were collected. All were evaluable for OS (350 patients) and four of these also were evaluable for DFS (273 patients). A significant benefit from AC was noted both in OS (RR, 0.74; 95% CI, 0.62–0.88 [P = 0.001]) and DFS (RR, 0.65; CI 0.54–0.78, [P < 0.001]). No significant heterogeneity was found.


Although the results of the analysis favor the use of AC, larger studies are needed to identify the role of AC in the treatment of muscle-invasive bladder carcinoma. Ongoing trials (European Organization for Research and Treatment of Cancer [EORTC]/Southwest Oncology Group [SWOG] and the Italian Multicentric Study) can help to better identify patients who can benefit from this approach. Cancer 2006. © 2006 American Cancer Society.

Bladder carcinoma represents the fifth most common cancer in the U.S. and it is estimated that the worldwide frequency is more than 300,000 new cases per year.1 Approximately 60% of patients with bladder carcinoma will develop muscle-invasive disease and many patients with superficial tumors will develop invasive disease.2

The treatment of muscle-invasive bladder carcinoma should include: 1) the eradication of local disease; 2) the elimination of potential micrometastases; and 3) maintenance of the best quality of life without compromising survival. Radical cystectomy is considered the cornerstone of the treatment of muscle-invasive bladder carcinoma, but although new surgical techniques3, 4 have reduced the morbidity and mortality related with this procedure, the 5-year survival rates for all stages range from 48–66%.5–7 Thus, the majority of patients with high-grade invasive disease will subsequently die of metastatic disease within 3 years. Neoadjuvant chemotherapy given before local treatment with the aim of reducing primary tumor volume and the control of micrometastatic disease has been widely used in recent decades. Recently, two metaanalysis demonstrated a significant benefit in overall survival (OS) from neoadjuvant cisplatin-based chemotherapy followed by local treatment compared with local treatment alone.8, 9

This encouraging result could favor considering neoadjuvant chemotherapy followed by radical surgery as the new standard treatment of muscle-invasive bladder carcinoma, but some points should be considered: 1) the staging procedures are made on a clinical basis and prognostic factors are unknown until surgery; 2) a large number of patients who do not benefit from neoadjuvant chemotherapy will be given a useless and even toxic treatment, with a delay of definitive treatment; 3) neoadjuvant chemotherapy rarely reaches one of its main objectives, which is the sparing of the bladder; and 4) we do not know how to select patients who can benefit from neoadjuvant chemotherapy.

Adjuvant chemotherapy (AC) given after surgery has the potential of being administered to accurately selected patients based on known prognostic factors for recurrence, without delaying definitive treatment; furthermore, micrometastases are treated at a very low volume.

Logothethis et al.10 first reported that patients with extravesical disease or positive lymph nodes treated with surgery followed by AC had a 2-year disease-free survival (DFS) rate that was higher than a historic control group of patients (87% vs. 37%) treated at the same institution with surgery alone.

Several randomized trials to date have been undertaken to investigate the use of AC.11–15 Three studies found no statistically relevant difference between AC and cystectomy alone, whereas two studies reported a survival benefit of AC. The studies analyzed individually provided insufficient evidence to support the routine use of AC due to small sample sizes, early cessation of patient entry, confusing analysis and terminology, and questionable conclusions.16

In today's evidence-based medicine, one of the strongest contributions to care changes is provided by level-1 recommendations, arising from large randomized clinical trials (RCTs) or metaanalyses.17 A metaanalysis is particularly needed in those clinical situations in which RCTs have small sample sizes, did not reach their endpoints, or when there are conflicting results.18 As is known, metaanalysis based on individual patient data (IPD) in the adjuvant setting is the most powerful method to calculate time to recurrence or death, but this analysis is limited by the difficulty of data collection, especially when trials are too old for data to be available.18

The objective of this pooled analysis was to verify whether AC is able to increase the DFS and OS of patients with muscle-invasive bladder carcinoma who had undergone radical cystectomy.


Our analysis was conducted in four steps18: definition of the outcomes (definition of the question that the analysis was designed to answer); definition of the criteria applied for the selection of the eligible trials; definition of the search strategy; and a detailed description of the statistical methods used. All calculations were independently performed by two investigators.

Outcome Definition

We considered chemotherapy arms as experimental and control arms as standard comparator arms. The analysis was conducted to determine significant differences in the DFS and OS.

Trial Identification Criteria

We collected all prospective and randomized trials published as formal articles in peer-reviewed journals or presented at the American Society of Clinical Oncology (ASCO), European Cancer Conference (ECCO), and European Society for Medical Oncology (ESMO). In these trials, patients with muscle-invasive bladder carcinoma who were not previously treated were randomized to receive either chemotherapy or no chemotherapy after radical cystectomy. Any chemotherapy regimen was considered eligible.

Search Strategy

The deadline for trial collection was December 31, 2004. RCTs were collected through a search of the MEDLINE (PubMed: available at URL: [accessed December 31, 2005]), ASCO (available at URL: [accessed December 31, 2005]), ESMO (available at URL: [accessed December 31, 2005]), and the Federation of European Cancer Societies (FECS; available at URL: [accessed December 31, 2005]) websites. The keywords used for research were: bladder cancer, urothelial, locally advanced, operable, chemotherapy, adjuvant, review, metaanalysis, pooled analysis, randomized, comprehensive review, systematic review. In addition to computer browsing, review and original articles were also scanned in the references chapter to search for missing trials. Furthermore, lectures from major meetings (ASCO, ESMO, and ECCO) having “chemotherapy for bladder carcinoma” or “adjuvant treatment for bladder carcinoma” as topics were checked.

Statistical Analysis

Data extraction methods

The analysis was performed using an event-based approach: a few events (chemotherapy and control) were extracted when available for each outcome (DFS and OS). If not available, we estimated the number of events by assuming an exponential distribution of survival, starting from the median DFS and OS as reported in the trials.19 In the case of missing time to event (median DFS and OS), these were directly extracted from Kaplan–Meier curves shown in the study reports of Parmar et al.20

Statistical method

We calculated variances as indicated for relative risk ratio (RR).21 The log of RR was estimated for each considered endpoint. Using this method, it was possible to apply a fixed-effect and a random-effect model as well according to the Mantel–Haentzel and the inverse variance estimation.19 The 95% confidence intervals (95% CI) were then calculated using critical values of the normal distribution.

The heterogeneity between trials was tested with Q statistics, computing the square distance of each study from the combined effect, and weighting these values with the inverse of variance of each study. The Q statistics were then compared with the chi-square distribution with k-1 degrees of freedom, in which k was the number of studies. The significance of the heterogeneity test suggests selecting the random-effect estimation for a more appropriate evaluation of the results. Absolute benefits for each outcome were calculated according to Parmar et al.22

The results are depicted in both figures as conventional-metaanalysis forest plots. The pooled analysis calculations were accomplished using Comprehensive Meta-Analysis software, v. 1.0.23 (CMA, Biostat, Englewood, NJ).


Six trials matched the entry criteria; one of these23 was excluded from the analysis because chemotherapy was given in both arms. Because some studies were dated, a search for updated articles also was performed.5, 24–26 Information regarding patients, treatments, outcomes, DFS, and OS were extracted from the text or extrapolated from the presented curves.

The trial characteristics are listed in Table 1. A total of 350 patients was included in 5 randomized Phase III trials. The median number of patients per arm was 36 (range, 23–48 patients); the numbers of patients enrolled in chemotherapy and control arms were 167 (median, 35 patients; range, 25–44 patients) and 183 (median, 40 patients; range, 23–48 patients), respectively. Moreover, the number of patients who actually received chemotherapy was 141, which is 84.4% of the total number of patients enrolled in the 5 studies. All RCTs were evaluable for OS analysis; data for the calculation of the DFS were available for 273 patients enrolled in 4 trials.

Table 1. Trials' Characteristics
ReferenceSkinner et al., 199111Stockle et al., 199212Studer et al., 199413Freiha et al., 199614Bono et al., 198915Total
  1. CT: chemotherapy; S: surgery alone; NR: not reached; pN: pathologic node; -: negative; + positive; CISCA: cisplatin, doxorubicin, and cyclophosphamide; MVA(E)C: methotrexate, vinblastine, doxorubicin (epirubicin), and cyclophosphamide; CIS: cisplatin; CIS + MTX: cisplatin plus methotrexate; OS: overall survival; DFS: disease-free survival.

No. of patients44472623374025253548167183
Patients treated33 18 30 25 35 141 
Median age in yrs (range)61 (22–75)62 (30–73)61 (32–67)63 (36–74)64 (54–73)61 (41–65)59 (41–76)64 (49–78)NR   
Male/female34/1035/1222/419/430/727/1323/222/3NR 109/23103/32
Planned cycles4 3 3 4 4   
Patients ≥ 50% CT29 patients 18 patients 26 patients 23 patients 33 patient 129 
OS71%   50%58%  15%57%  54%63 mos  36 mos51%  62%  
 (3 yr)(estimated at 3 yr)(5 yr)(median)(events at 5 yr)  
DFS70%  46%66%  15%NR 37 mos  12 mos51%  56%  
 (3 yr)(estimated at 3 yr) (median)(events at 5 yr)  
Median follow-up in mos168  overall54    5769   overall51    6469   overall  

A significant benefit of AC was seen with regard to OS (RR, 0.74; 95% CI, 0.62–0.88 [P = 0.001]) compared with DFS (RR, 0.65; 95% CI, 0.54–0.78 [P < 0.0001]) (Figs. 1, 2). The heterogeneity test was not significant for either OS (P = 0.75) or DFS (P = 0.10) (Table 2). The 5-year OS and DFS absolute benefit in favor of AC was 11% and 16%, respectively. The number of patients needed to treat to obtain a benefit for one patient were nine patients and six patients, respectively, for OS and DFS.

Figure 1.

Relative risk ratio forest plot for overall survival.

Figure 2.

Relative risk ratio forest plot for disease-free survival.

Table 2. Combined Results
 PatientsRCTsRR (95% CI)Chi-square testHet. (chi-square test2)Absolute benefit.Patients/ 1 benefit
  1. RCTs: randomized clinical trials; RR: relative risk ratio; 95% CI: 95% confidence intervals Het.: heterogeneity test; Patients/1 benefit: patients to treat for 1 benefit; OS: overall survival; DFS: disease-free survival.

OS35050.74 (0.62–0.88)0.0010.8011%9
DFS27340.65 (0.54–0.78)<0.0010.1016%6


The objective of the current systematic analysis was to review whether chemotherapy in addition to radical cystectomy could provide an advantage in terms of DFS and OS in patients with muscle-invasive bladder carcinoma. When analyzed individually, all the trials found in our search provided questionable conclusions, as reported also by Sylvester and Sternberg.16

Given the limitations and drawbacks when adopting the pooled analysis method, in our opinion this approach is relevant, especially in bladder carcinoma. The recent IPD metaanalysis collecting all RCTs exploring the addition of neoadjuvant chemotherapy to radical cystectomy8 reported the same overall results as the subsequent abstracted-data metaanalysis on the same issue.9 Although the latter study collected the Intergroup 0080 Trial data (which were not gathered in the previous IPD metaanalysis), all the explored outcomes provided the same ratios and significance when compared with IPD analysis. Moreover, the difficulties in collecting RCTs (especially for old studies such those for AC) allowed us to use this approach.

The results obtained with our analysis demonstrated statistical significance for both OS (P = 0.001) and DFS (P < 0.001) in favor of radical cystectomy followed by AC when compared with radical cystectomy alone. Although not resulting from IPD analysis, both the lack of heterogeneity and the equivalence between fixed and random effect models, as well as the agreement between the two methods, suggest that AC after radical surgery in muscle-invasive bladder carcinoma is strongly generated.

These results, although interesting, should be evaluated carefully before considering AC as a current management policy in the treatment of patients with muscle-invasive bladder carcinoma. This is because the number of trials, and consequently the number of patients enrolled in them, were too low to sustain this treatment modality as a standard practice, notwithstanding that the type of chemotherapy administered in the trials analyzed in only one case11 can be considered a standard regimen (methotrexate, vinblastine, doxorubicin (epirubicin), and cyclophosphamide [MVA(E)C]) for advanced disease, whereas in the other trials the regimens utilized were considered inferior in terms of response rate and survival in the treatment of patients with advanced and/or metastatic disease.27, 28

However, considering that the presence of micrometastasis is one of the major causes of death in patients with this disease, chemotherapy given in the perioperative setting should nonetheless be considered a fundamental part of the integrated treatment of muscle-invasive bladder carcinoma.

If we determine that radical cystectomy is still the gold standard for local treatment of the disease, the issue becomes whether chemotherapy should be given before or after surgery. The metaanalysis of neoadjuvant chemotherapy demonstrates that there is a beneficial effect with regard to OS for patients receiving chemotherapy, but this effect is restricted to a 5–6.5% increase in OS with a 10–13% reduction in the risk of death in patients receiving cisplatin-combination chemotherapy.8, 9 This means that for many patients neoadjuvant chemotherapy is useless and even toxic, and can lead to a delay in the definitive treatment of local disease. Moreover, to our knowledge, even for patients who derive a beneficial effect from neoadjuvant chemotherapy, there are no data presented to date concerning the possibility of bladder-sparing surgery.

The results achieved with the current analysis of AC demonstrate that chemotherapy given to a subset of patients with known prognostic factors for recurrence can be beneficial with regard to both DFS and OS, thereby reducing the number of patients receiving a useless treatment.

We are aware that metaanalyses performed on abstracted data should be regarded as only a first step toward generating a hypotheses to be tested by IPD metaanalysis,29 but this most likely is the only way to analyze old trials and is a good starting point for new, larger, randomized trials. Nevertheless, the results from a metaanalysis of approximately 400 patients are underpowered to detect an absolute benefit of 11%, which requires that greater than 1000 patients to be reliable.

Another issue also remains to be explored. Although no statistical heterogeneity was found, the most significant one (clinical heterogeneity), which cannot be quantified, varies great across all collected trials. Heterogeneity exists between trials concerning surgical technique, the chemotherapy regimen and agents, and, above all, patient selection (i.e., lymph node status at the time of surgery). Furthermore, we are not currently up to date regarding the errors and problems in the design, accrual, or the eventual early stopping of these trials.

Moreover, old trials cannot be easily controlled and validated for their results, as well as data quality. The quality of every metaanalysis is strongly linked to the quality of the gathered trials.

With this perspective, the results of the current study should not have any direct implication for clinical practice unless they are confirmed by eventual positive data emerging from ongoing randomized clinical trials that are appropriately designed and powered to detect this benefit.

Three large cooperative Phase III studies currently are ongoing. Two of these compare surgery plus AC versus chemotherapy at the time of disease recurrence (Italian Multicentric Study and European Organization for Research and Treatment of Cancer [EORTC]/Southwest Oncology Group [SWOG]); the other study (Spanish Oncologic Genito-Urinary Group [SOGUG]) compares surgery plus AC versus surgery alone. We believe the results from these trials or a metaanalysis will better define the exact role of adjuvant chemotherapy.