Presented at the 2008 Annual Meeting of the American Society of Clinical Oncology; May 30 to June 3, 2008; Chicago, IL.
Neoadjuvant chemotherapy improves the survival of patients with high-risk urothelial cancer. However, the lack of curative alternatives to cisplatin-based chemotherapy is limiting for patients with neuropathy or hearing loss. Sequential chemotherapy also has not been well studied in the neoadjuvant setting. The authors explored sequential neoadjuvant ifosfamide-based chemotherapy in a patient cohort at high risk of noncurative cystectomy.
Patients with muscle-invasive cancer and lymphovascular invasion, hydronephrosis, clinical T3b and T4a (cT3b-4a) disease (defined as a 3-dimensional mass on examination under anesthetic or invasion into local organs), micropapillary tumors, or upper tract disease received 3 cycles of combined ifosfamide, doxorubicin, and gemcitabine followed by 4 cycles of combined cisplatin, gemcitabine, and ifosfamide. The primary endpoint was downstaging to pT1N0M0 disease or lower.
At a median follow-up of 85.3 months, the 5-year overall survival (OS) and disease-specific survival (DSS) rates for all 65 patients were 63% and 68%, respectively (95% confidence interval: 5-year OS rate, 0.52%-0.76%; 5-year DSS rate, 0.58%-0.81%). Pathologic downstaging to pT1N0 disease or lower occurred in 50% of patients who underwent cystectomy and in 60% of patients who underwent nephroureterectomy and was correlated with the 5-year OS rate (pT1N0 disease or lower, 87%; pT2-pT3aN0 disease, 67%; and pT3b disease or higher or lymph node-negative disease, 27%; P ≤ .001 for pT1 or lower vs pT2 or higher). Variant histology was associated with an inferior 5-year DSS rate (50% vs 83% in pure transitional cell carcinoma; P = .02). The most frequent grade 3 toxicities were infection (38%), febrile neutropenia (22%), and mucositis (18%). There were 3 grade 4 toxicities (myocardial infarction, thrombocytopenia, and vomiting) and 1 grade 5 toxicity in a patient who refused antibiotics for pneumonia.
Neoadjuvant chemotherapy is the standard for treating surgically resectable, invasive urothelial cancer1, 2 However, neoadjuvant chemotherapy remains largely underused because of concern that the risks, especially the toxic effects of chemotherapy, may outweigh the benefits of treatment. At a previous plenary session of the American Society of Clinical Oncology, the merits of neoadjuvant chemotherapy were debated.3, 4 Much work is needed to improve survival and reduce toxicity for our patients with bladder cancer.
Sequential chemotherapy, as one potential method of improving survival, has not yet been explored in the neoadjuvant setting. Clinical trials in the metastatic setting have not observed any definitive benefit and have been limited by their fixed schedules.5 Using a sequence incorporating a for-cause determination, in which chemotherapy that is working is continued longer and is switched early when there is a lack of response, may provide additional benefit.6 By using a for-cause, sequential approach in the metastatic setting, we have observed a potential improvement in surgical consolidation of lymph node metastases.7
On the other side of the neoadjuvant debate is the toxicity of therapy. With the understanding that many patients are not candidates for cisplatin-based chemotherapy as a result of pre-existing peripheral neuropathy and/or hearing loss, we developed chemotherapy regimens using the alkylating agent ifosfamide as the backbone. Preclinical data suggested that there was enhanced DNA damage with combinations that incorporated a nucleoside (gemcitabine) with an alkylating agent (ifosfamide or cisplatin).8 Two combinations arose from this process, 1 that maximized the alkylating agent and minimized the nucleoside (combined ifosfamide, doxorubicin, and gemcitabine [IAG]),9 and the other that maximized the nucleoside agent gemcitabine with much lower doses of alkylating agents (combined cisplatin, gemcitabine, and ifosfamide [CGI]).10 By using this combination sequentially, we hoped to overcome potential chemotherapy-resistant cells, with the objective of improving response and enhancing the long-term survival for these patients. In addition, because this sequence used a much lower cumulative dose of cisplatin, we hoped to avoid peripheral neuropathy and provide an alternative active regimen for patients with pre-existing peripheral neuropathy or baseline hearing loss.
Previous work from our group also suggested that clinical and pathologic features predict for a high likelihood of extra-organ extension with pathologic upstaging in up to 86% of patients.11 This suggests the existence of a group of patients at particularly high risk from their cancer who would be candidates for aggressive sequential chemotherapy. Here, we the report final results from this phase 2 clinical trial of sequential neoadjuvant chemotherapy in patients with high-risk, surgically resectable urothelial cancer.
MATERIALS AND METHODS
From August 2001 to April 2006, 65 patients enrolled onto this M. D. Anderson Cancer Center Institutional Review Board-approved clinical trial. All patients had surgically resectable urothelial carcinoma and were selected because they had a high risk of undergoing noncurative resection by virtue of the following features: a 3-dimensional mass on examination under anesthesia performed after thorough transurethral resection (TUR); tumor invasion into the prostatic stroma, vagina, or uterus (clinical T4a [cT4a] disease); lymphovascular invasion; hydronephrosis; micropapillary histology; or high-grade upper tract cancer of the ureter or renal pelvis. Patients who were not conventionally considered resectable with curative intent (those with tumors fixed to the pelvic sidewall [cT4b disease] or with any evidence of lymph node involvement or distant metastases) were not eligible for this clinical trial. No prior systemic chemotherapy was allowed; however, patients may have received any prior intravesical therapy.
Patients had adequate physiologic reserve, with a Zubrod performance status ≤2, or 3 if of recent onset and caused entirely by the cancer and not by a comorbid medical condition. Adequate bone marrow reserve was defined as an absolute neutrophil count >1800/μL at baseline and a platelet count >150,000. Adequate liver function required a transaminase ≤3 times the upper limit of normal with a conjugated bilirubin ≤1.5 mg/dL or total bilirubin ≤2.5 mg/dL. Patients were required to have a creatinine clearance either measured or calculated with the Cockcroft-Gault equation of ≥45 mL/minute. Patients who had an abnormal electrocardiogram or a history of heart disease were required to have an ejection faction ≥50% to be eligible for this trial.
Sequential chemotherapy consisted of 3 cycles of IAG followed by 4 cycles of CGI. If an inadequate response was observed during the repeat 6 week cystoscopy, then patients were switched after 2 cycles IAG to 6 cycles of CGI.
Initially, patients received the following doses for IAG: ifosfamide 2000 mg/m2 daily with 2-mercaptoethane sulfonate sodium (MeSNa) on days 1 through 4, doxorubicin 50 mg/m2 on day 3, and gemcitabine at 200 mg/m2 on days 2 and 4. However, 4 of the first 10 patients experienced toxicity, especially myelosuppression, necessitating a dose reduction. The starting doses were modified in subsequent patients as follows: ifosfamide 1500 mg/m2 infused over 3 hours daily on days 1 through 4; MeSNa 225 mg/m2 infused over 15 minutes at hours zero, 3, 7, and 11 daily on days 1 through 4; doxorubicin 45 mg/m2 over 15 minutes as a peripheral intravenous infusion or over 12 to 18 hours infused through a central line on day 3 only; and gemcitabine 150 mg/m2 infused over 30 minutes on days 2 and 4 only.
While they were receiving IAG, patients were hydrated aggressively, frequently using a sodium acetate infusion. Patients received methylene blue with or without albumin in the event of neurologic toxicity, and the ifosfamide dose was reduced in subsequent cycles if neurologic symptoms recurred despite treatment with methylene blue. Serum creatinine was monitored on a daily basis, and treatment was withheld if renal insufficiency developed. Nearly all patients with hydronephrosis had a nephrostomy tube placed before chemotherapy, even in the setting of normal serum creatinine, to help maintain renal function. This inpatient regimen was repeated at 3-week intervals with growth factor support.
Patients received the following doses for CGI: gemcitabine 900 mg/m2 over 90 minutes on day 1, ifosfamide 1000 mg/m2 over 60 minutes on day 1, followed by cisplatin 50 mg/m2 in 500 mL normal saline with 12.5 g Mannitol over 3 hours on day 1. MeSNa was not required with this low dose of ifosfamide. The cisplatin typically was followed by a mannitol infusion consisting of 5% dextrose quarter-normal saline with mannitol 40 g/L for at least 2 or 3 L. This inpatient or outpatient regimen was repeated at 2-week intervals with growth factor support as necessary.
Within 6 weeks of study entry, patients had either a computed tomography or magnetic resonance imaging scans of the abdomen and pelvis, a chest x-ray or computed tomography scan of the chest, and an electrocardiogram within 6 months. A baseline bone scan was only required in the presence of bone pain or elevated calcium or alkaline phosphatase levels. Before registration, all patients were required to undergo examination by a member of the Department of Urology, to undergo cystoscopy with a thorough TUR of bladder tumor, and to be examined under anesthesia to evaluate for the presence of a 3-dimensional mass and for bladder mobility. After 6 weeks of treatment, patients underwent repeat cystoscopy and TUR to evaluate for response. All patients were recommended for cystectomy regardless of the degree of response observed in the bladder.
A major response was defined as downstaging to pathologic T1N0 (pT1N0) disease or lower at the time of cystectomy. Responses of lesser magnitude were scored as treatment failures. Progression was counted as any objective sign of recurrent or progressive disease. If symptoms predated histologic or radiographic confirmation of recurrence, then the date of progression was taken as the date of onset of symptoms.
The primary endpoint of this trial was downstaging to pT1N0 disease or lower after chemotherapy, an endpoint well described as a surrogate for eventual cure. The Simon 2-stage minimax design was used.12 Assuming an a priori response rate of 35% (P0 = .35) with an expected response rate of 50% (P1 = .50) with neoadjuvant therapy, this study required 49 patients to provide 80% power with α = .10. In the second stage, the study required at least 22 responses in 49 patients to be considered successful for further study. An additional 16 patients were accrued to account for patients who refused surgery, those who progressed to unresectable tumors, or those who received treatment for upper-tract urothelial cancer. All patients who received any chemotherapy were included in the final analysis. The probability of early termination under P0 was P0 = .46, and the expected sample size was 40.8.
Kaplan-Meier curves were used to estimate the distribution of overall survival (OS) and disease-specific survival (DSS). The 5-year survival probability and its corresponding 95% confidence interval (CI) were provided for each study cohort. Nonparametric log-rank tests were used to compare survival curves according to pathologic stage or histology. A multivariate Cox proportional hazards model was used to explore the effects of tumor characteristics on survival. The proportional hazards assumption was verified using the test statistic based on weighted residuals. All statistical tests were 2-sided at a significance level of P = .05.
Baseline characteristics for all 65 patients are listed in Table 1. Please note that all patients had at least 1 of the high-risk features noted above (see Materials and Methods). All patients had a histologic diagnosis of urothelial carcinoma; and 43% of patients had another variant histology, such as micropapillary, squamous, adenocarcinoma, or sarcomatoid changes associated with their urothelial tumor. In all instances, except in the setting of micropapillary tumors, the variant histology must have been present in <50% of the tumor specimen to be considered eligible for this trial.
Overall, pathologic downstaging to pT1 disease or lower occurred in 30 of 60 patients (50%) who had primary bladder tumors (Table 2) and in 3 of 5 patients who had primary tumors of the renal pelvis, with pT0 rates of 43% and 0%, respectively. Downstaging was correlated with 5-year OS (log-rank P ≤ .001): Patients who were downstaged to pT1N0 disease or lower, to pT2-T3aN0 disease, and to pT3b disease and lower or lymph node-negative disease had 5-year OS rates of 87%, 67%, and 27%, respectively (Fig. 1). For patients who completed 3 cycles of IAG before CGI, the rates for those with pT0N0 disease and pT1N0 disease or lower were 60% and 65%, respectively. For those who received ≤2 cycles of IAG and switched early to CGI because of a lack of response or toxicity, the rates for pT0N0 disease and for pT1N0 disease or lower were 17% and 26%, respectively. This likely reflects the poor biology observed in nonresponding patients, who typically had aggressive tumors and poor outcomes. There were only 2 patients who failed to respond to IAG and had pT0 disease after CGI. The absence of residual disease on repeat, postchemotherapy cystoscopy is not a guarantee of achieving pT0N0 status, as observed in other studies. In our experience, even when there was no visible tumor remaining on repeat cystoscopy, 42% of patients had residual tumor present at cystectomy.
Table 2. Clinical Versus Pathologic Stage at Cystectomy
Two patients with cT4aN0 disease and 2 patients with cT2N0 disease refused surgery, and there was 1 toxic death (a patient with cT3bN0 disease).
For all enrolled patients, the combined 5-year OS and DSS rates were 63% and 68%, respectively (95% CI: 5-year OS rate, 52%-76%; 5-year DSS rate, 58%-81%; median follow-up, 85.3 months) (Fig. 2). In the patients who underwent lymph node dissection for primary bladder tumors, the median number of lymph nodes removed was 12 (range, 3-39 lymph nodes). Three patients did not undergo lymph node dissection because of previous abdominal surgery or previous lymph node dissection (ie, previous prostate cancer). All patients who progressed did so within 18 months of their surgery except for 1 late progression in the lung at 53 months. There was some concern that the lung finding reflected a second primary cancer; however, because the scans were not available for review at our institution, we counted this patient as having progressive urothelial cancer.
The presence of variant histology was associated with an inferior 5-year DSS rate of 50% (95% CI, 35%-72%) compared with 83% (95% CI, 71%-96%) for pure transitional cell carcinoma (TCC) (log-rank P = .02) (Fig. 3). Although there was a similar trend of histology for OS, the difference in OS was not statistically significant (log-rank P = .1451). The presence of micropapillary histology was associated with a 5-year OS and DSS rate of 54% (95% CI, 33%-89%).
In the multivariate Cox model, whether a patient with primary bladder cancer (n = 60) was downstaged to pT1N0M0 disease or lower after chemotherapy was a statistically significant predictor (relative risk, 0.13; 95% CI, 0.04-0.40 for downstaging vs no downstaging; P < .001), and pure TCC histology also was a significant factor affecting DSS (relative risk, 0.35; 95% CI, 0.14-0.89 for TCC vs mixed histology; P = .03); whereas other tumor characteristics (clinical stage at diagnosis, hydronephrosis, micropapillary tumor, lymphovascular invasion, and performance status) were not statistically significant in the multivariate models. The proportional hazards model assumption was satisfied for both risk factors. In the multivariate Cox model, tumor histology (TCC vs mixed) did not have a statistically significant association with the OS endpoint, whereas patient downstaging was always associated significantly with the OS endpoint. Similar findings were observed when patients with upper-tract urothelial cancer were included in the analysis (n = 65), in which a final pathology of pT1N0M0 disease or lower remained statistically significant (P < .001), whereas pure TCC histology was marginally significant (P = .07).
In total, the 65 patients received 166 cycles of IAG and 196 cycles of CGI. The only treatment-related death was from pneumonia during neutropenia, which occurred in Cycle 1. This patient refused additional therapy, including antibiotics, and quickly died of the pneumonia. Three patients experienced grade 4 toxicities (myocardial infarction, platelet transfusion, and vomiting). The most frequent grade 3 toxicities (Table 3) were infection (38%), febrile neutropenia (22%), mucositis 18%, and platelet transfusion (12%). It is noteworthy that no patients developed peripheral neuropathy as a result of this chemotherapy. Three patients discontinued IAG after their first cycle, including 1 who experienced a fatal toxicity as described earlier, a second patient who developed renal insufficiency, and a third patient who had rapid tumor progression. Eleven patients required dose reductions of IAG, whereas 10 patients required dose reductions of CGI. Forty patients completed all 3 cycles of IAG. Of the 25 patients who transitioned early to CGI, 13 transitions were because of the lack of response, 3 transitions were by patient choice, and 9 transitions were because of toxicity or physician concern about the patient's ability to complete all 3 cycles. Seven of these 9 patients received an ifosfamide dose of 2000 mg/m2. After the first 10 patients were treated, the ifosfamide dose was decreased to 1500 mg/m2, which was much more tolerable.
Table 3. Chemotherapy-Related Toxicity
No. of Patients (%)
Abbreviations: PRCB, packed erythrocytes.
Neutropenic fever/pneumonia developed in 1 patient who refused additional treatment, including antibiotics.
Additional grade 3 toxicities that occurred in 1 patient each were arthralgia, dyspnea, rash, hallucination, thrombosis/embolus, and vasovagal reaction.
The median hospital stay for the patients who underwent cystectomy was 8 days (range, 4-38 days) (Table 4,). There were no perioperative deaths. Six patients received 5 or more units of packed erythrocytes postoperatively; however, most of those patients were anemic as a result of their neoadjuvant chemotherapy. Two patients underwent repeat surgery within the perioperative period, including 1 surgery for a bowel obstruction that required lysis of adhesions with an area of bowel torsion and 1 surgery for a patient who required embolectomy of an arterial thrombus in their aortofemoral graft. A full listing of perioperative complications within a 30-day period are listed in Table 4. Ten patients did not undergo surgery: Six patients declined surgery despite our firm recommendation, whereas 4 patients progressed (2 in the bone and 2 elsewhere in the bladder and lung). Unfortunately, all patients who declined surgery had a relapse of their cancer.
Table 4. Surgical Parameters and Perioperative Morbidity
Postsurgery, N = 55
No. of Patients (%)
Abbreviations: PRCBs, packed erythrocytes.
One patient with aortobifemoral grafts required embolectomy.
Ileal complications were those that required a delay in discharge >10 days after surgery.
Small bowel obstructions were adhesions and bowel torsion that required surgery.
Overall, 69% of patients in our study were cured of their high-risk bladder cancer. It is noteworthy that our results confirmed that neoadjuvant chemotherapy resulting in no residual muscle-invasive cancer in the cystectomy specimen is highly associated with OS and DSS. Our current results are similar to the findings from our previous randomized trial of neoadjuvant versus adjuvant methotrexate, vinblastine, doxorubicin, and cisplatin (M-VAC) in a similar high-risk cohort11 and to the findings from the Southwest Oncology Group Intergroup Study,1 which was a retrospective study that demonstrated a survival benefit for patients with pT1N0 disease or lower.13 The consistency of these findings firmly establishes that a pathologic response is a useful, early surrogate for achieving an eventual cure.
In the current trial, we used sequential chemotherapy in the hopes of enhancing pathologic response and overall cure. Compared with other trials in the metastatic setting in which fixed sequences were used,5 in the current trial, we used a for-cause approach for switching to the second sequence.6 Patients underwent repeat cystoscopy after the first 2 cycles of chemotherapy and, if they responded, then received additional IAG; nonresponders were switched earlier to CGI. However, the similar pathologic downstaging rates and survival in previous neoadjuvant trials suggested that there was no additional benefit from this sequential approach. It is possible that patients who did not respond to chemotherapy had such poor biology that they would not have responded to any chemotherapy. Only 2 of our patients who were switched early for failing to respond to IAG had pT0 disease after they received CGI. It is also possible that the combination of CGI, which uses lower doses of cisplatin and ifosfamide, may not be sufficient to overcome resistance to the initial chemotherapy. Therefore, sequencing with a high-dose ifosfamide/low-dose gemcitabine regimen and a low-dose ifosfamide/high-dose gemcitabine regimen may not have been as complementary as we had originally anticipated.
Although IAG was associated with toxicity, it is noteworthy that there was no peripheral neuropathy. Despite the toxicity, 100% of patients received at least 1 cycle of IAG, 93% received at least 2 cycles, and 61% received 3 cycles. Although the sequential therapy we investigated does not appear to be suitable for use outside the setting of dedicated, multispecialty care, IAG may be an attractive option for fit patients with pre-existing peripheral neuropathy or hearing loss who are at risk for worsening of these conditions with the receipt of cisplatin-based chemotherapy. However, our experience suggests that, to tolerate high-dose ifosfamide in this elderly patient group, patients must have adequate renal function (defined as a creatinine clearance of approximately ≥50 mL/minute), and nephrostomy tubes must be used in the setting of obstructed kidneys. We also observed that this regimen is difficult for most patients in their 80s to tolerate, and we would not recommend its use in this age group.
We used clinical criteria to select for patients with a higher risk of having extra-organ extension or lymph node involvement at cystectomy. In a previous clinical trial in which patients were randomized either to undergo initial cystectomy or to receive neoadjuvant chemotherapy, the presence of these high-risk features was associated with clinical up-staging in 86% of patients who underwent initial surgery.11 Patients who had muscle-invasive tumors only (without high-risk features) were not eligible for this trial; and, at our institution, these patients traditionally have undergone initial cystectomy followed by the receipt of adjuvant chemotherapy in the presence of extra-organ extension or lymph node involvement. A retrospective review of our experience with this strategy suggests that we can achieve similar long-term outcomes, sparing many patients from the toxicity of systemic chemotherapy (article in preparation). This is especially important when avoiding toxicity of chemotherapy in a population of patients that frequently includes elderly individuals with comorbid medical conditions.
However, the application of our high-risk criteria may be difficult to apply in the community setting. For example, distinguishing between lymphovascular invasion and a retraction artifact in a TUR specimen requires an experienced pathologist. Examination under anesthesia to determine whether there is tumor beyond the wall of the bladder requires a urologist skilled in this examination and a thorough TUR. The treatment we describe is a truly multidisciplinary collaboration between urologists, pathologists, and genitourinary medical oncologists that may be difficult to achieve outside the context of a tertiary center.
Although the current results may be slightly better than historic expectation, it is clear that the cohort we studied continues to have a disappointing prognosis despite very aggressive application of chemotherapy and surgery. Current treatment fails for approximately 33% of patients with high-risk bladder cancer, and those who have persistent extravesical or lymph node disease despite receiving primary chemotherapy have an especially poor prognosis. The ability to identify patients who are unlikely to benefit from primary chemotherapy would allow us to explore alternative strategies and is an important research objective. One such alternative, based on our previous finding that vascular endothelial growth factor expression is a powerful predictor of outcome in this cohort,14 is the combination of bevacizumab with neoadjuvant chemotherapy. This is the current strategy under investigation in our center.
In conclusion, similar response and survival were observed with the sequential IAG/CGI combination in our patients as compared with clinical trials with neoadjuvant M-VAC in a similar historical cohort. The toxicity profile of the sequential combination provides a useful alternative in the setting of peripheral neuropathy or hearing loss. However, the use of high-dose ifosfamide can be difficult in this patient population, requiring dedicated toxicity management, and it may be more suitable to a tertiary care facility. The presence of advanced disease despite preoperative chemotherapy continues to indicate a poor prognosis, and the ability to select these patients before chemotherapy would provide a cohort for initial surgery or chemotherapy combinations to overcome their resistance pathways. These results reinforce previous experience suggesting that pathologic downstaging to <pT1N0 or lower is a useful surrogate for achieving an eventual cure.
This work was supported by funding from the M. D. Anderson Cancer Center Charlene Curley Fund.