Sequential adjuvant chemotherapy after surgical resection of high-risk urothelial carcinoma

Abstract

BACKGROUND:

Despite definitive surgery, the survival of patients with high-risk urothelial carcinoma (UC) is poor. Adjuvant cisplatin-based chemotherapy may be beneficial, but it is restricted by the need for normal renal function (RF). Sequential administration of adjuvant chemotherapy facilitates drug delivery and improves survival in patients with breast cancer. The objective of this study was to evaluate the feasibility and survival impact of adjuvant, sequential chemotherapy in patients with high-risk UC.

METHODS:

Fifty patients were treated on 2 simultaneous protocols between 1997 and 2004. The patients on Protocol A (normal RF) received doxorubicin and gemcitabine (AG) followed by paclitaxel and cisplatin. The patients on Protocol B (impaired RF) received AG followed by paclitaxel plus carboplatin. Overall survival (OS) and disease-specific survival (DSS) were compared with a group of 203 contemporary control patients who had similar pathology and RF and who underwent surgery alone.

RESULTS:

The median follow-up of protocol patients was 6.5 years (range, 0.9-8.6 years), and 25 patients remained alive. The median follow-up of the control group was 4.7 years (0.0-9.2), and 68 patients remained alive. The median OS for patients on Protocol A was greater than that for controls who had good RF (4.6 years vs 2.5 years; P = .03). The median OS for patients on Protocol B was greater than that for controls who had impaired RF (3.4 years vs 2 years; P = .04). DSS for the protocol and matched control groups was similar (good RF: 4.6 years vs 3 years; P = .24; impaired RF: 3.4 years vs 3.3 years; P = .40).

CONCLUSIONS:

In this nonrandomized study, adjuvant, sequential chemotherapy for patients with high-risk UC did not improve DSS over that observed with surgery alone. Cancer 2009. © 2009 American Cancer Society.

Urothelial cancer (UC) of the bladder is the fifth most common cancer in the United States, and 68,810 new cases are expected in 2008.1 Approximately 33% of patients have tumors that invade the muscularis mucosa, for which the standard surgical treatment is radical cystectomy and bilateral pelvic lymphadenectomy. With surgery alone, there is a high rate of recurrence; and, ultimately 20% to 30% of patients with pathologic T2 (pT2) tumors, 40% to 60% of patients with pT3 tumors, and 70% to 90% of patients with pT4 tumors will develop distant metastases.2 Five-year survival rates after cystectomy of 36% to 48% have been reported, and worse survival has been reported in the range of 25% to 35% for patients with extravesical disease.3-6

High-risk patients frequently receive perioperative chemotherapy that targets micrometastases to reduce the risk of recurrence. Several clinical trials and 3 meta-analyses have demonstrated a survival benefit for neoadjuvant chemotherapy with an approximate 5% absolute improvement in overall survival (OS) at 5 years.7-12 Six studies investigating adjuvant chemotherapy in UC have been published. Two studies demonstrated a survival benefit, and 4 did not. All 6 trials have been criticized for methodological flaws.13-18 Nonetheless, a 25% relative decrease in the risk of death for adjuvant therapy compared with no therapy (hazard ratio [HR], 0.75; 95% confidence interval [CI], 0.60-0.96; P = .02) was suggested in 1 meta-analysis, although that study was limited by size (n = 491) and the poor quality of trial data on which it was based.19 That meta-analysis did not include the largest trial ever performed, which recently reported no benefit with immediate adjuvant chemotherapy for patients with UC.18

Mathematical models long have formed the basis of chemotherapy development. The Skipper-Schabel- Wilcox model introduced in 1871 established combination chemotherapy based on the premise that each dose of chemotherapeutic agent kills a constant fraction of tumor and that multiple drugs are required to kill cells with heterogeneous sensitivity. The “log-kill” theory assumed a constant and exponential growth pattern and proposed that cure could be achieved by exposing sensitive cells to a sufficient duration of treatment. This hypothesis worked for leukemia, because the growth dynamics of the cancer are constant.20

However, many tumors do not grow exponentially and, instead, fit a growth pattern described by Gompertz whereby tumor doubling time increases with tumor size. Norton and Simon expanded on this theory and hypothesized that the fraction of cells killed falls as the number of cells increases.21 They proposed that, between doses of chemotherapy, small tumors grow more quickly, relative to their size, than larger tumors; and the authors suggested that increasing dose density would minimize the opportunity for regrowth between cycles. Sequential administration of chemotherapy achieves this aim.

Only cisplatin-based regimens have demonstrated benefit in the perioperative treatment of UC, a disease with a median age of 68 years at diagnosis. The glomerular filtration rate deteriorates with age, and many patients have comorbid medical illnesses that further compromise renal function (RF). Poor RF limits the administration of cisplatin, and it is not known whether carboplatin is as effective in the adjuvant setting. For metastatic disease, small phase 2 studies evaluating carboplatin regimens have reported similar response rates with less toxicity but with shorter median survival than typically is observed with cisplatin-based regimens.22-24

At Memorial Sloan-Kettering Cancer Center (MSKCC) between 1997 and 2004, we tested the feasibility and outcomes of dose-dense, sequential, adjuvant chemotherapy using 2 prospective phase 2 protocols. Both protocols had similar eligibility criteria with the exception of RF, and RF decided which platinum was used. Herein, we report the toxicity, OS, and disease-specific survival (DSS) of these 2 studies compared with contemporary controls who had similar pathology and RF and who underwent cystectomy and lymphadenectomy but who did not receive perioperative chemotherapy.

MATERIALS AND METHODS

Eligibility Criteria

Our study population included patients with complete resection of locally advanced UC with negative margins, histologically confirmed by the Pathology Service at MSKCC. Pathologic stages included Tany,N+,M0; T3b,N0,M0; T4a,N0,M0 for the bladder. For UC of the renal pelvis or ureter, pathologic stages included Tany,N+,M0; T3,N0,M0; and T4,N0,M0. Patients on Protocol A had normal RF (serum creatinine <1.5 mg/dL or calculated creatinine clearance ≥60 mL/minute using the Cockroft-Gault equation). Patients on Protocol B had impaired RF (serum creatinine >1.5 mg/dL but ≤2.5mg/dL and calculated creatinine clearance <60 mL/minute but >30 mL/minute), or a prior nephrectomy. The control group consisted of similarly staged patients with normal and impaired RF who underwent cystectomy by the same urologists during the same exact time period but who did not receive perioperative chemotherapy (Table 1).

Table 1. Patient Characteristics by Renal Function

Variable	No. of Patients With Good Renal Function (%)		P
	Control Group, n=116	Protocol Group, n=30
SD indicates standard deviation; T0, noninvasive tumor; T1, tumor invades subepithelial connective tissue; T2, tumor invades muscularis propria; T3, tumor invades perivesical tissue; T4, tumor invades other pelvic organs; N0, no lymph involvement; N+, metastasis in a single lymph node ≤2 cm in greatest dimension (N1), or metastasis in a single lymph node ≥2 cm but ≤5 cm in greatest dimension or multiple lymph nodes all ≤5 cm in greatest dimension (N2). * P value was based on the t test. Other P values were based on the Fisher test. † Nx was not included in the test.
Age, y			.02*
Mean±SD	65±11	60±13
Median [range]	67 [34-88]	58 [31-99]
Sex			.35
Women	30 (26)	5 (17)
Men	86 (74)	25 (83)
Tumor classification			.25
T0, T1, T2	14 (12)	6 (20)
T3, T4	102 (88)	24 (80)
Lymph node involvement			.99
N0	55 (48)	14 (48)
N+	59 (52)	15 (52)
Nx	2	1
Variable	No. of Patients With Impaired Renal Function (%)		P
	Control Group, n=87	Protocol Group, n=90
Age, y			<.0001*
Mean±SD	75±9	65±9
Median [range]	77 [47-91]	66 [38-78]
Sex			>.99
Women	20 (23)	5 (25)
Men	67 (77)	15 (75)
Tumor classification			.43
T0, T1, T2	8 (9)	3 (15)
T3, T4	79 (91)	17 (85)
Lymph node involvement			.18
N0	52 (60)	7 (41)
N+	34 (40)	10 (59)
Nx†	1	3

Eligibility criteria included age ≥18 years, Karnofsky performance status (KPS) ≥60%, granulocytes ≥1500 cells/mm³, platelets ≥150,000 cells/mm³, bilirubin <1.5 times normal, alkaline phosphatase <2 times normal, aspartate aminotransferase <2 times normal, and normal cardiac function (New York Heart Association functional Class <III/IV heart disease without serious cardiac arrhythmias). Patients were excluded if they had nontransitional cell histology, uncontrolled infection, evidence of another active cancer, or previous chemotherapy or irradiation to the bladder. Both protocols were approved by the Memorial Hospital Institutional Review Board (IRB), and all patients provided written consent before enrollment. A waiver of approval from the IRB was obtained for analysis of the control group.

Treatment

All patients started treatment with AG. Eighteen of 50 patients received gemcitabine 1000 mg/m² weekly for 6 weeks followed by doxorubicin 60 mg/m² biweekly for 8 weeks. Because of the extensive time required for this initial component of treatment (14 weeks), the schedule was optimized to a 10-week schedule in both protocols by giving both drugs simultaneously with doxorubicin 50 mg/m² and gemcitabine 2000 mg/m² every other week for 5 cycles (AG) with granulocyte-colony–stimulating factor (G-CSF) support after each dose. After 5 cycles of AG, patients on Protocol A received 4 cycles of paclitaxel 150 mg/m² as a 3-hour infusion plus cisplatin either at 1 dose of 60 mg/m² or at a dose of 20 mg/m² daily for 3 days every 14 days. Patients on Protocol B received AG for 5 cycles followed by 12 weekly cycles of paclitaxel 65 mg/m² plus carboplatin (area under the curve, 1.7). G-CSF was used in both arms. Attempts were made to treat all patients within 3 months of surgery. Both studies were pilot trials and were designed to estimate the feasibility, safety, and preliminary efficacy of dose-dense adjuvant chemotherapy. Protocol A was a phase 1/2 study that was planned to accrue up to 55 patients, depending on the number of patients who were treated on the phase 1 portion. This number was estimated to allow a noncomparative estimation of efficacy with 95% CIs of ±20%. Only 30 patients were accrued on this study. Protocol B was a phase 2 study that was planned to accrue up to 45 patients. If neutropenic fever was observed in <5 patients, then we could be 95% confident that the true incidence of a major toxicity was <7%. Only 20 patients were accrued to this study. To enable a clinically meaningful comparison, we combined the 2 protocols and compared those patients with MSKCC patients who had pathology and RF similar to those of the patients who were on the protocol that included surgery but no perioperative chemotherapy during the same period as the ongoing protocols and who were followed expectantly.

Toxicity/Dose Modification

Toxicity was graded in accordance with the Common Terminology Criteria, version 2.0. Dose modifications and delays in treatment were specified for potential toxicities. Dose modifications for AG in both protocols were reported previously.25, 26

Dose modifications for paclitaxel were similar in both protocols. If the granulocyte count was <1500 cells/mm³ or if the platelet count was <100,000 cells/mm³, then chemotherapy was held until counts recovered. The paclitaxel dose was reduced by 20% for grade ≥3 nonhematologic toxicity or febrile neutropenia. Dose adjustments for platinum analogue-specific toxicities were protocol specific. In Protocol A, cisplatin was held for grade 3/4 renal toxicity, and subsequent doses of cisplatin were divided over 3 days with predose mannitol 12.5 g. If RF failed to recover, then no further cisplatin was given. In Protocol B, carboplatin was reduced by 20% for grade ≥3 mucositis, febrile neutropenia, or grade 4 thrombocytopenia. Patients with grade ≥3 neurotoxicity were removed from protocol.

Patient Evaluation

Baseline evaluation of patients included a history and physical examination; complete blood count, clinical chemistry, urinalysis, electrocardiogram, chest x-ray, computed tomography (CT) scan of the abdomen and pelvis, and radionuclide left ventriculogram or echocardiogram and bone scans when indicated.

Clinical follow-up was performed 1 month after the completion of adjuvant therapy, every 3 months for 2 years, every 6 months for 3 additional years, and yearly thereafter. CT scans of the abdomen and pelvis were obtained every 6 months for 2 years after therapy, then yearly for 3 years. Patients in the control group were followed with CT scans every 3 months for the first 18 months, every 6 months for the next 18 months, and yearly thereafter.

Biostatistics

The primary objective of this study was to compare the OS and DSS of protocol patients with those observed in patients who underwent surgery alone during the time of protocol accrual (1997-2004). The patient population was stratified and analyzed according to good RF versus impaired RF. OS and DSS were calculated using the Kaplan-Meier method and were compared using the log-rank test. OS was defined as the time from surgery to the date of death from any cause; and DSS was defined as the time from surgery to the date of death caused by UC, censoring patients who remained alive or patients who died of other/unknown causes. A sensitivity analysis was performed using a competing risk method27 by considering all causes of death, excluding UC as a competing risk, and censoring patients who remained alive at last follow-up. All analyses were done within SAS version 9.1 (SAS Institute, Cary, NC) and R (version 2.5.1; R Foundation for Statistical Computing, Vienna, Austria).

RESULTS

Between August 1997 and July 2004, 20 patients were accrued on Protocol A (9 bladder primaries, 11 upper tract), and 30 patients were accrued on Protocol B (28 bladder primaries, 2 upper tract). The median age of protocol patients was 62 years (range, 31-99 years). Twenty-five patients remained alive after a median follow-up of 6.5 years (95%CI: 5.8-7.1 years). The median interval from the date of surgery to the start of treatment was 2.1 months (95% CI, 0.9-4.6 months). The median OS was 4.5 years (95% CI, 2.6 to not achieved [NA]). There was no difference in OS when the patients who had good RF were compared with the patients who had impaired RF (P = .8).

The MSKCC institutional database identified 1036 patients who underwent cystectomy or nephroureterectomy at our center between May 1997 and July 2004. Two hundred three patients (116 with good RF and 87 with impaired RF) met the pathologic stage inclusion criteria for Protocols A and B but did not receive perioperative chemotherapy. The median age of this control group was 70 years (range, 34-91 years), which was statistically higher than the median age of the protocol patients (P < .0001). Age, however, was not a predictor of OS when it was analyzed as a continuous variable (P = .6) or as a dichotomous variable (age ≥65 years vs <65 years; P = .6). There was no significant difference in stage (P = .4) or lymph node involvement (P = .3) between the control patients and the protocol patients.

Sixty-eight of these patients remained alive at a median follow-up of 4.7 years (95% CI, 4.1 ∼ 5.5 years) (Table 1). There was no statistically significant difference (P = .5) in median OS between control patients who had normal RF (2.5 years; 95%CI, 1.6-3.1 years) and control patients who had impaired RF (2 years; 95%CI, 1.4-2.9 years).

An analysis was performed comparing outcomes among protocol patients and controls stratified by RF. OS for protocol patients was greater than OS for controls (good RF: 4.6 years [95% CI. 2.3 years to NA] vs 2.5 years [95% CI, 1.7-3.3 years]; P = .03; poor RF: 3.4 years [95% CI, 2.6 years to NA] vs 2 years [95% CI, 1.5-3.2 years]; P = .04). DSS among protocol patients and control patients was similar (impaired RF: 3.4 years vs 3.3 years; P = .4; Good RF: 4.6 years vs 3 years; P = .24; Fig. 1). This observation of improved OS without improved DSS may have been affected by missing data. Disease-specific events were available for 100% of protocol patients but for only 73% of pathology-matched controls. Twenty-five patients on protocol died from UC (15 patients with good RF and 10 patients with impaired RF), and there were 99 deaths from UC in the control group. 2

Figure 1. Disease-specific survival is illustrated for patients with good renal function (RF). CI indicates confidence interval; inf, infinity.

Figure 2. Disease-specific survival is illustrated for patients with impaired real function (RF). CI indicates confidence interval; inf, infinity.

Safety and Tolerability

Forty patients (80%) required treatment delay or cancellation. Reasons for those modifications included neutropenia, thrombocytopenia, anemia, infection, diarrhea, abdominal cramping, renal insufficiency, palpitations, carboplatin allergy, mucositis, and scheduling issues. There were 24 treatment-related hospital admissions for 19 patients: The reasons for those admissions were neutropenic fever (5 episodes), non-neutropenic infection (11 episodes), dehydration (3 episodes), rectal bleeding (1 episode), renal impairment (2 episodes), treatment (1 episode), and stomatitis (1 episode). Treatment was stopped prematurely for 1 patient because of a protocol error.

Neutropenia was the most common toxicity (14 episodes of grade 3 neutropenia and 3 episodes grade 4 neutropenia). Seven patients experienced grade 3 anemia, and 2 patients developed grade 3 thrombocytopenia. Nonhematologic grade 3 and 4 toxicities were uncommon and included grade 3 thromboses (4 patients; 2 on each protocol). Severe nausea, vomiting, fatigue, and mucositis all were observed, although infrequently. No patient developed grade 3/4 cardiac toxicity during treatment; after a median 4.6 year follow-up, there was no reported delayed cardiac toxicity. There was no treatment-induced grade 3/4 neuropathy, although 7 patients developed grade 2 neuropathy, and 39 patients developed grade 1 neuropathy.

DISCUSSION

This study failed to demonstrate an improvement in DSS for adjuvant treatment with dose-dense, sequential chemotherapy compared with contemporary patients who had similar pathology and underwent cystectomy and lymphadenectomy alone. OS was improved across both treatment groups. Although this lack of DSS benefit may have been a consequence of deficient data for the surgery-only group, we cannot recommend further pursuit of this approach in the adjuvant setting. For patients with breast cancer, the North American Intergroup confirmed that adjuvant dose-dense chemotherapy was superior to conventional treatment, with observed reductions in the annual odds of disease recurrence by 26% and in the annual odds of death by 31%.28 Dose intensification with high-dose methotrexate, vinblastine, doxorubicin, and cisplatin (MVAC) has been investigated in metastatic UC and, compared with standard MVAC, high-dose treatment resulted in an improvement in OS at a median follow-up of 7.3 years (HR for death, 0.76; 95% CI, 0.58-0.99; 5-year survival rate, 22% vs 14%; P = .042).29 It is possible that this difference in improved survival may be related to more rapid dosing of all 4 drugs, including cisplatin.

Aggressive, sequential chemotherapy in this patient population is associated with substantial toxicity, poor tolerance, and delays in initiation because of postoperative surgical complications. A treatment modification was required in 80% of protocol patients, and a delay in starting chemotherapy up to 4.6 months after surgery was observed. A recent review of 1142 patients who underwent radical cystectomy at MSKCC demonstrated that approximately 30% (347 patients) experienced a complication within 90 days after discharge that was severe enough to interfere with the administration of adjuvant therapy.30 The sequential approach in the current study attempted to deliver optimal doses over a shorter time, resulting in greater “dose density.” This rationale is undermined if surgical complications prevent treatment administration in a timely manner. With regard to neoadjuvant chemotherapy, the US Intergroup study demonstrated that preoperative treatment with an aggressive 4-drug regimen did not impact negatively on planned surgery.31

There is a need for effective, noncisplatin-based treatment for UC; and, to our knowledge, this study is the first trial to compare a carboplatin-based regimen for the adjuvant treatment of UC with surgery alone. Trials of carboplatin-based regimens in patients with advanced UC and normal RF have produced encouraging overall response rates (30%-40%); however, complete response rates and median survival rates are lower than those produced with cisplatin-based regimens.22-24 In the current trial, the carboplatin-based regimen demonstrated a significant improvement in OS over surgery alone. However, it is possible that patients who chose this adjuvant treatment had better outcomes because of unknown contributing factors, such as less medical comorbidities, thus extending OS but not DSS. We believe that a perioperative chemotherapy benefit should be defined by improvements in both DSS and OS and that a standard analysis for such studies should include a competing risk analysis that considers death from another cause as a competing risk.

The current results underscore the idea that a “plateau of efficacy” has been reached in chemotherapy used in the treatment of UC. We combined agents based on their nonoverlapping toxicities, different mechanisms of action, and synergism when administered at maximum doses but failed to demonstrate a definitive survival benefit. Triplet chemotherapy regimens in metastatic UC have been similarly disappointing. A European Organization for Research and Treatment of Cancer multicenter, phase 3 trial of over 600 patients demonstrated no significant difference in either PFS or OS for paclitaxel, cisplatin, and gemcitabine compared with cisplatin and gemcitabine alone. Recent reports of activity for new classes of drugs in metastatic UC offer hope of an alternative therapeutic approach.32, 33 Combinations of traditional chemotherapy drugs with these newer agents may improve outcomes in UC, which has been the case with bevacizumab in colon, lung, and breast cancer.34-36

Our results are limited for several reasons. We chose a cohort of stage-matched, contemporary controls, but a nonrandomized, retrospective comparison inherently is flawed and compares unfavorably with a prospective study. In addition, patients were followed more frequently if they were on protocol, which may have enabled earlier detection of recurrence. The effect of surgical quality on survival in UC has been well documented.37-39 We attempted to control for this by using a control group who underwent surgery performed by the same urologists during the same period. Our analysis did not stratify for the number of lymph nodes resected, because those data were not available in all patients; and, because of the small protocol group of 50 patients stratified into 2 groups based on RF, further subset analysis would be error-prone. The only significant difference between the 2 groups was the younger median age of the protocol group (age 62 years vs age 70 years; P < .0001). Although age was not a significant predictor of survival, it is possible that control patients who died of other causes contributed to a lower survival rate in the control group, resulting in an OS benefit for protocol patients without an improvement in DSS. Disease-specific events were available in 100% of protocol patients compared with only 73% of controls, and cause of death was not known for 27 patients (13%) in the control group. Results could have been affected if these patients died of bladder cancer. Nonetheless, the difference that we report between OS and DSS emphasizes the importance of reporting DSS for all such studies.

In summary, we report that adjuvant treatment with a 4-drug, sequential chemotherapy regimen failed to improve DSS in patients who had high-risk, extravesical UC compared with the DSS observed for contemporary controls who underwent surgery alone. Therapy was poorly tolerated, illustrating the difficulty in delivering intensive chemotherapy in a timely manner to an older patient population in the postoperative setting. The optimal trial design for studying sequential therapy would be a random assignment study, and such a trial was initiated within the US intergroup mechanism. However that trial was closed early because of poor accrual. These study results, plus the closure of the intergroup trial, effectively terminate the investigation of intensive-dose, sequential chemotherapy in UC. Phase 3 trials in Europe continue to study whether adjuvant, conventional-dose chemotherapy confers a survival benefit. Conversely, our institutional investigational approach to perioperative therapy trials has switched to the neoadjuvant setting in which benefit is supported by level 1 evidence.40 Newer drugs incorporated in these trials may achieve benefit that is not observed with combinations of multiple older chemotherapy agents.

Conflict of Interest Disclosures

Supported in part by Bristol Myers Squibb and Eli Lilly.