Neoadjuvant chemotherapy improves survival of patients with upper tract urothelial carcinoma


  • Arthur Gelmis reviewed the article for readability.



High-grade upper tract urothelial carcinoma (UTUC) is frequently upstaged after surgery and is associated with uniformly poor survival. Neoadjuvant chemotherapy may offer a way to improve clinical outcomes. The authors compared the survival rates of patients with UTUC who received neoadjuvant chemotherapy before surgery with the rates among patients who did not.


A retrospective review was conducted of patients with high-risk UTUC who received neoadjuvant chemotherapy followed by surgery from 2004 to 2008 (study group) compared with a matched cohort who underwent initial surgery from 1993 to 2003 (control group). Fisher exact tests, Wilcoxon rank-sum tests, and Kaplan-Meier methods were used. The log-rank test and Cox proportional-hazards models were used to evaluate the association of the 2 outcomes with patient, treatment, and tumor characteristics in univariate and multivariate models.


Of 112 patients, there were 31 in the study group and 81 in the control group. Patients who received neoadjuvant chemotherapy had improved overall survival (OS) and disease-specific survival (DSS) with a 5-year DSS rate of 90.1% and a 5-year OS rate of 80.2% versus DSS and OS rates of 57.6% for those who underwent initial surgery (P = .0204 and P = .0015, respectively). In multivariate analyses, the neoadjuvant group had a lower risk of mortality (OS: hazard ratio, 0.42 [P = .035]; DSS: hazard ratio, 0.19 [P = .006]).


Neoadjuvant chemotherapy improved the survival of patients with UTUC compared with a matched historic cohort of patients who underwent initial surgery. Patients with high-risk UTUC should be considered for neoadjuvant chemotherapy in view of the limited opportunity to administer effective cisplatin-based chemotherapy after nephroureterectomy. Cancer 2014;120:1794–1799. © 2014 American Cancer Society.


The current standard treatment for upper tract urothelial carcinoma (UTUC) is radical nephroureterectomy. Results from single-institution,[1, 2] multi-institution,[3] and population-based[4-6] studies have consistently demonstrated poor survival for patients with muscle-invasive and nonorgan-confined UTUC and for patients who have lymph node involvement identified after extirpative treatment. Initial support for the use of perioperative chemotherapy for UTUC was provided by studies of bladder cancer indicating improved survival in patients who received neoadjuvant, cisplatin-based chemotherapy.[7, 8] The strongest argument for the use of neoadjuvant, as opposed to adjuvant, chemotherapy in patients with UTUC is based on the high incidence of baseline and subsequent decline in renal function after nephroureterectomy. The use of adjuvant chemotherapy in this population is limited by the significant loss of renal function that occurs after surgery. Studies have demonstrated that >50% of patients who present with UTUC have chronic kidney disease, which worsens after nephroureterectomy and precludes postnephroureterectomy cisplatin-based chemotherapy in the majority of patients; indeed, only 19% to 22% of patients remain eligible for such therapy based on current renal function standards.[9, 10]

We previously reported significant rates of disease downstaging and a 14% complete response rate after neoadjuvant chemotherapy among patients with high-risk UTUC who underwent nephroureterectomy compared with the rates for a matched historic control group of UTUC patients who underwent surgery without receiving neoadjuvant chemotherapy.[11] An earlier study by Igawa et al of 15 patients who had locally advanced UTUC produced a 13% complete remission rate and suggested improved survival.[12] The objective of the current study was to determine whether neoadjuvant chemotherapy confers a demonstrable survival benefit compared with initial surgery in patients with high-risk disease.


On the basis of our finding that patient survival had not improved over several decades[1] and in light of the limitations of postoperative chemotherapy, in 2004, we began uniformly offering neoadjuvant chemotherapy to patients with UTUC who presented with high-risk features at our institution. The criteria used to identify high-risk UTUC patients to be considered for neoadjuvant chemotherapy were a biopsy specimen that exhibited high-grade tumor,[13-15] sessile tumor architecture,[16, 17] and large tumor burden (measurable on axial imaging).[18, 19] This became our standard practice based on the few retrospective studies available and expert opinion based on experience with urothelial cell carcinoma of the bladder. Patients who elect to undergo initial surgery despite the recommendation of neoadjuvant chemotherapy are not explicitly accounted for in the current analysis, but they form a very small proportion of the patients who are offered the neoadjuvant approach.

Thus, the study group comprised patients with UTUC who received neoadjuvant chemotherapy followed by radical nephroureterectomy at our institution from 2004 to 2008. These patients are the same as those in our initial report,[11] with the exception that only patients with clinically negative lymph node (cN0) status were included in the current analysis.

The control group consisted of patients with UTUC who underwent a nephroureterectomy at our institution from 1993 to2003, a period during which nephroureterectomy was offered to >90% of UTUC patients at our institution.[1] For the current retrospective study, we rereviewed the initial biopsy findings and included only those patients who had truly high-grade disease on the basis of the 2004 update to the World Health Organization tumor classification system,[20] and we also ensured that all patients had cN0 status.

For the current study, from the patients' records, we obtained and evaluated multiple clinical and pathologic features, including patient data (age, sex, Eastern Cooperative Oncology Group performance status), tumor data (laterality, radiographic tumor size, prior history of bladder cancer, location of tumor, tumor architecture), treatment (type and courses of neoadjuvant chemotherapy, lymphadenectomy performed), pathology (pathologic classification, pathologic lymph node classification, number of lymph nodes removed; and presence of extranodal extension, lymphovascular invasion, carcinoma in situ, and multifocality), and survival (disease-specific survival [DSS] and overall survival [OS]). Pathologic complete remission was defined as the absence of any identifiable malignancy in all resected specimens. This retrospective study was approved by our institutional review board with waiver of informed consent.

The Fisher exact test and the Wilcoxon rank-sum test were used to compare categorical and continuous patient characteristics, respectively. The Kaplan-Meier method was used to estimate the probability of OS and DSS rates starting from surgery. Patients who were alive at the last follow-up or who were lost-to-follow-up or died for other reasons were censored. The log-rank test and a Cox proportional-hazards model were used to evaluate the association of these 2 time-to-event outcomes with patient characteristics, treatments, and tumor characteristics.

Clinically relevant variables and variables that were significant in univariate analysis were included in the multivariate model. Age, neoadjuvant chemotherapy, and sessile architecture were significant on univariate analysis for OS and/or DSS. The number of lymph nodes removed was added to the model as a surrogate to control differences over time in surgical management and principles of lymphadenectomy for upper tract disease. The mere performance of a lymphadenectomy as well as the total number of lymph nodes removed was not significantly associated with DSS or OS on univariate analysis. In a retrospective study, Roscigno et al recently reported that at least 8 lymph nodes were necessary to consider lymphadenectomy to be sufficient in patients with UTUC[21]; therefore, we used this cutoff value in our analysis as a surrogate to account for differences in surgical technique. Lymphadenectomy in the study group included the paracaval or para-aortic lymph nodes with or without interaortocaval lymph nodes in those who had tumors above the mid-ureter; whereas patients who had distal ureteral tumors underwent pelvic lymphadenectomy.

SAS software 9.3 (SAS Institute Inc., Cary, NC) and S Plus software 8.2 (TIBCO Software Inc., Palo Alto, Calif) were used for statistical analyses. P values < .05 were considered statistically significant.


The study group consisted of 31 patients who received neoadjuvant chemotherapy, and the control group consisted of 81 patients who underwent surgery without receiving neoadjuvant chemotherapy. Table 1 lists the results of baseline, surgical, and tumor characteristics of the 2 groups based on Fisher exact and Wilcoxon rank-sum tests. Differences in laterality seemed spurious, and the lower incidence of lymphadenectomy in the control group reflects changing trends in treatment of the disease. Twenty patients (24.7%) in the control group received adjuvant chemotherapy; whereas, in the study group, none received adjuvant chemotherapy. There was no statistically difference (P = .416) in tumor size (mean ± standard deviation) between 42 of 81 patients who underwent initial surgery (4.1 ± 2.1 cm) and 19 of 31 patients who received neoadjuvant chemotherapy (3.7 ± 1.3 cm).

Table 1. Differences in Patient and Tumor Characteristics Between Patients Who Did and Did Not Receive Neoadjuvant Chemotherapy: Fisher Exact and Wilcoxon Rank-Sum Tests
CharacteristicNo. of Patients (%)Pa
Surgery, n = 81Neoadjuvant Chemotherapy, n = 31
  1. Abbreviations: ECOG, Eastern Cooperative Oncology Group.

  2. a

    P values in boldface indicate a statistically significant difference.

Age: Median [range], y68 [41-85]70 [32-85].9663
Nonwhite6 (7.4)5 (16.1).1739
White75 (92.6)26 (83.9) 
Women33 (40.7)12 (38.7)1
Men48 (59.3)19 (61.3) 
ECOG performance status   
041 (50.6)18 (60).4009
140 (49.4)12 (40) 
Left47 (58)11 (35.5).037
Right34 (42)20 (64.5) 
Renal pelvis47 (58)21 (67.7).3551
Ureter24 (29.6)9 (29) 
Ureteroenteric anastomosis10 (12.3)1 (3.2) 
History of bladder cancer   
No36 (44.4)16 (51.6).531
Yes45 (55.6)15 (48.4) 
No31 (38.3)5 (16.1).0259
Yes50 (61.7)26 (83.9) 
Tumor architecture   
Papillary32 (39.5)11 (35.5).8287
Sessile49 (60.5)20 (64.5) 
Adjuvant chemotherapy   
No61 (75.3)31 (100).0015
Yes20 (24.7)0 (0) 
Lymphovascular invasion   
No38 (46.9)21 (67.7).0584
Yes43 (53.1)10 (32.3) 
Carcinoma in situ   
No33 (40.7)19 (61.3).0592
Yes48 (59.3)12 (38.7) 
No42 (51.9)21 (67.7).1426
Yes39 (48.1)10 (32.3) 

Neoadjuvant therapy consisted of a cisplatin-containing regimen in 21 patients (standard or dose-dense methotrexate-vinblastine-doxorubicin-cisplatin, gemcitabine-cisplatin, or cisplatin-gemcitabine-ifosfamide) or high-dose ifosfamide-doxorubicin-gemcitabine in 3 patients. Kidney-sparing therapy (primarily gemcitabine-paclitaxel-doxorubicin) was received by 7 patients. All patients who were started on neoadjuvant chemotherapy were able to complete a median number of 4 cycles (interquartile range, 4-5 cycles; range, 2-6 cycles) before surgical extirpation. No patient was precluded from surgery because of preoperative chemotherapy.

Significant differences in disease staging were observed between the study population and the control group for individual stages and various stage subgroupings. For example, there were significantly lower rates of muscle-invasive disease (P = .0017) and organ-confined disease (P = .0024) in the study population versus the control group when evaluating only those who had pathologically lymph node-negative (pN0) disease (Table 2). Downstaging remained significant when pN-positive patients were included (P = .0001 and P = .0005, respectively). There was no difference in the rates of pN1/pN2 disease between the control group and the study groups (18.5% and 6.5%, respectively; P = .2218). The 5-year OS and DSS rates were 80.2% and 90.1%, respectively, in the neoadjuvant group versus 57.6% and 57.6%, respectively, in the initial surgery group (Figs. 1, 2).

Table 2. Pathologic Tumor Classification in Patients Without Lymph Node Disease Who Did and Did Not Receive Neoadjuvant Chemotherapy
pT ClassificationNo. of Patients (%)P
Surgery, n = 66Neoadjuvant Chemotherapy, n = 29
  1. Abbreviations: pT classification, pathologic tumor classification; Tis, tumor in situ.

T00 (0)4 (13.8).0011
Ta5 (7.6)4 (13.8) 
Tis6 (9.1)2 (6.9) 
T19 (13.6)9 (31) 
T215 (22.7)6 (20.7) 
T329 (43.9)3 (10.3) 
T42 (3)1 (3.4) 
Noninvasive: T0, Tis, Ta11 (16.7)10 (34.5).0049
Invasive—any: T1-T455 (83.3)19 (65.5) 
Nonmuscle invasive: T0, Tis, Ta, T120 (30.3)19 (65.5).0017
Muscle invasive: T2-T446 (69.7)10 (34.5) 
Organ confined: T0, Tis, Ta, T1-T235 (53)25 (86.2).0024
Nonorgan confined: T3-T431 (47)4 (13.8) 
Figure 1.

Overall survival is illustrated in relation to the receipt of neoadjuvant chemotherapy. Num. indicates number; CI, confidence interval.

Figure 2.

Disease-specific survival (DSS) is illustrated in relation to the receipt of neoadjuvant chemotherapy. Num. indicates number; CI, confidence interval.

Multivariate analyses were performed to assess the effects of individual factors on outcomes. Variables that were significant and clinically relevant in univariate analyses were included in multivariate models, as described above (see Materials and Methods). Cox regression models for OS indicated that neoadjuvant chemotherapy had a significant influence on OS (Table 3). Age, the number of lymph nodes removed, and tumor architecture were not significant in the model. Similar analyses for DSS indicated the significant influence of neoadjuvant chemotherapy and tumor architecture (Table 3).

Table 3. Multivariate Cox Model for 5-Year Overall Survival and Disease-Specific Survival
VariableHR (95% CI)Pa
  1. Abbreviations: CI, confidence interval; HR, hazard ratio.

  2. a

    P values in boldface indicate a statistically significant difference.

Overall survival  
Age1.02 (0.998-1.05).075
Neoadjuvant chemotherapy0.42 (0.19-0.94).035
≥8 Lymph nodes removed0.75 (0.40-1.40).370
Sessile tumor architecture1.16 (0.69-1.96).580
Disease-specific survival  
Age1.01 (0.98-1.04).560
Neoadjuvant chemotherapy0.19 (0.06-0.61).006
≥8 Lymph nodes removed0.54 (0.24-1.23).140
Sessile tumor architecture2.77 (1.30-5.89).008


Neoadjuvant chemotherapy was associated with improved OS and DSS compared with a matched historic cohort of patients who underwent initial surgery. Our results validate the initial observations reported by Igawa et al[12] and confirm the validity of using pathologic outcomes as a reasonable surrogate for outcomes in patients with UTUC. Prior studies demonstrated that the majority of patients with UTUC present with chronic kidney disease, and an even greater proportion cannot receive effective postoperative chemotherapy when adverse pathologic features are present.[9, 10] This finding may explain the conflicting and largely inconclusive reports from previous studies on the utility of adjuvant therapy for UTUC patients.[22-27]

Despite significant improvements in imaging and technologic advances in ureteroscopy (both presumably enabling earlier disease detection and treatment), not only have survival rates not improved for patients with UTUC,[1] but they may be worsening.[28] These developments indicate that the treatment paradigm for UTUC may need to shift from reflexive initial surgery to more accurate, thoughtful risk stratification with the consideration of neoadjuvant chemotherapy for patients who are classified as high risk.

Accurate clinical risk stratification becomes essential to avoid overtreatment and to identify the patients who are most likely to benefit from neoadjuvant chemotherapy. To aid with clinical risk stratification, it has been demonstrated in 2 preoperative nomograms that using a combination of various clinically available factors—such as biopsy grade, tumor architecture, results of selective cytology, and imaging findings like hydronephrosis—can provide independent prognostic value.[29, 30] These tools can help make the selection of the most appropriate treatment for a patient more systematic and more accurate than has historically been possible.

The limitations of the current study included the retrospective nature of the analysis. We performed pathologic reanalysis of any equivocal biopsy results and excluded those that indicated tumors that were not high grade. The survival rates for our control population were similar to the survival rates in other published studies of patients with UTUC, supporting the valid use of this group as an historic, matched control group. In a study of 1363 patients with UTUC who underwent radical nephroureterectomy, Margulis et al reported that patients with high-grade disease had a 5-year survival rate of 57.2%, which is very similar to our findings.[3] Likewise, in a study of 252 patients with UTUC who were treated surgically, Hall et al reported DSS rates from 72.6% to 40.5% for patients who had T2 through T4 disease.[2] Those data support the validity of our control population. In contrast to pathologic outcomes reported in bladder cancer neoadjuvant trials, documented downstaging for each individual patient with UTUC is not possible given the inaccuracy of initial clinical staging. It is believed that clinical stage has a misclassification bias of approximately 45%, and current expert opinion as well as many retrospective studies have demonstrated that tumor grade is a stronger predictor of high-risk disease with a strong association to advanced pathologic stage, recurrence, and outcome.[1, 3] Therefore, it is difficult to control for true clinical stage in this disease, which adds to the limitations of our findings and conclusions. In addition, a small number of patients did have a history of bladder cancer, and although this was not significantly different between groups, it may confound findings and outcomes related to DSS. Thus, matched historic cohorts like the group used in this study provide the best available data for assessing patient outcomes.

Nevertheless, only a prospective and, ideally, randomized study can definitively validate these findings. The design of such trials is made complicated by variations in the chemotherapy regimens needed, as observed in our study. Given the relatively advanced age of patients with UTUC and the high rates of comorbidity (including baseline renal dysfunction and cardiac disease), the current dogma of devising narrow inclusion criteria for the sake of study population homogeneity would render the recruitment of patients with this rare disease even more difficult, and the subsequent results would be inapplicable to a large proportion of patients. It is on this basis that a call has been made to consider in future trials more practical designs that allow for the diverse comorbidities frequently observed in patients with UTUC. In the meantime, the results of the current study provide a strong foundation for urologists and medical oncologists who seek to improve the outcomes of patients with UTUC to consider applying accurate clinical risk stratification and to offer neoadjuvant chemotherapy to patients with high-risk features.


Neoadjuvant chemotherapy in patients with high-risk UTUC resulted in significantly higher survival rates than surgery without neoadjuvant chemotherapy in a matched historic group.


This research was supported in part by the National Institutes of Health through The University of Texas MD Anderson's Cancer Center Support Grant CA016672.


Dr. Kamat reports being a lecturer and participant at meetings sponsored by Photocure, Cubist, Sanofi, and Taris; and he reports grants from Abbott, Cubist, and FDK.