Int J Clin Oncol DOI 10.1007/s10147-014-0695-1
ORIGINAL ARTICLE
Preoperative risk stratification for cancer-specific survival of patients with upper urinary tract urothelial carcinoma treated by nephroureterectomy Kazutoshi Fujita • Motohide Uemura • Yoshiyuki Yamamoto • Go Tanigawa Wataru Nakata • Mototaka Sato • Akira Nagahara • Hiroshi Kiuchi • Yasutomo Nakai • Kiyomi Matsumiya • Seiji Yamaguchi • Norio Nonomura
•
Received: 31 January 2014 / Accepted: 31 March 2014 Ó Japan Society of Clinical Oncology 2014
Abstract Background This study aimed to identify preoperative parameters for predicting cancer-specific survival (CSS) in patients with upper urinary tract urothelial carcinoma (UTUC) who have undergone radical nephroureterectomy (RNU). Methods The preoperative clinical and laboratory records of 357 UTUC patients who underwent RNU at three different institutions were retrospectively reviewed (256, training set; 101, test set). Univariate and multivariate analyses were performed on the training set data to identify preoperative prognostic factors, using which a risk stratification model was developed. The model was validated using test set data. Results In univariate analysis, clinical T stage classification and preoperative concentrations of hemoglobin, C-reactive protein, sodium, and albumin showed significant association with CSS. Multivariate analysis showed that low preoperative sodium and hemoglobin concentrations were significantly associated with a poor prognosis. A risk stratification model was developed using the preoperative sodium (\141 mEq/L) and hemoglobin concentrations (below normal). Three subgroups were formed depending
K. Fujita (&) M. Uemura W. Nakata M. Sato A. Nagahara H. Kiuchi Y. Nakai N. Nonomura Department of Urology, Osaka University Graduate School of Medicine, 2-2 Yamada-oka, Suita, Osaka 565-0871, Japan e-mail: [email protected] Y. Yamamoto K. Matsumiya Department of Urology, Osaka Police Hospital, Osaka, Japan G. Tanigawa S. Yamaguchi Department of Urology, Osaka General Medical Center, Osaka, Japan
on the presence of no (favorable group), one (intermediate), or two (poor) prognostic factors, and the 5-year CSS estimates were found to be 96.5, 75.5, and 47.0 %, respectively (P \ 0.01). The risk model was significantly associated with the adverse pathological findings of stage pT3 or more and lymphovascular invasion (P = 0.005). Conclusion We identified low preoperative sodium and hemoglobin concentrations as prognostic factors for patients with UTUC treated with RNU. Our risk stratification model may help physicians design a therapeutic strategy. Keywords Prognosis Upper urinary tract urothelial carcinoma Nephroureterectomy Sodium Hemoglobin
Introduction Radical nephroureterectomy (RNU) with excision of the bladder cuff and resection of the retroperitoneal lymph node is the gold standard for treatment of non-metastatic upper urinary tract urothelial carcinoma (UTUC). However, a significant proportion of patients who have undergone RNU die from disease recurrence. Thus, RNU alone is not adequate for patients with high-risk UTUC, and it is reasonable to consider perioperative chemotherapy to improve patient survival [1]. Neoadjuvant chemotherapy is considered preferable for patients with UTUC over adjuvant chemotherapy, because RNU causes loss of renal function, whereby renal function is inadequate for cisplatin-based chemotherapy [2]. Neoadjuvant chemotherapy has been proven effective for invasive urothelial carcinoma of the bladder, and it is therefore expected to be effective for UTUC [3]. However, accurate preoperative staging is more difficult in UTUC than in bladder cancer, in which
123
Int J Clin Oncol
clinical staging can be accurately determined using magnetic resonance imaging (MRI) and transurethral resection of a bladder tumor. Thus, preoperative parameters to predict poor prognosis must be identified in order to prevent overtreatment of patients with low-risk UTUC. Several such parameters associated with poor prognosis have been reported, but many include pathological findings, such as CpT3 staging or lymphovascular invasion [4]. Recently, preoperative risk stratification on the basis of the clinical T stage, voided urine cytology, and white blood cell (WBC) count has been proposed as a method of predicting survival after RNU [5]. We previously reported that the preoperative sodium concentration was associated with poor prognosis in combination with postoperative pathological parameters [6]. The purpose of the present study is to identify preoperative parameters that predict survival in patients with UTUC who have undergone RNU.
classification; and follow-up data. Clinical T classification was determined by enhanced CT. Laboratory data, including the serum WBC count and platelet counts and levels of hemoglobin, C-reactive protein (CRP), lactate dehydrogenase (LDH), sodium, modified calcium, creatinine, total protein, albumin, alkaline phosphatase, aspartate transaminase (AST), alanine transaminase, and total bilirubin were determined from the serum chemistry tests conducted at admission for surgery. Patients were generally followed up every 3 months during years 0–2 postoperatively, every 6 months during years 2–5, and every 6–12 months thereafter. Tumor recurrence was defined as the development of recurrence at RNU site, lymph node metastasis, and/or visceral metastasis. The present study was approved by the Osaka University Graduate School of Medicine Institutional Review Board, the Osaka General Medical Center Institutional Review Board, and the Osaka Police Hospital Institutional Review Board, and written informed consent was obtained from all patients.
Materials and methods Statistical analysis Patients We retrospectively analyzed the medical records of 357 Japanese men who had undergone RNU between 1998 and 2012 at the Osaka University Hospital, Osaka General Medical Center, and Osaka Police Hospital (256 cases in Osaka University Hospital and Osaka General Medical Center for the training set; 101 cases in Osaka Police Hospital for the test set). The training set data were used for development of a risk stratification model, while the test set data were used for model validation. Seven patients who had distant metastasis and had undergone RNU for palliative treatment were excluded (five from the training set and two from the test set). Twenty-nine patients who had radiographic lymph node involvement were also subsequently excluded (17 from the training set; 12 from the test set). Four patients in the training set who were receiving hemodialysis were excluded because of abnormal laboratory data and 10 patients whose laboratory data were not found in their medical records were excluded (four from the training set, and six from the test set). Thus, the remaining 226 patients in the training set and 80 patients in the test set were examined. In the training set, 140 (62 %) patients underwent laparoscopic RNU, 86 (38 %) underwent open RNU, and 164 (72 %) underwent concomitant retroperitoneal lymphadenectomy. The following clinical data and radiological findings obtained by intravenous pyelography, retrograde pyelography, or computed tomography (CT) were obtained from medical records: age; sex; history of previous bladder tumor; results of voided urinary cytology; presence of hydronephrosis; clinical tumor, lymph node, metastasis (TNM)
123
The results are presented as median values (range). Univariate and multivariate analyses were carried out using the Cox proportional hazards method with stepwise forward selection to predict recurrence-free survival (RFS) and cancer-specific survival (CSS). The variables entered into the model for RFS analysis were patient age; gender; clinical T classification; and preoperative hemoglobin, sodium, and albumin levels. Those entered into the model for CSS analysis were patient age; gender; clinical T classification; and preoperative hemoglobin, CRP, sodium, and albumin levels. The proportional hazards assumptions for the Cox regression model fit were tested and confirmed. The association of serum sodium or hemoglobin levels with CSS was examined using time-dependent Cox models with adjustment for potential confounders. The Kaplan– Meier estimate of survival curve and log-rank test were used to explore the association between the parameters and patient survival. The patients were stratified into favorable (0 point), intermediate (1 point), and poor risk groups (2 points) according to the number of positive prognostic factors (low serum sodium concentration and low hemoglobin concentration). The cutoff level of sodium concentration was set at 131 mEq/L, and those of hemoglobin were set at 13.8 mg/dL for men and 12.0 mg/dL for women. External validation of the risk stratification model was performed separately for 80 patients with UTUC who underwent RNU at the Osaka Police Hospital. All probability (P) values were two-sided, with statistical significance being set at P \ 0.05. All statistical analyses were performed using SPSS version 11.0.1 (SPSS, Chicago, IL, USA), GraphPad Prism 5 (GraphPad Software, La Jolla,
Int J Clin Oncol
CA, USA), and R version 3.0.1 with the Smooth HR package (Artur Agostinho Arau´jo and Luı´s Meira-Machado, 2013); smoothHR: smooth hazard ratio curves taking a reference value. R package version 1.0 URL http://CRAN. R-project.org/package=smoothHR.
Table 1 Clinical and pathological characteristics of patients who underwent nephroureterectomy for upper tract urothelial carcinoma Training set n
226
Age (years) [median (range)]
70 (37–90)
Test set 80 70 (50–91)
Gender, n (%)
Results
Male
153 (68)
68 (85)
73 (32)
12 (15)
108 (48) 110 (49)
45 (56) 30 (38)
8 (3)
5 (6)
No
177 (78)
63 (79)
Yes
49 (22)
17 (21)
Negative
119 (53)
35 (44)
Positive
107 (47)
45 (56)
No
159 (70)
44 (55)
Yes
67 (30)
32 (40)
0 (0)
4 (5)
Female
Table 1 shows the clinical and pathological characteristics of the patients. The median age was 70 years (range 37–90), and 155 of the 226 patients (67.9 %) were men. The median follow-up time was 41 months (range 1–164). Of the 226 patients, 50 (22.1 %) experienced disease recurrence and 43 (19.0 %) died from UTUC during follow-up. The overall 2- and 5-year RFS estimates were 82.4 % [95 % confidence interval (CI) 76.4–87.0] and 75.5 % (95 % CI 68.4–81.1), respectively, while the overall 2- and 5-year CSS estimates were 89.0 % (95 % CI 83.7–92.6) and 77.7 % (95 % CI 70.3–83.4), respectively. Univariate and multivariate analyses were performed to identify preoperative predictors of disease recurrence and cancer-specific mortality (Table 2). In the univariate analysis, age, clinical T stage classification, preoperative hemoglobin concentration, preoperative sodium concentration, and preoperative albumin concentration were found to be significantly associated with disease recurrence (P \ 0.05). Multivariate analysis with these variables and gender showed that patient age and serum sodium concentration were significantly associated with RFS (P \ 0.05). Additionally, in the univariate analysis, clinical T stage classification and the preoperative levels of hemoglobin, CRP, sodium, and albumin were found to be significantly associated with cancer-specific mortality (P \ 0.05). Furthermore, multivariate analysis with these variables, age, and gender showed that the preoperative sodium and hemoglobin levels were significantly associated with CSS (P \ 0.05). The association of serum sodium and hemoglobin levels with cancer-specific mortality is shown in Fig. 1. In the training cohort, the preoperative serum sodium concentration ranged from 133 to 147 mEq/L. It was below normal (i.e., \136 mEq/L) in only three cases (1.2 %). Lower preoperative serum sodium levels showed a significant association with higher mortality even after multivariate adjustment for age, sex, and hemoglobin level. The threshold for the serum sodium was set at 141 mEq/L, which is the median of the logarithm of the hazard ratio. Patients with serum sodium levels of \141 mEq/L had an unadjusted cancer-specific mortality hazard ratio of 3.51 (95 % CI 1.93–6.38) (P \ 0.001) when compared with those whose serum sodium levels were C141 mEq/L. Lower preoperative hemoglobin concentration also showed
Tumor location, n (%) Renal pelvis Ureter Multiple Previous bladder cancer, n (%)
Urinary cytology, n (%)
Hydronephrosis, n (%)
Unknown Clinical T stage, n (%) \T3
205 (91)
70 (88)
CT3 Neoadjuvant chemotherapy
21 (9) 8 (4)
10 (12) 0 (0)
Adjuvant chemotherapy
44 (19)
29 (36)
Ta
56 (25)
20 (25)
Tis
15 (7)
4 (5)
T1
42 (18)
9 (11)
T2
40 (17)
12 (15)
T3
66 (29)
33 (41)
T4
7 (3)
2 (3)
155 (69)
79 (99)
9 (4)
1 (1)
62 (27)
0 (0)
Pathological T classification, n (%)
pN stage, n (%) Negative Positive Unknown Tumor grade, n (%) G1
17 (8)
8 (10)
G2 G3
105 (46) 102 (45)
43 (53) 29 (36)
2 (1)
1 (1)
Absent
173 (77)
47 (59))
Present
45 (19)
31 (39)
8 (4)
2 (2)
Unknown Lymphovascular invasion, n (%)
Unknown
an association with higher mortality after adjustment for age, sex, and sodium level. The hemoglobin level of the median logarithm of hazard ratio was 12.7 mg/dL.
123
Int J Clin Oncol Table 2 Univariate and multivariate analyses predicting disease recurrence and cancerspecific mortality in 139 patients who underwent nephroureterectomy for upper tract urothelial carcinoma
Univariate
Multivariate
HR (95 % CI)
P value
HR (95 % CI)
P value
Age, years
1.03 (1.00–1.07)
0.035
1.04 (1.00–1.08)
0.018
Gender (female vs. male)
1.15 (0.64–2.07)
0.61
–
–
Urine cytology (positive vs. negative)
0.92 (0.58–1.48)
0.75
–
–
–
–
0.83 (0.74–0.94)
0.003
–
–
Recurrence
Tumor location (ureter vs. pelvis)
1.57 (0.88–2.78)
0.12
(multiple vs. pelvis)
2.25 (0.54–9.33)
0.26
Previous bladder cancer
1.14 (0.55–2.36)
0.71
Clinical T classification (CcT3 vs. BcT2)
2.41 (1.13–5.17)
0.023
Hydronephrosis
1.49 (0.83–2.67)
0.17
White blood cells (per 1 9 103/lL)
1.06 (0.937–1.21)
0.31
Hemoglobin (per 1 g/dL) Platelets (per 1 9 104/lL)
0.84 (0.72–0.98) 1.00 (0.99–1.00)
0.028 0.30
CRP (per 1 mg/dL)
1.10 (0.99–1.23)
0.073
Sodium (per 1 mEq/L)
0.84 (0.75–0.93)
0.002
Corrected calcium (per 1 mg/dL)
1.14 (0.51–2.56)
0.74
Creatinine (per 1 mg/dL)
1.29 (0.71–2.33)
0.39
Lactate dehydrogenase (per 1 IU/L)
0.99 (0.99–1.00)
0.31
Total protein (per 1 g/dL)
0.64 (0.37–1.11)
0.11
Albumin (per 1 g/dL)
0.38 (0.17–0.87)
0.021
AST (per 1 IU/L)
0.97 (0.94–1.01)
0.18
ALT (per 1 IU/L)
0.96 (0.93–1.00)
0.055
Alkaline phosphatase (per 1 IU/L)
1.00 (0.99–1.00)
0.93
Total bilirubin (per 1 mg/dL)
0.29 (0.06–1.23)
0.095
Age, years
1.02 (0.98–1.06)
0.18
–
–
Gender (female vs. male)
0.94 (0.50–1.76)
0.87
–
–
Urine cytology (positive vs. negative) Tumor location (ureter vs. pelvis)
1.06 (0.64–0.17) 1.48 (0.81–2.73)
0.81 0.19
(multiple vs. pelvis)
1.48 (0.20–10.8)
0.69
Previous bladder cancer
1.04 (0.46–2.36)
0.90
Clinical T classification (CcT3 vs. BcT2)
2.62 (1.16–5.91)
0.020
–
–
0.82 (0.70–0.96)
0.019
Cancer-specific mortality
AST aspartate aminotransferase, ALT alanine aminotransferase
123
Hydronephrosis
1.44 (0.77–2.70)
0.25
White blood cells (per 1 9 103/lL)
1.06 (0.92–1.22)
0.40
Hemoglobin (per 1 g/dL)
0.83 (0.71–0.98)
0.029
Platelets (per 1 9 104/lL)
1.00 (0.99–1.00)
0.47
CRP (per 1 mg/dL)
1.99 (1.05–3.79)
0.035
–
–
Sodium (per 1 mEq/L)
0.83 (0.74–0.93)
0.002
0.85 (0.76–0.95)
0.004
Corrected calcium (per 1 mg/dL)
0.87 (0.38–2.014)
0.53
–
–
Creatinine (per 1 mg/dL)
1.35 (0.70–2.57)
0.36
Lactate dehydrogenase (per 1 IU/L)
0.99 (0.99–1.00)
0.85
Total protein (per 1 g/dL)
0.63 (0.35–1.11)
0.13
Albumin (per 1 g/dL) AST (per 1 IU/L)
0.29 (0.12–0.69) 0.97 (0.94–1.01)
0.005 0.23
ALT (per 1 IU/L)
0.97 (0.94–1.00)
0.091
Alkaline phosphatase (per 1 IU/L)
1.00 (0.99–1.00)
0.88
Total bilirubin (per 1 mg/dL)
0.39 (0.08–1.80)
0.22
Int J Clin Oncol
Fig. 1 Association of cancer-specific mortality hazard ratio with preoperative serum sodium concentration (a) and hemoglobin concentration (b) in 226 patients with UTUC who underwent RNU.
Fig. 2 Kaplan–Meyer estimates for each risk group [no (dotted curve), one (dashed curve), or two risk factors (solid curve)] in the training set a (log-rank test: P \ 0.0001) and the test set b (log-rank test: P \ 0.05)
The serum sodium or hemoglobin concentration shown corresponds to the median of the logarithm of the hazard ratio
A risk model was constructed using the preoperative sodium (\141 mEq/L) and hemoglobin level (below normal) obtained from the multivariate analysis used to predict CSS. Patients were stratified into favorable, intermediate, and poor risk groups according to the presence of no, one, or two prognostic factors, respectively. Figure 2a shows the Kaplan–Meier estimates for CSS by risk stratification. The CSS estimates in 5 years were 96.5 % (95 % CI 87.0–98.0) for the favorable risk group, 75.5 % (95 % CI 64.1–83.7) for the intermediate risk group, and 47.0 % (95 % CI 27.1–64.7) for the poor risk group [log-rank test; intermediate vs. favorable: HR 5.41 (95 % CI 1.59–7.61); poor vs. intermediate: HR 2.98 (95 % CI 1.99–9.01)]. Adverse pathological findings of CpT3 staging or lymphovascular invasion were found in 12 of 60 patients (20.0 %) in the favorable risk group, 44 of 118 patients (37.2 %) in the intermediate risk group, and 27 of 48 patients (56.2 %) in the poor risk group (Cochran–Armitage test for trend, P = 0.0001). Multivariate analysis with the logistic regression model adjusted by age, gender, and clinical T classification showed that risk classification with preoperative sodium and hemoglobin levels was significantly associated with pathological findings of CpT3 or lymphovascular invasion (P = 0.005). To validate the risk stratification, 80 patients with UTUC who underwent RNU at Osaka Police Hospital were examined as the test set. The median age was 70 years (range 50–91), and 68 of the 80 patients (83.9 %) were men. The median follow-up time was 32 months (range 1–149). Of the 80 patients, 21 (26.2 %) experienced
123
Int J Clin Oncol
disease recurrence and 12 (15.0 %) died from UTUC during follow-up. Twelve patients were stratified to the favorable risk group, with a 5-year CSS estimate of 100 %. Forty patients were stratified to the intermediate risk group, with a 5-year CSS estimate of 81.4 % (95 % CI 59.6–92.1). Lastly, 27 patients were stratified to the poor risk group, with a 5-year CSS estimate of 67.4 % (95 % CI 39.0–84.7). Figure 2b shows the Kaplan–Meier estimates for CSS by risk classification (log-rank test for trend, P = 0.036).
Discussion It has been reported that, generally, invasive UTUC has a poor prognosis, and approximately 40 % die because of recurrence after RNU [1]. Thus, the demand for preoperative prediction of patient survival, which will aid in the selection of candidates for neoadjuvant chemotherapy, is high. Many postoperative parameters have been reported as prognostic factors, but they could not be used for the selection of patients for neoadjuvant chemotherapy. Neoadjuvant chemotherapy is preferred to adjuvant chemotherapy because patients lose renal function after RNU, making it difficult for them to receive the cisplatin-based chemotherapy. The criteria for patient selection for neoadjuvant chemotherapy are yet to be determined, but it has been recommended that patients with invasive UTUC should receive neoadjuvant chemotherapy [1]. Currently, radiological examination is the only method of preoperatively predicting the pathological stage. CT has been used for preoperative staging of UTUC. However, its accuracy for predicting the pathological TNM stage was reported to be only 59 %; furthermore, under-staging was reported in 16 % cases and non-detection in 24 % [7]. In the present study, among the 23 patients who were preoperatively diagnosed with stage cT3 disease or more, 15 patients (65.2 %) had shown stage pT3 or more on pathological examination, but 8 patients (34.8 %) were preoperatively over-staged. In contrast, 58 (26.1 %) of 203 patients who were not preoperatively diagnosed with stage cT3 disease or more showed stage pT3 or more on pathological examination. Because of these CT inaccuracies for clinical staging, the CT findings did not remain significant after the multivariate analysis including other parameters. Hydronephrosis among the preoperative CT findings was also reported to be associated with a poor prognosis [8, 9], but it was not observed in our cohort. In the multivariate analysis in the present study, preoperative sodium and hemoglobin concentrations in their continuous form were found to be prognostic parameters for patients with non-metastatic UTUC. The preoperative serum sodium concentration ranged from 133 to 147 mEq/L in the test cohort; it was below the normal range (i.e., \136 mEq/L) in only three
123
cases (1.2 %), and the cutoff level of sodium concentration based on the normal range could not be used. The serum sodium concentration in the extracellular fluid volume is strictly regulated, although it does fluctuate from day to day. On analyzing the association between the mortality hazard ratio and continuous sodium concentration, we found an inverse S-shaped association. In other words, a low serum sodium concentration was associated with a high hazard ratio. The sodium concentration which corresponded to the median mortality hazard ratio was 141 mEq/L. The equivalent value for hemoglobin was 12.7 mg/dL. We set the cutoff level of hemoglobin as the lower limit of the normal range, which differs between men and women (men 13.8 mg/dL; women 12 mg/dL). Risk stratification with these two preoperative markers could predict the CSS of patients with non-metastatic UTUC and was associated with the pathological findings of stage pT3 or more and lymphovascular invasion. Recently, Sakano et al. [5] reported preoperative risk stratification with clinical T stage, voided urine cytology, and WBC count, but the preoperative sodium concentration was not included in their analysis, which may explain the difference in the risk stratification models between their study and ours. Since the 5-year estimates of cancer-specific mortality in the favorable risk group in the present study were only 3.6 % in the training set and 0 % in the test set, we could conclude that patients in this group do not require neoadjuvant chemotherapy. The mechanism underlying the anemia and low sodium concentration observed in UTUC patients with a poor prognosis remains to be elucidated. Hyponatremia has been reported to be associated with poor prognosis in several cancers. More than 40 % of patients with small cell lung carcinoma had hyponatremia (\135 mEq/L), which was associated with poorer prognosis. Hyponatremia in patients with small cell lung carcinoma has been explained to be caused by a syndrome of inappropriate secretion of antidiuretic hormone. In localized renal cell carcinoma, patients with lower preoperative sodium concentration (\139 mEq/L) had significantly poorer prognosis than patients with higher preoperative sodium concentration (C139 mEq/L) [10]. The authors suggested that the renal tumors might be inducing a disturbance of the renin– angiotensin–aldosterone axis, which controls the sodium and the extracellular fluids. Hyponatremia has been also reported be a prognostic factor in metastatic renal cell carcinoma [11, 12] and hepatocellular carcinoma [13]. Interleukin-6 (IL-6), one of the key inducers of cancerrelated inflammation, can cause hyponatremia via vasopressin secretion [14]. It can also cause anemia in cancer patients through both complex physiological mechanisms as well as genetic regulatory aspects of erythropoiesis. Hematuria is also likely to be the cause of anemia in
Int J Clin Oncol
patients with UTUC [15]. A previous univariate analysis showed that anemia was associated with the prognosis of UTUC patients [16]. CRP, another preoperative prognostic marker [17], has also shown an association with survival in the univariate analysis, although not in the multivariate analysis, in the present study. Additionally, CRP production from hepatocytes is stimulated by IL-6. Several studies have reported that elevated IL-6 levels are associated with elevated CRP levels in various cancers [18]. It is possible that the preoperative prognostic markers of low serum sodium concentration, anemia, and elevated CRP level indicate IL-6 production by the tumor and inflammatory infiltrate cells, and the IL-6-related pathway may be a new therapeutic target for aggressive UTUC. Indeed, STAT3 expression is a prognostic factor of bladder cancer, and STAT3 is an important agent in the IL-6 pathway [19]. High urinary IL-6 levels were found to be a poor prognostic factor in patients with urothelial carcinoma of the bladder [20]. Thus, future studies should examine whether serum or urinary IL-6 levels in patients with UTUC are associated with survival, like the sodium and hemoglobin concentrations are. Despite its novel findings, our study does have some limitations. First, the preoperative CT findings did not remain significant in the multivariate analysis for CSS. Several patients had renal insufficiency preoperatively and were not suited to undergoing contrast-enhanced examination. This renal insufficiency for using contrast material may have led to the low accuracy of CT for preoperative staging. Preoperative staging by MRI might be more accurate, and the results might be used as a prognostic factor. Although this study proposes a risk stratification model for selection of patients for neoadjuvant chemotherapy, only 8 of the 226 patients (4 %) in our cohort received neoadjuvant chemotherapy. Therefore, it was difficult to assess the association between the effect of neoadjuvant chemotherapy on survival and risk stratification using preoperative sodium and hemoglobin concentrations. Thus, no concrete recommendations can be made from our findings in terms of the selection of candidates for neoadjuvant chemotherapy. Subsequent studies should examine whether the prognosis of patients with intermediate or poor risk based on our risk stratification model can be improved by neoadjuvant chemotherapy. Lastly, although we validated the risk stratification model using a test set from a different institution, this was nonetheless a retrospective study including only Japanese patients. Further large-scale studies should be performed to further validate the proposed model. In conclusion, the present multi-institutional study showed that low preoperative sodium and hemoglobin concentrations, which can be easily and routinely measured, were prognostic factors in patients with UTUC undergoing RNU. We also proposed a risk stratification model for predicting CSS based
on these two preoperative variables, which were also associated with adverse pathological findings. Our risk stratification model may help physicians design a therapeutic strategy. Conflict of interest Norio Nonomura received lecture fees from Novartis Pharma and Pfizer Pharmaceutical.
References 1. Audenet F, Yates DR, Cussenot O et al (2013) The role of chemotherapy in the treatment of urothelial cell carcinoma of the upper urinary tract (UUT-UCC). Urol Oncol 31:407–413 2. Kaag MG, O’Malley RL, O’Malley P et al (2010) Changes in renal function following nephroureterectomy may affect the use of perioperative chemotherapy. Eur Urol 58:581–587 3. Babjuk M, Compe E, Zigeuner R et al (2013) European guidelines on upper tract urothelial carcinomas: 2013 update. Eur Urol 63:1059–1071 4. Margulis V, Shariat SF, Matin SF et al (2009) Outcomes of radical nephroureterectomy: a series from the upper tract urothelial carcinoma collaboration. Cancer 115:1224–1233 5. Sakano S, Matsuyama H, Kamiryo Y et al (2013) Risk group stratification based on preoperative factors to predict survival after nephroureterectomy in patients with upper urinary tract urothelial carcinoma. Ann Surg Oncol 20:4389–4396 6. Fujita K, Tanigawa G, Imamura R et al (2013) Preoperative serum sodium is associated with cancer-specific survival in patients with upper urinary tract urothelial carcinoma treated by nephroureterectomy. Int J Urol 20:594–601 7. Scolieri MJ, Paik ML, Brown SL et al (2000) Limitations of computed tomography in the preoperative staging of upper tract urothelial carcinoma. Urology 56:930–934 8. Ng CK, Shariat SF, Lucas SM et al (2011) Does the presence of hydronephrosis on preoperative axial CT imaging predict worse outcomes for patients undergoing nephroureterectomy for uppertract urothelial carcinoma? Urol Oncol 29:27–32 9. Cho KS, Hong SJ, Cho NH et al (2007) Grade of hydronephrosis and tumor diameter as preoperative prognostic factors in ureteral transitional cell carcinoma. Urology 70:662–666 10. Vasudev NS, Brown JE, Brown SR et al (2008) Prognostic factors in renal cell carcinoma: association of preoperative sodium concentration with survival. Clin Cancer Res 14:1775–1781 11. Jeppesen AN, Jensen HK, Donskov F et al (2010) Hyponatremia as a prognostic and predictive factor in metastatic renal cell carcinoma. Br J Cancer 102:867–872 12. Kawashima A, Tsujimura A, Takayama H et al (2012) Impact of hyponatremia on survival of patients with metastatic renal cell carcinoma treated with molecular targeted therapy. Int J Urol 19:1050–1057 13. Huo T-I, Lin H-C, Hsia C-Y et al (2008) The MELD-Na is an independent short- and long-term prognostic predictor for hepatocellular carcinoma: a prospective survey. Dig Liver Dis 40:882–889 14. Swart RM, Hoorn EJ, Betjes MG et al (2011) Hyponatremia and inflammation: the emerging role of interleukin-6 in osmoregulation. Nephron Physiol 118:45–51 15. Dicato M, Plawny L, Diederich M (2010) Anemia in cancer. Ann Oncol 21(Suppl 7):vii167–vii172 16. Akdogan B, Dogan HS, Eskicorapci SY et al (2006) Prognostic significance of bladder tumor history and tumor location in upper tract transitional cell carcinoma. J Urol 176:48–52. doi:10.1016/ S0022-5347(06)00511-8
123
Int J Clin Oncol 17. Obata J, Kikuchi E, Tanaka N et al (2013) C-reactive protein: a biomarker of survival in patients with localized upper tract urothelial carcinoma treated with radical nephroureterectomy. Urol Oncol 31:1725–1730 18. Guthrie GJK, Roxburgh CSD, Horgan PG et al (2013) Does interleukin-6 link explain the link between tumour necrosis, local and systemic inflammatory responses and outcome in patients with colorectal cancer? Cancer Treat Rev 39:89–96
123
19. Mitra AP, Pagliarulo V, Yang D et al (2013) Generation of a concise gene panel for outcome prediction in urinary bladder cancer. J Clin Oncol 27:3929–3937 20. Seguchi T, Yokokawa K, Sugao H et al (1992) Interleukin-6 activity in urine and serum in patients with bladder carcinoma. J Urol 148:791–794
ORIGINAL ARTICLE
Preoperative risk stratification for cancer-specific survival of patients with upper urinary tract urothelial carcinoma treated by nephroureterectomy Kazutoshi Fujita • Motohide Uemura • Yoshiyuki Yamamoto • Go Tanigawa Wataru Nakata • Mototaka Sato • Akira Nagahara • Hiroshi Kiuchi • Yasutomo Nakai • Kiyomi Matsumiya • Seiji Yamaguchi • Norio Nonomura
•
Received: 31 January 2014 / Accepted: 31 March 2014 Ó Japan Society of Clinical Oncology 2014
Abstract Background This study aimed to identify preoperative parameters for predicting cancer-specific survival (CSS) in patients with upper urinary tract urothelial carcinoma (UTUC) who have undergone radical nephroureterectomy (RNU). Methods The preoperative clinical and laboratory records of 357 UTUC patients who underwent RNU at three different institutions were retrospectively reviewed (256, training set; 101, test set). Univariate and multivariate analyses were performed on the training set data to identify preoperative prognostic factors, using which a risk stratification model was developed. The model was validated using test set data. Results In univariate analysis, clinical T stage classification and preoperative concentrations of hemoglobin, C-reactive protein, sodium, and albumin showed significant association with CSS. Multivariate analysis showed that low preoperative sodium and hemoglobin concentrations were significantly associated with a poor prognosis. A risk stratification model was developed using the preoperative sodium (\141 mEq/L) and hemoglobin concentrations (below normal). Three subgroups were formed depending
K. Fujita (&) M. Uemura W. Nakata M. Sato A. Nagahara H. Kiuchi Y. Nakai N. Nonomura Department of Urology, Osaka University Graduate School of Medicine, 2-2 Yamada-oka, Suita, Osaka 565-0871, Japan e-mail: [email protected] Y. Yamamoto K. Matsumiya Department of Urology, Osaka Police Hospital, Osaka, Japan G. Tanigawa S. Yamaguchi Department of Urology, Osaka General Medical Center, Osaka, Japan
on the presence of no (favorable group), one (intermediate), or two (poor) prognostic factors, and the 5-year CSS estimates were found to be 96.5, 75.5, and 47.0 %, respectively (P \ 0.01). The risk model was significantly associated with the adverse pathological findings of stage pT3 or more and lymphovascular invasion (P = 0.005). Conclusion We identified low preoperative sodium and hemoglobin concentrations as prognostic factors for patients with UTUC treated with RNU. Our risk stratification model may help physicians design a therapeutic strategy. Keywords Prognosis Upper urinary tract urothelial carcinoma Nephroureterectomy Sodium Hemoglobin
Introduction Radical nephroureterectomy (RNU) with excision of the bladder cuff and resection of the retroperitoneal lymph node is the gold standard for treatment of non-metastatic upper urinary tract urothelial carcinoma (UTUC). However, a significant proportion of patients who have undergone RNU die from disease recurrence. Thus, RNU alone is not adequate for patients with high-risk UTUC, and it is reasonable to consider perioperative chemotherapy to improve patient survival [1]. Neoadjuvant chemotherapy is considered preferable for patients with UTUC over adjuvant chemotherapy, because RNU causes loss of renal function, whereby renal function is inadequate for cisplatin-based chemotherapy [2]. Neoadjuvant chemotherapy has been proven effective for invasive urothelial carcinoma of the bladder, and it is therefore expected to be effective for UTUC [3]. However, accurate preoperative staging is more difficult in UTUC than in bladder cancer, in which
123
Int J Clin Oncol
clinical staging can be accurately determined using magnetic resonance imaging (MRI) and transurethral resection of a bladder tumor. Thus, preoperative parameters to predict poor prognosis must be identified in order to prevent overtreatment of patients with low-risk UTUC. Several such parameters associated with poor prognosis have been reported, but many include pathological findings, such as CpT3 staging or lymphovascular invasion [4]. Recently, preoperative risk stratification on the basis of the clinical T stage, voided urine cytology, and white blood cell (WBC) count has been proposed as a method of predicting survival after RNU [5]. We previously reported that the preoperative sodium concentration was associated with poor prognosis in combination with postoperative pathological parameters [6]. The purpose of the present study is to identify preoperative parameters that predict survival in patients with UTUC who have undergone RNU.
classification; and follow-up data. Clinical T classification was determined by enhanced CT. Laboratory data, including the serum WBC count and platelet counts and levels of hemoglobin, C-reactive protein (CRP), lactate dehydrogenase (LDH), sodium, modified calcium, creatinine, total protein, albumin, alkaline phosphatase, aspartate transaminase (AST), alanine transaminase, and total bilirubin were determined from the serum chemistry tests conducted at admission for surgery. Patients were generally followed up every 3 months during years 0–2 postoperatively, every 6 months during years 2–5, and every 6–12 months thereafter. Tumor recurrence was defined as the development of recurrence at RNU site, lymph node metastasis, and/or visceral metastasis. The present study was approved by the Osaka University Graduate School of Medicine Institutional Review Board, the Osaka General Medical Center Institutional Review Board, and the Osaka Police Hospital Institutional Review Board, and written informed consent was obtained from all patients.
Materials and methods Statistical analysis Patients We retrospectively analyzed the medical records of 357 Japanese men who had undergone RNU between 1998 and 2012 at the Osaka University Hospital, Osaka General Medical Center, and Osaka Police Hospital (256 cases in Osaka University Hospital and Osaka General Medical Center for the training set; 101 cases in Osaka Police Hospital for the test set). The training set data were used for development of a risk stratification model, while the test set data were used for model validation. Seven patients who had distant metastasis and had undergone RNU for palliative treatment were excluded (five from the training set and two from the test set). Twenty-nine patients who had radiographic lymph node involvement were also subsequently excluded (17 from the training set; 12 from the test set). Four patients in the training set who were receiving hemodialysis were excluded because of abnormal laboratory data and 10 patients whose laboratory data were not found in their medical records were excluded (four from the training set, and six from the test set). Thus, the remaining 226 patients in the training set and 80 patients in the test set were examined. In the training set, 140 (62 %) patients underwent laparoscopic RNU, 86 (38 %) underwent open RNU, and 164 (72 %) underwent concomitant retroperitoneal lymphadenectomy. The following clinical data and radiological findings obtained by intravenous pyelography, retrograde pyelography, or computed tomography (CT) were obtained from medical records: age; sex; history of previous bladder tumor; results of voided urinary cytology; presence of hydronephrosis; clinical tumor, lymph node, metastasis (TNM)
123
The results are presented as median values (range). Univariate and multivariate analyses were carried out using the Cox proportional hazards method with stepwise forward selection to predict recurrence-free survival (RFS) and cancer-specific survival (CSS). The variables entered into the model for RFS analysis were patient age; gender; clinical T classification; and preoperative hemoglobin, sodium, and albumin levels. Those entered into the model for CSS analysis were patient age; gender; clinical T classification; and preoperative hemoglobin, CRP, sodium, and albumin levels. The proportional hazards assumptions for the Cox regression model fit were tested and confirmed. The association of serum sodium or hemoglobin levels with CSS was examined using time-dependent Cox models with adjustment for potential confounders. The Kaplan– Meier estimate of survival curve and log-rank test were used to explore the association between the parameters and patient survival. The patients were stratified into favorable (0 point), intermediate (1 point), and poor risk groups (2 points) according to the number of positive prognostic factors (low serum sodium concentration and low hemoglobin concentration). The cutoff level of sodium concentration was set at 131 mEq/L, and those of hemoglobin were set at 13.8 mg/dL for men and 12.0 mg/dL for women. External validation of the risk stratification model was performed separately for 80 patients with UTUC who underwent RNU at the Osaka Police Hospital. All probability (P) values were two-sided, with statistical significance being set at P \ 0.05. All statistical analyses were performed using SPSS version 11.0.1 (SPSS, Chicago, IL, USA), GraphPad Prism 5 (GraphPad Software, La Jolla,
Int J Clin Oncol
CA, USA), and R version 3.0.1 with the Smooth HR package (Artur Agostinho Arau´jo and Luı´s Meira-Machado, 2013); smoothHR: smooth hazard ratio curves taking a reference value. R package version 1.0 URL http://CRAN. R-project.org/package=smoothHR.
Table 1 Clinical and pathological characteristics of patients who underwent nephroureterectomy for upper tract urothelial carcinoma Training set n
226
Age (years) [median (range)]
70 (37–90)
Test set 80 70 (50–91)
Gender, n (%)
Results
Male
153 (68)
68 (85)
73 (32)
12 (15)
108 (48) 110 (49)
45 (56) 30 (38)
8 (3)
5 (6)
No
177 (78)
63 (79)
Yes
49 (22)
17 (21)
Negative
119 (53)
35 (44)
Positive
107 (47)
45 (56)
No
159 (70)
44 (55)
Yes
67 (30)
32 (40)
0 (0)
4 (5)
Female
Table 1 shows the clinical and pathological characteristics of the patients. The median age was 70 years (range 37–90), and 155 of the 226 patients (67.9 %) were men. The median follow-up time was 41 months (range 1–164). Of the 226 patients, 50 (22.1 %) experienced disease recurrence and 43 (19.0 %) died from UTUC during follow-up. The overall 2- and 5-year RFS estimates were 82.4 % [95 % confidence interval (CI) 76.4–87.0] and 75.5 % (95 % CI 68.4–81.1), respectively, while the overall 2- and 5-year CSS estimates were 89.0 % (95 % CI 83.7–92.6) and 77.7 % (95 % CI 70.3–83.4), respectively. Univariate and multivariate analyses were performed to identify preoperative predictors of disease recurrence and cancer-specific mortality (Table 2). In the univariate analysis, age, clinical T stage classification, preoperative hemoglobin concentration, preoperative sodium concentration, and preoperative albumin concentration were found to be significantly associated with disease recurrence (P \ 0.05). Multivariate analysis with these variables and gender showed that patient age and serum sodium concentration were significantly associated with RFS (P \ 0.05). Additionally, in the univariate analysis, clinical T stage classification and the preoperative levels of hemoglobin, CRP, sodium, and albumin were found to be significantly associated with cancer-specific mortality (P \ 0.05). Furthermore, multivariate analysis with these variables, age, and gender showed that the preoperative sodium and hemoglobin levels were significantly associated with CSS (P \ 0.05). The association of serum sodium and hemoglobin levels with cancer-specific mortality is shown in Fig. 1. In the training cohort, the preoperative serum sodium concentration ranged from 133 to 147 mEq/L. It was below normal (i.e., \136 mEq/L) in only three cases (1.2 %). Lower preoperative serum sodium levels showed a significant association with higher mortality even after multivariate adjustment for age, sex, and hemoglobin level. The threshold for the serum sodium was set at 141 mEq/L, which is the median of the logarithm of the hazard ratio. Patients with serum sodium levels of \141 mEq/L had an unadjusted cancer-specific mortality hazard ratio of 3.51 (95 % CI 1.93–6.38) (P \ 0.001) when compared with those whose serum sodium levels were C141 mEq/L. Lower preoperative hemoglobin concentration also showed
Tumor location, n (%) Renal pelvis Ureter Multiple Previous bladder cancer, n (%)
Urinary cytology, n (%)
Hydronephrosis, n (%)
Unknown Clinical T stage, n (%) \T3
205 (91)
70 (88)
CT3 Neoadjuvant chemotherapy
21 (9) 8 (4)
10 (12) 0 (0)
Adjuvant chemotherapy
44 (19)
29 (36)
Ta
56 (25)
20 (25)
Tis
15 (7)
4 (5)
T1
42 (18)
9 (11)
T2
40 (17)
12 (15)
T3
66 (29)
33 (41)
T4
7 (3)
2 (3)
155 (69)
79 (99)
9 (4)
1 (1)
62 (27)
0 (0)
Pathological T classification, n (%)
pN stage, n (%) Negative Positive Unknown Tumor grade, n (%) G1
17 (8)
8 (10)
G2 G3
105 (46) 102 (45)
43 (53) 29 (36)
2 (1)
1 (1)
Absent
173 (77)
47 (59))
Present
45 (19)
31 (39)
8 (4)
2 (2)
Unknown Lymphovascular invasion, n (%)
Unknown
an association with higher mortality after adjustment for age, sex, and sodium level. The hemoglobin level of the median logarithm of hazard ratio was 12.7 mg/dL.
123
Int J Clin Oncol Table 2 Univariate and multivariate analyses predicting disease recurrence and cancerspecific mortality in 139 patients who underwent nephroureterectomy for upper tract urothelial carcinoma
Univariate
Multivariate
HR (95 % CI)
P value
HR (95 % CI)
P value
Age, years
1.03 (1.00–1.07)
0.035
1.04 (1.00–1.08)
0.018
Gender (female vs. male)
1.15 (0.64–2.07)
0.61
–
–
Urine cytology (positive vs. negative)
0.92 (0.58–1.48)
0.75
–
–
–
–
0.83 (0.74–0.94)
0.003
–
–
Recurrence
Tumor location (ureter vs. pelvis)
1.57 (0.88–2.78)
0.12
(multiple vs. pelvis)
2.25 (0.54–9.33)
0.26
Previous bladder cancer
1.14 (0.55–2.36)
0.71
Clinical T classification (CcT3 vs. BcT2)
2.41 (1.13–5.17)
0.023
Hydronephrosis
1.49 (0.83–2.67)
0.17
White blood cells (per 1 9 103/lL)
1.06 (0.937–1.21)
0.31
Hemoglobin (per 1 g/dL) Platelets (per 1 9 104/lL)
0.84 (0.72–0.98) 1.00 (0.99–1.00)
0.028 0.30
CRP (per 1 mg/dL)
1.10 (0.99–1.23)
0.073
Sodium (per 1 mEq/L)
0.84 (0.75–0.93)
0.002
Corrected calcium (per 1 mg/dL)
1.14 (0.51–2.56)
0.74
Creatinine (per 1 mg/dL)
1.29 (0.71–2.33)
0.39
Lactate dehydrogenase (per 1 IU/L)
0.99 (0.99–1.00)
0.31
Total protein (per 1 g/dL)
0.64 (0.37–1.11)
0.11
Albumin (per 1 g/dL)
0.38 (0.17–0.87)
0.021
AST (per 1 IU/L)
0.97 (0.94–1.01)
0.18
ALT (per 1 IU/L)
0.96 (0.93–1.00)
0.055
Alkaline phosphatase (per 1 IU/L)
1.00 (0.99–1.00)
0.93
Total bilirubin (per 1 mg/dL)
0.29 (0.06–1.23)
0.095
Age, years
1.02 (0.98–1.06)
0.18
–
–
Gender (female vs. male)
0.94 (0.50–1.76)
0.87
–
–
Urine cytology (positive vs. negative) Tumor location (ureter vs. pelvis)
1.06 (0.64–0.17) 1.48 (0.81–2.73)
0.81 0.19
(multiple vs. pelvis)
1.48 (0.20–10.8)
0.69
Previous bladder cancer
1.04 (0.46–2.36)
0.90
Clinical T classification (CcT3 vs. BcT2)
2.62 (1.16–5.91)
0.020
–
–
0.82 (0.70–0.96)
0.019
Cancer-specific mortality
AST aspartate aminotransferase, ALT alanine aminotransferase
123
Hydronephrosis
1.44 (0.77–2.70)
0.25
White blood cells (per 1 9 103/lL)
1.06 (0.92–1.22)
0.40
Hemoglobin (per 1 g/dL)
0.83 (0.71–0.98)
0.029
Platelets (per 1 9 104/lL)
1.00 (0.99–1.00)
0.47
CRP (per 1 mg/dL)
1.99 (1.05–3.79)
0.035
–
–
Sodium (per 1 mEq/L)
0.83 (0.74–0.93)
0.002
0.85 (0.76–0.95)
0.004
Corrected calcium (per 1 mg/dL)
0.87 (0.38–2.014)
0.53
–
–
Creatinine (per 1 mg/dL)
1.35 (0.70–2.57)
0.36
Lactate dehydrogenase (per 1 IU/L)
0.99 (0.99–1.00)
0.85
Total protein (per 1 g/dL)
0.63 (0.35–1.11)
0.13
Albumin (per 1 g/dL) AST (per 1 IU/L)
0.29 (0.12–0.69) 0.97 (0.94–1.01)
0.005 0.23
ALT (per 1 IU/L)
0.97 (0.94–1.00)
0.091
Alkaline phosphatase (per 1 IU/L)
1.00 (0.99–1.00)
0.88
Total bilirubin (per 1 mg/dL)
0.39 (0.08–1.80)
0.22
Int J Clin Oncol
Fig. 1 Association of cancer-specific mortality hazard ratio with preoperative serum sodium concentration (a) and hemoglobin concentration (b) in 226 patients with UTUC who underwent RNU.
Fig. 2 Kaplan–Meyer estimates for each risk group [no (dotted curve), one (dashed curve), or two risk factors (solid curve)] in the training set a (log-rank test: P \ 0.0001) and the test set b (log-rank test: P \ 0.05)
The serum sodium or hemoglobin concentration shown corresponds to the median of the logarithm of the hazard ratio
A risk model was constructed using the preoperative sodium (\141 mEq/L) and hemoglobin level (below normal) obtained from the multivariate analysis used to predict CSS. Patients were stratified into favorable, intermediate, and poor risk groups according to the presence of no, one, or two prognostic factors, respectively. Figure 2a shows the Kaplan–Meier estimates for CSS by risk stratification. The CSS estimates in 5 years were 96.5 % (95 % CI 87.0–98.0) for the favorable risk group, 75.5 % (95 % CI 64.1–83.7) for the intermediate risk group, and 47.0 % (95 % CI 27.1–64.7) for the poor risk group [log-rank test; intermediate vs. favorable: HR 5.41 (95 % CI 1.59–7.61); poor vs. intermediate: HR 2.98 (95 % CI 1.99–9.01)]. Adverse pathological findings of CpT3 staging or lymphovascular invasion were found in 12 of 60 patients (20.0 %) in the favorable risk group, 44 of 118 patients (37.2 %) in the intermediate risk group, and 27 of 48 patients (56.2 %) in the poor risk group (Cochran–Armitage test for trend, P = 0.0001). Multivariate analysis with the logistic regression model adjusted by age, gender, and clinical T classification showed that risk classification with preoperative sodium and hemoglobin levels was significantly associated with pathological findings of CpT3 or lymphovascular invasion (P = 0.005). To validate the risk stratification, 80 patients with UTUC who underwent RNU at Osaka Police Hospital were examined as the test set. The median age was 70 years (range 50–91), and 68 of the 80 patients (83.9 %) were men. The median follow-up time was 32 months (range 1–149). Of the 80 patients, 21 (26.2 %) experienced
123
Int J Clin Oncol
disease recurrence and 12 (15.0 %) died from UTUC during follow-up. Twelve patients were stratified to the favorable risk group, with a 5-year CSS estimate of 100 %. Forty patients were stratified to the intermediate risk group, with a 5-year CSS estimate of 81.4 % (95 % CI 59.6–92.1). Lastly, 27 patients were stratified to the poor risk group, with a 5-year CSS estimate of 67.4 % (95 % CI 39.0–84.7). Figure 2b shows the Kaplan–Meier estimates for CSS by risk classification (log-rank test for trend, P = 0.036).
Discussion It has been reported that, generally, invasive UTUC has a poor prognosis, and approximately 40 % die because of recurrence after RNU [1]. Thus, the demand for preoperative prediction of patient survival, which will aid in the selection of candidates for neoadjuvant chemotherapy, is high. Many postoperative parameters have been reported as prognostic factors, but they could not be used for the selection of patients for neoadjuvant chemotherapy. Neoadjuvant chemotherapy is preferred to adjuvant chemotherapy because patients lose renal function after RNU, making it difficult for them to receive the cisplatin-based chemotherapy. The criteria for patient selection for neoadjuvant chemotherapy are yet to be determined, but it has been recommended that patients with invasive UTUC should receive neoadjuvant chemotherapy [1]. Currently, radiological examination is the only method of preoperatively predicting the pathological stage. CT has been used for preoperative staging of UTUC. However, its accuracy for predicting the pathological TNM stage was reported to be only 59 %; furthermore, under-staging was reported in 16 % cases and non-detection in 24 % [7]. In the present study, among the 23 patients who were preoperatively diagnosed with stage cT3 disease or more, 15 patients (65.2 %) had shown stage pT3 or more on pathological examination, but 8 patients (34.8 %) were preoperatively over-staged. In contrast, 58 (26.1 %) of 203 patients who were not preoperatively diagnosed with stage cT3 disease or more showed stage pT3 or more on pathological examination. Because of these CT inaccuracies for clinical staging, the CT findings did not remain significant after the multivariate analysis including other parameters. Hydronephrosis among the preoperative CT findings was also reported to be associated with a poor prognosis [8, 9], but it was not observed in our cohort. In the multivariate analysis in the present study, preoperative sodium and hemoglobin concentrations in their continuous form were found to be prognostic parameters for patients with non-metastatic UTUC. The preoperative serum sodium concentration ranged from 133 to 147 mEq/L in the test cohort; it was below the normal range (i.e., \136 mEq/L) in only three
123
cases (1.2 %), and the cutoff level of sodium concentration based on the normal range could not be used. The serum sodium concentration in the extracellular fluid volume is strictly regulated, although it does fluctuate from day to day. On analyzing the association between the mortality hazard ratio and continuous sodium concentration, we found an inverse S-shaped association. In other words, a low serum sodium concentration was associated with a high hazard ratio. The sodium concentration which corresponded to the median mortality hazard ratio was 141 mEq/L. The equivalent value for hemoglobin was 12.7 mg/dL. We set the cutoff level of hemoglobin as the lower limit of the normal range, which differs between men and women (men 13.8 mg/dL; women 12 mg/dL). Risk stratification with these two preoperative markers could predict the CSS of patients with non-metastatic UTUC and was associated with the pathological findings of stage pT3 or more and lymphovascular invasion. Recently, Sakano et al. [5] reported preoperative risk stratification with clinical T stage, voided urine cytology, and WBC count, but the preoperative sodium concentration was not included in their analysis, which may explain the difference in the risk stratification models between their study and ours. Since the 5-year estimates of cancer-specific mortality in the favorable risk group in the present study were only 3.6 % in the training set and 0 % in the test set, we could conclude that patients in this group do not require neoadjuvant chemotherapy. The mechanism underlying the anemia and low sodium concentration observed in UTUC patients with a poor prognosis remains to be elucidated. Hyponatremia has been reported to be associated with poor prognosis in several cancers. More than 40 % of patients with small cell lung carcinoma had hyponatremia (\135 mEq/L), which was associated with poorer prognosis. Hyponatremia in patients with small cell lung carcinoma has been explained to be caused by a syndrome of inappropriate secretion of antidiuretic hormone. In localized renal cell carcinoma, patients with lower preoperative sodium concentration (\139 mEq/L) had significantly poorer prognosis than patients with higher preoperative sodium concentration (C139 mEq/L) [10]. The authors suggested that the renal tumors might be inducing a disturbance of the renin– angiotensin–aldosterone axis, which controls the sodium and the extracellular fluids. Hyponatremia has been also reported be a prognostic factor in metastatic renal cell carcinoma [11, 12] and hepatocellular carcinoma [13]. Interleukin-6 (IL-6), one of the key inducers of cancerrelated inflammation, can cause hyponatremia via vasopressin secretion [14]. It can also cause anemia in cancer patients through both complex physiological mechanisms as well as genetic regulatory aspects of erythropoiesis. Hematuria is also likely to be the cause of anemia in
Int J Clin Oncol
patients with UTUC [15]. A previous univariate analysis showed that anemia was associated with the prognosis of UTUC patients [16]. CRP, another preoperative prognostic marker [17], has also shown an association with survival in the univariate analysis, although not in the multivariate analysis, in the present study. Additionally, CRP production from hepatocytes is stimulated by IL-6. Several studies have reported that elevated IL-6 levels are associated with elevated CRP levels in various cancers [18]. It is possible that the preoperative prognostic markers of low serum sodium concentration, anemia, and elevated CRP level indicate IL-6 production by the tumor and inflammatory infiltrate cells, and the IL-6-related pathway may be a new therapeutic target for aggressive UTUC. Indeed, STAT3 expression is a prognostic factor of bladder cancer, and STAT3 is an important agent in the IL-6 pathway [19]. High urinary IL-6 levels were found to be a poor prognostic factor in patients with urothelial carcinoma of the bladder [20]. Thus, future studies should examine whether serum or urinary IL-6 levels in patients with UTUC are associated with survival, like the sodium and hemoglobin concentrations are. Despite its novel findings, our study does have some limitations. First, the preoperative CT findings did not remain significant in the multivariate analysis for CSS. Several patients had renal insufficiency preoperatively and were not suited to undergoing contrast-enhanced examination. This renal insufficiency for using contrast material may have led to the low accuracy of CT for preoperative staging. Preoperative staging by MRI might be more accurate, and the results might be used as a prognostic factor. Although this study proposes a risk stratification model for selection of patients for neoadjuvant chemotherapy, only 8 of the 226 patients (4 %) in our cohort received neoadjuvant chemotherapy. Therefore, it was difficult to assess the association between the effect of neoadjuvant chemotherapy on survival and risk stratification using preoperative sodium and hemoglobin concentrations. Thus, no concrete recommendations can be made from our findings in terms of the selection of candidates for neoadjuvant chemotherapy. Subsequent studies should examine whether the prognosis of patients with intermediate or poor risk based on our risk stratification model can be improved by neoadjuvant chemotherapy. Lastly, although we validated the risk stratification model using a test set from a different institution, this was nonetheless a retrospective study including only Japanese patients. Further large-scale studies should be performed to further validate the proposed model. In conclusion, the present multi-institutional study showed that low preoperative sodium and hemoglobin concentrations, which can be easily and routinely measured, were prognostic factors in patients with UTUC undergoing RNU. We also proposed a risk stratification model for predicting CSS based
on these two preoperative variables, which were also associated with adverse pathological findings. Our risk stratification model may help physicians design a therapeutic strategy. Conflict of interest Norio Nonomura received lecture fees from Novartis Pharma and Pfizer Pharmaceutical.
References 1. Audenet F, Yates DR, Cussenot O et al (2013) The role of chemotherapy in the treatment of urothelial cell carcinoma of the upper urinary tract (UUT-UCC). Urol Oncol 31:407–413 2. Kaag MG, O’Malley RL, O’Malley P et al (2010) Changes in renal function following nephroureterectomy may affect the use of perioperative chemotherapy. Eur Urol 58:581–587 3. Babjuk M, Compe E, Zigeuner R et al (2013) European guidelines on upper tract urothelial carcinomas: 2013 update. Eur Urol 63:1059–1071 4. Margulis V, Shariat SF, Matin SF et al (2009) Outcomes of radical nephroureterectomy: a series from the upper tract urothelial carcinoma collaboration. Cancer 115:1224–1233 5. Sakano S, Matsuyama H, Kamiryo Y et al (2013) Risk group stratification based on preoperative factors to predict survival after nephroureterectomy in patients with upper urinary tract urothelial carcinoma. Ann Surg Oncol 20:4389–4396 6. Fujita K, Tanigawa G, Imamura R et al (2013) Preoperative serum sodium is associated with cancer-specific survival in patients with upper urinary tract urothelial carcinoma treated by nephroureterectomy. Int J Urol 20:594–601 7. Scolieri MJ, Paik ML, Brown SL et al (2000) Limitations of computed tomography in the preoperative staging of upper tract urothelial carcinoma. Urology 56:930–934 8. Ng CK, Shariat SF, Lucas SM et al (2011) Does the presence of hydronephrosis on preoperative axial CT imaging predict worse outcomes for patients undergoing nephroureterectomy for uppertract urothelial carcinoma? Urol Oncol 29:27–32 9. Cho KS, Hong SJ, Cho NH et al (2007) Grade of hydronephrosis and tumor diameter as preoperative prognostic factors in ureteral transitional cell carcinoma. Urology 70:662–666 10. Vasudev NS, Brown JE, Brown SR et al (2008) Prognostic factors in renal cell carcinoma: association of preoperative sodium concentration with survival. Clin Cancer Res 14:1775–1781 11. Jeppesen AN, Jensen HK, Donskov F et al (2010) Hyponatremia as a prognostic and predictive factor in metastatic renal cell carcinoma. Br J Cancer 102:867–872 12. Kawashima A, Tsujimura A, Takayama H et al (2012) Impact of hyponatremia on survival of patients with metastatic renal cell carcinoma treated with molecular targeted therapy. Int J Urol 19:1050–1057 13. Huo T-I, Lin H-C, Hsia C-Y et al (2008) The MELD-Na is an independent short- and long-term prognostic predictor for hepatocellular carcinoma: a prospective survey. Dig Liver Dis 40:882–889 14. Swart RM, Hoorn EJ, Betjes MG et al (2011) Hyponatremia and inflammation: the emerging role of interleukin-6 in osmoregulation. Nephron Physiol 118:45–51 15. Dicato M, Plawny L, Diederich M (2010) Anemia in cancer. Ann Oncol 21(Suppl 7):vii167–vii172 16. Akdogan B, Dogan HS, Eskicorapci SY et al (2006) Prognostic significance of bladder tumor history and tumor location in upper tract transitional cell carcinoma. J Urol 176:48–52. doi:10.1016/ S0022-5347(06)00511-8
123
Int J Clin Oncol 17. Obata J, Kikuchi E, Tanaka N et al (2013) C-reactive protein: a biomarker of survival in patients with localized upper tract urothelial carcinoma treated with radical nephroureterectomy. Urol Oncol 31:1725–1730 18. Guthrie GJK, Roxburgh CSD, Horgan PG et al (2013) Does interleukin-6 link explain the link between tumour necrosis, local and systemic inflammatory responses and outcome in patients with colorectal cancer? Cancer Treat Rev 39:89–96
123
19. Mitra AP, Pagliarulo V, Yang D et al (2013) Generation of a concise gene panel for outcome prediction in urinary bladder cancer. J Clin Oncol 27:3929–3937 20. Seguchi T, Yokokawa K, Sugao H et al (1992) Interleukin-6 activity in urine and serum in patients with bladder carcinoma. J Urol 148:791–794
