Development of a new outcome prediction model in carcinoma invading the bladder based on preoperative serum C-reactive protein and standard pathological risk factors: the TNR-C score

Authors


Georgios Gakis, Department of Urology, Eberhard-Karls University, Hoppe-Seyler Strasse 3, D-72076 Tuebingen, Germany.e-mail: georgios.gakis@med.uni-tuebingen.de

Abstract

Study Type – Prognosis (case series)

Level of Evidence 4

What’s known on the subject? and What does the study add?

There is increasing evidence for a prognostic significance of pretherapeutically elevated serum C-reactive protein levels in various cancers. However, little is known about its significance in patients with invasive bladder cancer. This study shows that serum CRP is an independent predictor for cancer-specific survival in bladder cancer, and its incorporation into a new outcome model (TNR-C Score) encompassing major pathological determinants for survival, increases significantly its predictive accuracy.

OBJECTIVE

• To assess the predictive value of preoperative C-reactive protein (CRP) in patients undergoing radical cystectomy (RC) for carcinoma invading the bladder in light of recent data showing it to be an independent indicator of adverse oncological outcome in other malignancies.

PATIENTS AND METHODS

• A contemporary, consecutive series of 246 patients undergoing RC and bilateral pelvic lymphadenectomy for bladder cancer between 1999 and 2009.

• Elevated CRP was defined as >0.5 mg/dL and was consistent during the study period. The median (range) follow-up was 30 (6–116) months.

• Kaplan–Meier analysis was used to estimate cancer-specific survival (CSS) using a log-rank test and Cox regression analysis for multivariate analysis of risk factors.

• Based on regression estimates of significant parameters in multivariate analysis, a new CRP-based scoring model was developed to predict cancer-specific outcomes. The predictive accuracy of the model was evaluated using the concordance index.

RESULTS

• The 3-year CSS was 74.0% in patients with normal and 44.0% with elevated CRP (P < 0.001).

• In multivariate analysis, CRP (P < 0.001; used as a continuous variable), tumour stage (P= 0.001), lymph-node density ≥0.09 (P= 0.02) and resection margin status (P < 0.001) were independent predictors of CSS.

• The 3-year CSS in patients with a score in the ranges 0–2, 3–6 and 7–10 was 80.5%, 44.9% and 7.1%, respectively (P < 0.001). Consideration of CRP in the final model increased its predictive accuracy by 4.9% with a concordance index of 0.788 (P= 0.01).

CONCLUSIONS

• This is the largest, contemporary series to date indicating that preoperative serum CRP is an independent risk factor for CSS.

• CRP may be a useful parameter to include in predictive bladder cancer nomograms.

Abbreviations
CRP

C-reactive protein

CSS

cancer-specific survival

RC

radical cystectomy

TURBT

transurethral resection of bladder tumour.

INTRODUCTION

Systemic inflammation is a common host reaction to carcinogenesis or cancer progression [1]. In tumour tissues, cancer cells are embedded in a microenvironment resembling chronic inflammation [2]. Serum C-reactive protein (CRP) is a highly sensitive indicator of acute or chronic inflammation [3]. It is released mainly by hepatocytes after interleukin-6 stimulation from the tumour microenvironment and acts as an opsonizing agent for cancer cell detection and elimination, but also can be released by tumour cells themselves, facilitating cancer cell survival by pleiotrophic effects [4]. Thus, elevated CRP levels are not only an epiphenomenon of the tumour microenvironment, but also may play an important role in the host tumour defence response [5,6].

Elevated CRP levels have been found in patients with malignant disease, implying a close link between inflammation and malignancy [5,7]. Moreover, in various cancers, including gastric, pancreatic, lung and ovarian, elevated CRP concentrations have been correlated independently with tumour load and disease progression [6,8–10].

A recent screening study among healthy individuals showed that elevated CRP indicates a higher risk of developing bladder cancer [11]. Additionally, in patients with carcinoma invading bladder muscle treated with chemoradiotherapy only, elevated CRP concentrations were associated with adverse outcome [12]. Similarly, in patients with upper-tract urothelial carcinoma undergoing surgery, increased CRP concentrations predicted poor survival [13].

Because there is little evidence available in the literature concerning the prognostic value of CRP in cystectomy patients, we analyzed the correlation between preoperative serum CRP and clinicopathological parameters in patients undergoing radical cystectomy (RC) for bladder cancer, aiming to determine its clinical significance as a potential predictor for poor oncological outcome.

PATIENTS AND METHODS

In the present retrospective study, we reviewed the clinical and pathological records of 246 consecutive patients who underwent RC and bilateral pelvic lymphadenectomy at our institution for carcinoma invading the bladder between 1999 and 2009. There were six patients (one with normal, five with elevated serum CRP) who were excluded because they had undergone platinum-based neoadjuvant chemotherapy in which cytotoxic effects may also increase CRP levels [14]. The present study was approved by the ethics committee of Tuebingen.

Serum CRP was measured 1–3 days before cystectomy by immunoturbidimetry. Elevated CRP was defined as >0.5 mg/dL [13,14], which was consistent throughout the study period. The clinical and pathological parameters assessed were: age at RC, gender, tumour stage, lymph node tumour involvement, lymph node density ≥0.09, tumour size, resection margin status, tumour grade, previous BCG therapy, postoperative chemotherapy, non-urothelial carcinoma pathology, median time between the last transurethral resection of bladder tumour (TURBT) and RC, number of TURBTs before RC, preoperatively elevated leukocyte count and tumour multifocality.

HISTOLOGICAL ASSESSMENT

The histological assessment was performed in a single pathology department and was based on the TNM classification approved by the American Joint Cancer Committee [15]. The pathological macro- and microscopic evaluation of cystectomy specimens included a cross-sectioning of the entire specimen with immunohistochemical staining for cytokeratin-5/6 and -20 to identify the presence of urothelial carcinoma [16]. Positive surgical margins were defined as the microscopic presence of malignant cells at the resection margins.

FOLLOW-UP

Electronic hospital charts and physician records were reviewed to determine clinical outcomes. Patients generally were seen postoperatively at least every 3–4 months for the first year, semi-annually for the second and third years, and annually thereafter. Follow-up examinations included radiological imaging with CT in all patients [17]. In addition to physical examination with laboratory testing, i.v. pyelography, cystoscopy, urine cytology, urethral washings and bone scintigraphy were carried out, if indicated [17]. Local recurrence was defined as recurrence in the surgical bed, and distant as recurrence at distant organs. The median (range) follow-up was 30 (6–116) months. Clinical outcomes were measured from the date of cystectomy to the date of first documented recurrence at CT, the date of death, or the date of last follow-up when the patient had not experienced disease recurrence.

STATISTICAL ANALYSIS AND MODEL DEVELOPMENT

For univariate analysis, Fisher’s exact test was used for nominal data and Student’s t-test for scaled data. For multivariate analysis, Cox proportional hazards analysis was carried out to evaluate risk factors for cancer-specific death. Kaplan–Meier plots were used to estimate and cancer-specific survival (CSS) using a log-rank test.

A new outcome prediction model for CSS was developed based on the regression coefficients from the final multivariate Cox proportional hazards model. The model building algorithm was based on the fifteen clinical and pathological risk factors for cancer-specific death that have been addressed in larger studies [18,19]. All these risk factors were evaluated in univariate analysis. As a second step, significant parameters of univariate analysis were investigated in multivariate analysis and used to create a new predictive model with and without the inclusion of CRP, as has been carried out in a similar analysis investigating the prognostic value of CRP in RCC [20]. In the final model, the regression coefficient of each parameter was divided by the coefficient of the parameter with the highest regression coefficient, multiplied by 4 and rounded to the nearest integer [20]. The predictive ability of the new model was evaluated using the concordance index (C-index) [21]. Receiver-operating curves were used to evaluate the predictive accuracy of the final model including standard pathological risk factors with and without the inclusion of CRP. In this context, a C-index of 1.0 indicates that the model discriminates perfectly between patients with different outcomes, whereas a value of 0.5 is equivalent to a toss of a coin, indicating that the model contains no predictive information [22].

P < 0.05 (two-sided) was considered statistically significant. Statistical analysis was performed using JMP, version 8.0.2. Values are given as the mean (sem) for all continuous variables or as median (range) for non-continuous variables.

RESULTS

Of the 246 patients, 120 (48.8%) had normal (≤0.05) and 126 had elevated (>0.05) CRP (51.2%). The CSS rate at 3 and 5 years in the total patient cohort was 60.2% and 56.0%, respectively. Cancer-specific death occurred in 21 patients with CRP ≤0.5 mg/dL and 53 patients with CRP >0.5 mg/dL. Substratified for patients with normal and elevated CRP, the 3/5-year CSS was 74.0/71.6% and 44.0/40.5%, respectively (P < 0.001) (Fig. 1). In univariate analysis, elevated CRP was associated with extravesical disease, increased lymph node density ≥0.09, increased tumour size (all P < 0.001), lymph node tumour involvement (P= 0.002), increased age (P= 0.003) and positive resection margins (P= 0.026). No significant differences were found between both groups with respect to gender, median time between last TURBT and RC, the number of TURBTs, tumour multifocality and grade, previous BCG therapy, non-urothelial carcinoma pathology, preoperatively elevated leucocyte count and postoperative chemotherapy (Table 1).

Figure 1.

Kaplan–Meier analysis for cancer-specific survival in patients with normal (≤0.5 mg/dL) and elevated C-reactive protein (CRP) (>0.5 mg/dL) ( P < 0.001).

Table 1.  Clinical and pathological parameters in patients with bladder cancer and normal (≤0.5 mg/dL) or elevated (>0.5 mg/dL) serum C-reactive protein (CRP)*
VariableNormal CRP (≤0.5 mg/dL)Elevated CRP (>0.5 mg/dL)P
  1. *Univariate analysis with Fisher’s exact/Student×s t-test. RC, radical cystectomy; TURBT, transurethral bladder resection; UC, urothelial carcinoma. Values in bold indicate a statistically significant difference.

Number of patients (%)120 (48.8)126 (51.2) 
Mean age at RC (years)64.768.00.003
Median (range)65 (43–84)69 (40–86) 
Gender   
 Male891020.28
 Female3124 
T stage   
 Organ-confined (≤ pT2b)7846<0.001
 Extravesical (≥ pT3a)4280 
Node-positive disease at RC20430.002
Lymph node density ≥0.09 1129<0.001
Mean tumour size (cm)2.93.8<0.001
 Median2.83.5 
 Range0.8–6.50.5–12 
Positive resection margins6170.026
Tumour grade   
 G1110.07
 G22940 
 G39675 
Previous BCG therapy26180.18
Postoperative chemotherapy17230.38
Non-UC pathology590.41
Mean time between last   
 TUR-BT and RC (days)56760.47
 Median (range)37 (8–392)32 (5–1281) 
Mean number of TURBTs before RC2.01.80.57
 Median (range)1 (1–12)1 (1–9) 
Preoperatively elevated leucocyte count (>10 000/mm3)39480.43
Tumour multifocality41370.67

In multivariate analysis, cancer-specific death was significantly associated with increased CRP, used as a continuous variable (hazard ratio, HR, 1.18; 95% CI, 1.09–1.27, P < 0.001), advanced tumour stage (HR, 2.52; 95% CI, 1.43–4.60, P= 0.001), increased lymph node density ≥0.09 (HR, 2.12; 95% CI, 1.15–3.75, P= 0.015) and positive resection margins (HR, 3.84, 95% CI, 1.84–7.72, P < 0.001). A new scoring model was developed to predict cancer-specific outcomes after radical cystectomy in carcinoma invading the bladder using the regression coefficients of the final multivariate model. The final model consisted of T-stage, lymph node density, resection margin status and CRP (Table 2). The score was calculated as 4 (if positive resection margins) + 3 (if ≥ pT3a) + 2 (if lymph node density ≥0.09) + 1 (if CRP >0.5 mg/dL) and 0 (if otherwise).

Table 2.  Multivariate analysis of risk factors for cancer-specific death in patients with bladder cancer
Full modelFinal model
VariableRegression coefficientHazard ratio (95% CI)PRegression coefficient (95% CI)Hazard ratio (95% CI)P
  1. Cont., continuous variable; CRP, C-reactive protein; LN, lymph node. Values in bold indicate a significant difference.

Serum CRP (cont.)0.1831.20 (1.10–1.30)0.00120.1681.18 (1.09–1.27)0.001
≥pT3a vs ≤pT2b0.5212.83 (1.54–5.45)0.00070.4632.52 (1.43–4.60)0.001
LN density ≥0.09 vs <0.090.4072.25 (1.17–4.20)0.01600.3762.12 (1.15–3.75)0.015
Positive vs negative resection margins0.6653.77 (1.69–8.11)0.00150.6733.84 (1.84–7.72)0.001
Age >65a vs ≤65a0.1981.49 (0.84–2.69)0.16
Tumour size >3 cm vs ≤3 cm0.1161.26 (0.74–2.13)0.38

The median (mean, range) score of the final model was 3 (2.85, 0–10). The 3/5-year CSS in patients with a score in the ranges 0–2 (low-risk), 3–6 (intermediate-risk) and 7–10 (high-risk) was 80.5/78.1%, 44.9/42.8% and 7.1/0%, respectively (P < 0.001) (Fig. 2). Consideration of CRP in the final model increased its predictive accuracy by 4.9% with a C-index of 0.788 (P= 0.01).

Figure 2.

Cancer-specific survival by score categories in the final tumour stage, lymph node density, resection margin status and serum C-reactive protein (TNR-C) model (differences between subgroups, P < 0.001).

DISCUSSION

In the present study, we investigated the prognostic value of preoperative serum CRP levels in patients with carcinoma invading the bladder because increasing evidence exists for adverse oncological outcome in patients with elevated CRP in other malignancies [6,8–10].

In univariate analysis, elevated CRP was strongly associated with advanced tumour stage, lymph node tumour involvement, lymph node density ≥0.09 and positive resection margins. Because all of these pathological parameters are of the utmost prognostic importance in carcinoma invading the bladder [18], serum CRP may reflect local tumour extent and aggressiveness. Positive correlations were also found for increased tumour size and age, which are parameters that have recently been suggested to predict adverse oncological outcome in carcinoma invading the bladder [23,24]. Interestingly, in univariate analysis, elevated CRP correlated better with lymph node density ≥0.09 than with lymph node tumour involvement. This finding may be indicative of a possible curative effect on micrometastases by pelvic lymphadenectomy [1]. Indeed, lymph node density (which actually reflects the relationship of lymph node tumour burden with the extent of pelvic lymphadenectomy) may be of superior prognostic importance than the presence of lymph node tumour involvement per se[25]. Therefore, in the multivariate model, lymph node density ≥0.09 was considered for further analysis. In addition, we used serum CRP as a continuous parameter because continuously coded parameters obviate the problem of lack of standardization of threshold values. This strengthens the prognostic importance of predictive markers [26].

In multivariate analysis, serum CRP (used as a continuous variable), advanced tumour stage (≥ pT3a), lymph node density ≥0.09 and positive resection margins were independent risk factors of cancer-specific death. Previous studies reported that the independent predictor status of a novel marker does not imperatively translate into better predictive accuracy when considered along with established predictors. Therefore, based on regression estimates of significant parameters in multivariate analysis [20], we developed a new and simple outcome prediction model after RC. This model accounts only for tumour stage, lymph node density, resection margin status and serum CRP. We term this scoring model the TNR-C. In this respect, because 74 cancer-specific deaths occurred in the present study population, the Cox regression model can sufficiently evaluate not more than four risk factors, which consisted of the four significant parameters of multivariate analysis [22]. The final model consisting of these four parameters yielded a high predictive accuracy of 0.788. Of note, the addition of serum CRP to these baseline pathological risk factors increased significantly the predictive accuracy by 4.9%. In other words, from a practical perspective, the consideration of CRP in the final model would result in the correct classification of 49 additional patients out of 1000.

There is some evidence in the literature with regard to the prognostic value of serum CRP in bladder cancer [7]. In a previous study evaluating the role of CRP in patients with bladder cancer undergoing chemoradiotherapy only, pretreatment levels indicated poor survival [12]. There is also evidence suggesting a correlation between elevated CRP >0.5 mg/dL and outcome in patients undergoing surgery for urothelial cancer of the upper urinary tract. Saito et al. [13] reported an association with poor survival in patients with upper-tract urothelial carcinoma. In 105 patients with urothelial carcinoma of the bladder (72% with superficial disease and 23% who underwent additional surgery or chemotherapy), Hilmy et al. [27] found tumour stage and elevated CRP >0.5 mg/dL to be the only independent parameters for decreased CSS.

The results obtained in the present study suggest that elevated preoperative CRP is an independent prognostic marker and may be used for improved risk stratification. Because clinical understaging in carcinoma invading the bladder can attain levels of up to 50%[18,28], preoperative CRP may serve as a risk factor for advanced disease and aid in the selection of patients for adjuvant or neoadjuvant chemotherapy. With regard to its effect on tumour downstaging in cystectomy candidates, CRP may also play a role as a response parameter after neoadjuvant chemotherapy [29].

The integration of CRP levels in current nomograms may potentially improve their predictive value. A recent study in 130 patients undergoing surgery for primary RCC found that, preoperatively, CRP was a strong predictor of metastasis and mortality, and advocated its incorporation into current clinical algorithms to maximize their predictive value [30]. Similarly, McArdle et al. [31] found significant correlations between elevated CRP levels and outcome in localized prostate cancer, concluding that the presence of a systemic inflammatory response at diagnosis predicted poor outcome independently of PSA and Gleason score.

The present study has limitations inherent to any retrospective analysis. First, CRP is a sensitive but non-specific serum biomarker for inflammation and tissue damage. Therefore, increased levels may be the result of any intervention before RC or subclinical infection. However, in patients from the present study with either elevated or normal CRP, neither the time from last TURBT, nor the number of previous TURBTs differed, suggesting that these confounders do not play a significant role in the interpretation of the results. In addition, there was no significant correlation between elevated CRP and leucocytes. Second, we did not assess postoperative CRP levels to determine their impact on outcomes. Because of the high morbidity associated with RC, acute or subacute inflammatory conditions (e.g. surgical infections or postoperative complications) may significantly increase CRP levels in the perioperative period [17]. CRP levels may also vary depending on the degree of surgical stress caused by different techniques (e.g. laparoscopic or open) [32]. In this respect, further prospective studies evaluating the impact of postoperative CRP levels in cystectomy patients would certainly be helpful. Third, the results of the present study await formal external validation before routine use in clinical practice and validation of this new model with established bladder cancer nomograms [33,34].

Unlike many of the other prognostic markers in bladder cancer, CRP levels, as well as the underlying inflammatory process, can potentially be modified. Therefore, it should be questioned whether CRP levels can be influenced therapeutically, thus possibly obtaining better tumour outcomes. Hence, whether CRP-lowering agents, such as cyclooxygenase inhibitors or statins, may have implications for bladder cancer therapy, as has been recently suggested for prostate cancer [35], remains to be addressed in future prospective trials.

To our knowledge, the present study is the largest contemporary series to date evaluating the role of pretreatment CRP levels on oncological outcome in a homogenous consecutive series of patients undergoing RC for carcinoma invading the bladder. The findings suggest that pretreatment CRP levels, which can be easily obtained in daily practice, provide additional independent prognostic information for CSS. Therefore, CRP may be a useful parameter to include in predictive bladder cancer nomograms. Further prospective studies should yield a better understanding of how the inflammatory process, and potentially CRP itself, can influence disease progression, and may point the way toward improved therapies and outcomes in bladder cancer.

CONFLICT OF INTEREST

None declared.

Ancillary