SEARCH

SEARCH BY CITATION

Keywords:

  • CRP;
  • inflammation;
  • recurrence;
  • accuracy;
  • prognosis;
  • kidney cancer

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Patients and Methods
  5. Results
  6. Discussion
  7. Conflict of Interest
  8. Details of Ethics Approval
  9. References

Objective

  • To validate high-sensitivity C-reactive protein (hs-CRP) serum levels as an independent marker for disease-free survival (DFS) in clinically localised clear cell renal cell carcinoma (ccRCC).

Patients and Methods

  • In all, 403 consecutive patients with clinically localised (T1–3N0M0) ccRCC treated by radical or partial nephrectomy were enrolled.
  • Preoperative serum levels of hs-CRP were evaluated as both a continuous and categorical variables.
  • Associations with clinical (age, gender) and pathological variables (T classification, grade, tumour necrosis) were assessed with the chi-square and Kruskal–Wallis tests.
  • Univariable and multivariable Cox proportional hazards models were fitted. The prognostic accuracy (PA) was assessed with Harrell's C-index.

Results

  • The mean hs-CRP level was 1.32 mg/dL. The hs-CRP levels were associated with T classification (P = 0.05), high-grade disease (P < 0.001) and tumour necrosis (P = 0.003).
  • After a median follow-up of 43 months, 41 patients (10.1%) had developed disease recurrence. With each unit increase in hs-CRP levels, the risk of recurrence increased by 10% (hazard ratio 1.10, P = 0.015).
  • The thresholds of 0.5 and 0.75 mg/dL showed the best discrimination for stratification of patients according to the probability of recurrence.
  • These categorically coded hs-CRP levels were identified as independent prognostic factors in multivariable analyses (P < 0.001) and led to a significant increase in the PA of a multivariable base model containing the variables of the ‘Stage, Size, Grade and Necrosis’ (SSIGN) score.

Conclusions

  • This study validates preoperative serum hs-CRP levels as independent prognostic factor after surgery for localised ccRCC.
  • Hs-CRP may be included in standard prognostic modelling after surgery and may guide surveillance and inclusion in adjuvant clinical trials.

Abbreviations
ccRCC

clear cell RCC

CRP

C-reactive protein

DFS

disease-free survival

hs

high-sensitivity (assay)

PA

prognostic accuracy

SSIGN

Stage, Size, Grade and Necrosis (score)

Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Patients and Methods
  5. Results
  6. Discussion
  7. Conflict of Interest
  8. Details of Ethics Approval
  9. References

Following curative surgery for clinically localised (N0M0) RCC, up to 30% of patients develop recurrent disease [1-3]. Currently there are no approved adjuvant therapies for patients at high risk of recurrence and postoperative surveillance remains the standard of care [4, 5]. Once metastatic, RCC has a poor prognosis, although recently approved systemic therapies are associated with prolonged survival in selected cohorts [6]. It is therefore of great importance to accurately stratify patients after surgery into those with high- or low-risk of recurrence, as both surveillance imaging and possible inclusion in adjuvant clinical trials are dependent on the risk profile [7]. Several prognostic models have been proposed to determine this risk profile, most of which include T classification and grade [1, 8]. The accuracy of standard prognostic factors is increased by additional clinical, pathological and molecular markers, including serum markers of systemic inflammatory response [9, 10].

In this regard, the prognostic value of the acute-phase reactant C-reactive protein (CRP) has been a matter of interest in several studies, as this marker can be assessed routinely. Preoperative serum CRP has been identified as a marker of disease-free survival (DFS) in patients with clinically localised RCC [2, 11, 12]. Also, CRP has been associated with prolonged survival in metastatic RCC treated with targeted therapy [13, 14], which has developed as first-line systemic therapy [15]. CRP has increased the prognostic accuracy (PA) of standard variables [10, 16], and has subsequently been included in the TNM-C scoring model [17]. However, for the latter analyses, patients at all stages were included or data did not rely on high-sensitivity (hs) assays.

The aims of the present study were to validate hs-CRP as a marker of DFS in patients with clinically localised RCC and to assess the PA of prognostic models with and without hs-CRP. For these aims, we studied a prospective cohort from a single institution.

Patients and Methods

  1. Top of page
  2. Abstract
  3. Introduction
  4. Patients and Methods
  5. Results
  6. Discussion
  7. Conflict of Interest
  8. Details of Ethics Approval
  9. References

Between 2002 and 2010, 931 patients underwent radical or partial nephrectomy for suspected RCC at our institution. Clinical and pathological characteristics, perioperative outcomes, survival and laboratory data of these patients were abstracted in the prospectively maintained institutional kidney cancer database. The database was queried for the purpose of this study. Patients with clinically localised, unilateral, clear cell RCC (ccRCC; pT1–3N0M0) were eligible. Patients with benign renal tumours, RCC of other subtypes, hereditary RCC, bilateral disease, synchronous metastases, positive surgical margins, follow-up duration of <3 months, missing hs-CRP levels, autoimmune diseases, concomitant second malignancies, bilateral renal tumours, and chronic inflammatory diseases were excluded. In all, 403 patients met inclusion criteria. The study was approved by the Institutional Review Board (protocol registration number 226/2009), and all patients gave informed consent for bloods to be taken.

Clinical and pathological characteristics, DFS and hs-CRP levels were extracted from the kidney cancer database. Clinical characteristics included age and gender. Pathological characteristics included T classification, tumour size according to the largest pathological diameter, nuclear grade, and histological tumour necrosis. T classification was assigned according to the 2010 staging manual [18], and grade according to Fuhrman criteria [19]. Grade was stratified as low (G1–2) or high grade (G3–4). Slides were all re-assessed by one expert uro-pathologist (A.H.).

Blood for preoperative hs-CRP measurement was taken by clean venepuncture at the time of admission. Blood was drawn into a serum Vacuette tube (Greiner Bio-One). Hs-CRP was measured with a fully automated particle enhanced immunonephelometry (N high sensitivity CRP, Dade Behring, Marburg, Germany) on a Behring nephelometer II (BN Systems, Orchard Park, NY). Hs-CRP levels were treated as continuous variables and further coded according to the thresholds 0.5, 0.75, 1.0, 2.0, 3.0, 4.0, and 5.0 mg/dL.

After surgery, patients were followed according to the recommendations of the European Association of Urology [20]. The study point of interest was DFS, which was calculated from the date of surgery to the date of disease recurrence. Disease recurrence was defined as the development of a new tumour in a patient ≥3 months after surgery, confirmed by clinical findings, pathological specimens or radiological reports. There were no local recurrences in ipsilateral kidneys. Metachronous tumour occurrence in the contralateral kidney was not treated as recurrence, but as de novo tumour.

Continuous data are presented as the mean (sd) and were compared with t-tests and one-way anova, as appropriate. Numbers and proportions are given for categorical variables, which were compared with chi-square tests.

The Kaplan–Meier method was used to estimate DFS, and survivor functions were compared with log-rank tests. Univariable Cox regression models were fitted using continuously and categorically coded hs-CRP levels. For categorically coded hs-CRP levels, different thresholds were assessed. Multivariable Cox proportional hazards models were fitted to identify independent prognostic factors. Schoenfeld's test was used to test the proportional hazards assumption in Cox models. The PA of Cox models was assessed with Harrell's C-Index, which is an approximation of the area under the curve for time-to-event data. A C-index of 0.5 is equal to chance discrimination and a C-index of 1.0 represents a perfect discrimination. The variables of the ‘Stage, Size, Grade and Necrosis’ (SSIGN) score [21] (T classification, tumour size, grade and necrosis) were selected as multivariable base Cox model. To show an improvement in PA with inclusion of the marker hs-CRP, Cox models with hs-CRP were also fitted. PAs were compared to the base model using likelihood ratio tests. All tests were two-sided, and the level of statistical significance was set at 0.05.

Results

  1. Top of page
  2. Abstract
  3. Introduction
  4. Patients and Methods
  5. Results
  6. Discussion
  7. Conflict of Interest
  8. Details of Ethics Approval
  9. References

Patient characteristics

There were 253 male (62.8%) and 150 female (37.2%) patients, who underwent surgery at a mean (sd) age of 62.1 (14.4) years. RCCs were classified as T1, T2 and T3 in 234 (58%), 26 (6.5%) and 143 (35.5%) cases, respectively. Tumour necrosis was seen in 126 (31.3%), and 76 tumours (18.9%) were high grade (G3–4). The mean tumour size was 4.6 (2.5) cm.

Association of hs-CRP with pathology

The mean (sd) hs-CRP level was 1.32 (2.23) mg/dL. High hs-CRP levels were associated with T classification (P = 0.05), high grade disease (P < 0.001) and tumour necrosis (P = 0.003) (Table 1). Additionally, there was a weak but statistically significant correlation with tumour size (ρ = 0.108, P = 0.030).

Table 1. Associations and correlations of hs-CRP levels with standard clinical and pathological variables
VariableCRPP
Age – ρ−0.0120.804
Tumour size – ρ0.1080.030
Mean (sd):  
Gender – 0.630
Male1.38 (2.69)
Female1.28 (1.91)
T classification: 0.053
pT11.10 (1.76)
pT21.37 (2.31)
pT31.67 (2.80)
Nuclear grade: <0.001
G1–G21.06 (1.71)
G3–G42.41 (3.54)
Tumour necrosis: 0.003
Absence1.09 (1.77)
Presence1.81 (2.95)

Association of hs-CRP with DFS

After a mean (sd) follow-up of 45 (36) months (median, 43 months), 41 patients (10.1%) had developed recurrent disease after a mean (median) duration of 31 (25) months. With each unit increase in hs-CRP levels, the risk of recurrence increased by 10% (hazard ratio 1.10, P = 0.015). For prognostic sub-stratification, C-indices for the thresholds 0.5, 0.75, 1.0, 2.0, 3.0, 4.0 and 5.0 mg/dL were assessed. In this analysis, 0.5 and 0.75 mg/dL showed the highest PA (Table 2). The PA for the 0.75 mg/dL threshold was similar to the one obtained for continuously coded hs-CRP (Table 2).

Table 2. Univariable Cox models for prediction of DFS, stratified according to different hs-CRP thresholds. PA are indicated
hs-CRP threshold, mg/dLN (%) with high hs-CRPHRPPA, %
  1. HR, hazard ratio.

Continuous1.100.01569.6
<0.5 vs ≥0.5135 (33.5)4.14<0.00168.3
<0.75 vs ≥0.75108 (26.8)4.90<0.00169.6
<1 vs ≥185 (21.1)4.56<0.00167.6
<2 vs ≥257 (14.1)3.240.00162.0
<3 vs ≥343 (10.7)2.650.01457.3
<4 vs ≥437 (9.2)2.600.02256.0
<5 vs ≥530 (7.4)3.330.00456.7

Hs-CRP levels were ≥0.5 and ≥0.75 mg/dL in 135 (33.5%) and 108 (26.8%) patients, respectively. Figure 1 shows the corresponding Kaplan–Meier plots. The 5-year DFS-probabilities (se) in patients with hs-CRP ≥ 0.5 vs <0.5 mg/dL were 74 (5)% vs 92 (2)%, respectively. When the threshold was 0.75 mg/dL, the 5-year DFS-probability (se) were 69 (6)% vs 92 (2)%, respectively. Both pairwise comparisons of survival difference reached statistical significance (P < 0.001, log-rank). However, in patients with elevated hs-CRP (≥0.5 and ≥0.75 mg/dL) a higher level of hs-CRP was not associated with worse DFS (P = 0.984 and P = 0.585, respectively).

figure

Figure 1. Kaplan–Meier plots depicting DFS, stratified according to the hs-CRP thresholds 0.5 mg/dL (A) and 0.75 mg/dL (B). Solid line: below the threshold. Dashed line: above the threshold.

Download figure to PowerPoint

A multivariable base model was fitted that relied on the variables of the SSIGN score. The PA of this base model was 81.8%. Continuously coded hs-CRP did not reach statistical significance and did not add significant value in discrimination (+0.4%, P = 0.814, Table 3). In contrast, the addition of categorically coded hs-CRP led to an increased PA of 84.8% (+3.0%) and 84.9% (+3.1%) for the 0.5 and 0.75 mg/dL thresholds, respectively (Table 3). The increases in PA were statistically significant (both P < 0.001). Both categorically coded hs-CRP levels were identified as independent prognostic factors in multivariable analyses (both P < 0.001, Table 3).

Table 3. Multivariable Cox models predicting DFS. Categorically coded hs-CRP levels were identified as independent prognostic factor. The PA of the base model increased significantly after inclusion of categorically coded hs-CRP. In contrast, continuously coded CRP was not a significant predictor of prognosis
 Base modelContinuous CRPCategorical CRP (0.5 mg/dL)Categorical CRP (0.75 mg/dL)
HRsePHRsePHRsePHRseP
  1. HR, hazard ratio.

T classification            
pT11.00  1.00  1.00  1.00  
pT21.531.070.5471.561.100.5301.651.170.4751.731.220.437
pT33.351.510.0083.333.330.0083.141.430.0123.241.480.010
Grade G3–42.270.780.0172.210.770.0232.230.770.0202.150.740.027
Tumour necrosis1.100.400.7831.080.390.8271.060.380.8681.130.400.734
Tumour size1.230.090.0031.220.090.5571.210.080.0051.190.080.016
CRP1.020.040.5573.441.12<0.0013.741.21<0.001
PA, %81.8  82.2  84.8  84.9  

Discussion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Patients and Methods
  5. Results
  6. Discussion
  7. Conflict of Interest
  8. Details of Ethics Approval
  9. References

The present study validates preoperative hs-CRP levels as an independent prognostic factor for DFS after surgery for clinically localised ccRCC. Hs-CRP improved the PA of established prognostic factors.

The present data corroborate previous studies showing that elevated CRP levels are associated with worse DFS, overall survival and DFS [10-12, 22]. Michigan et al. [22] studied 257 patients who underwent surgery for localised RCC. In multivariable analysis, preoperative CRP levels remained the only independent prognostic factor for overall survival (hazard ratio 1.035, P < 0.001). Komai et al. [23] reported on a cohort of 101 patients treated by nephrectomy. An elevated CRP was defined as 0.5 mg/dL before surgery. In all, 26% of patients had elevated CRP levels, which was comparable with the proportion reported in the present study. T classification and elevated CRP were identified as significant parameters in the multivariable model. Park et al. [2] stratified patients into those with no recurrence at ≤5 years, synchronous metastasis, early recurrence (<5 years) and late recurrence (>5 years). They were able to show that preoperative CRP levels differed significantly between the four groups, although the latter three groups showed comparable data. However, CRP was not included in the multivariable analysis.

Despite this growing body of evidence suggesting that CRP is an independent prognostic factor, few studies have investigated the improvement in PA in relation to established prognostic factors. It is possible that a factor is significant in multivariable modelling, but does not improve discrimination compared with a multivariable base model [24]. Karakiewicz et al. [10] were able to show in 313 patients at all stages that CRP improved the PA of a base model by 3.7%. Iimura et al. [17] reported an increase of 2.8% and 9.9% in the developing and validation cohort of the TNM-C score, respectively. Likewise, we obtained a significant increase in our cohort of patients with localised disease. Taken together, preoperative CRP is an independent prognostic factor for patients with RCC and further increases the PA of established prognostic factors, e.g. the SSIGN score. Widespread use of this marker as an adjunct to standard prognostic factors may be advised. CRP may assist in guiding surveillance and inclusion in adjuvant clinical trials. As a consequence of the present study, we now evaluate preoperative hs-CRP in every patient routinely. The impact of hs-CRP on the time and pattern of recurrence is now being evaluated to tailor future surveillance protocols.

As one factor alone is not sufficient to predict patients’ prognosis accurately, several have been combined in prognostic models which show increased PA [25]. For example, the University of California Integrated Staging System combines Eastern Cooperative Oncology Group (ECOG) performance status, TNM classification, and grade to stratify patients into low-, intermediate- and high-risk group of recurrence and mortality [1]. The SSIGN score uses TNM classification, tumour size, grade and necrosis for patients with the cc subtype [21]. In the present study, SSIGN score was used as multivariable base model. Importantly, PA can be further improved by integrating molecular markers into prognostic models, e.g. immunohistochemical and blood-based protein markers [17, 26]. In this regard, Iimura et al. [17] developed the TNM-C score, which contains TNM and categorised CRP levels based on a threshold of 0.5 mg/dL. Inclusion of CRP led to an increase of PA compared with TNM alone. The present study shows similar data.

A problem in the interpretation of studies on CRP may be due to thresholds. Categorisation of a continuous variable creates a ‘black box’ and one is asked to accept thresholds that may be not applicable to the patient's own. However, in clinical practice one relies on thresholds, although it is obvious that CRP levels represent a continuum. For research on CRP, several thresholds have been proposed to define high and low levels [12]. Several studies used the 0.5 mg/dL threshold, which proved to have a high PA in the present analysis [17, 23]. Johnson et al. [11] studied continuous CRP levels in 130 surgically treated patients. They identified continuous CRP as an independent prognostic factor for recurrence within the first postoperative year. In the present study, CRP levels were analysed both as a continuous and a categorical variable. Both showed similar PA; however, continuously coded CRP was not a significant parameter in the multivariable analysis. We further show that continuous CRP was not significantly associated with DFS in patients with CRP elevation and thus categorisation of CRP may be appropriate. The biological reasons for this finding are unknown, but it is possible that inflammatory tumour response impacts tumour biology only up to a certain extent.

Several limitations of the present study merit discussion. Although the CRP data were collected prospectively, the present study is retrospective in nature. There were only 41 patients with recurrent disease, which limited subgroup analyses. Multicentre studies on the group of high-risk patients would be of considerable interest. CRP was determined only preoperatively, but postoperative data may be also prognostically relevant [27]. However, postoperative CRP may be impacted by surgical complications. Additionally, CRP levels are elevated by several conditions, including infectious diseases, malignant tumours, autoimmune diseases and trauma [28]. We aimed to overcome this by using stringent inclusion/exclusion criteria, but contamination of the present data is possible. There was no inter-assay variability, as the same assay was used in every patient. However, pre-analytical steps (transport, centrifugation) were not standardised. Despite these limitations, the present study clearly shows that preoperative CRP is an independent prognostic factor for localised RCC.

In conclusion, the present study validates preoperative serum hs-CRP levels as an independent prognostic factor after surgery for localised ccRCC. Hs-CRP improves the PA of established prognostic factors. Widespread use of this marker as an adjunct to standard prognostic factors may therefore be advised.

Details of Ethics Approval

  1. Top of page
  2. Abstract
  3. Introduction
  4. Patients and Methods
  5. Results
  6. Discussion
  7. Conflict of Interest
  8. Details of Ethics Approval
  9. References

The study was approved by the Ethics Committee of the Medical University of Vienna, Vienna, Austria.

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. Patients and Methods
  5. Results
  6. Discussion
  7. Conflict of Interest
  8. Details of Ethics Approval
  9. References