Original and reviewed nuclear grading according to the Fuhrman system

A multivariate analysis of 388 patients with conventional renal cell carcinoma

Authors


Abstract

BACKGROUND

The objective of the current study was to evaluate the reproducibility of the Fuhrman nuclear grading system as well as its independent predictive value in a series of patients with conventional renal cell carcinoma (RCC).

METHODS

The authors selected 388 patients who had undergone surgical treatment for conventional RCC between 1986 and 2000. Pathology slides from the selected patients were reviewed by a single pathologist, who reassigned a Fuhrman nuclear grade and assessed the presence of tumor necrosis. The pathologist was blinded to both the original pathologic diagnosis and follow-up data. The κ statistic was used to evaluate concordance between original and reviewed nuclear grades. The log-rank test was used for univariate analyses, and a Cox proportional hazards model was used for multivariate analyses.

RESULTS

The original Fuhrman nuclear grade was Grade 1 (G1) in 111 patients (28.6%), G2 in 141 patients (36.3%), G3 in 108 patients (27.8%), and G4 in 28 patients (7.3%). After pathology slide review, nuclear grades were reassigned as follows: G1 in 49 patients (12.6%), G2 in 138 patients (35.6%), G3 in 150 patients (38.7%), and G4 in 51 patients (13.1%). The grade of concordance was moderate (κ = 0.44; P < 0.001). Univariate analyses identified three separate prognostic categories defined by nuclear grade (G1 and G2 vs. G3 vs. G4). Both the original and the reviewed Fuhrman nuclear grading systems were capable of independently predicting disease-specific survival in patients with conventional RCC.

CONCLUSIONS

The interobserver reproducibility of Fuhrman nuclear grading was moderate. The substantial overlap in survival curves for G1 and G2 tumors provided an opportunity to cluster those categories, and the resulting three-tiered nuclear grading system was an independent predictor of cause-specific survival in patients with conventional RCC. Other independent predictors of survival included pathologic stage and tumor necrosis status. Cancer 2005. © 2004 American Cancer Society.

Several systems have been proposed for the grading of renal cell carcinoma (RCC). It has been shown that nuclear grading systems are more predictive of disease-specific survival in patients with RCC than are grading systems based on the cytoplasmic and/or architectural features of tumor cells.1–4 Classification of RCC into four categories according to nuclear size and the presence of nucleoli, as reported by Fuhrman, is currently the most widely used grading protocol in North America and Europe.2, 4, 5

The most relevant drawback of the Fuhrman nuclear grading system is its lack of efficacy in establishing prognostic groups.4, 6 Several studies have reported differences in disease-specific survival after separating patients with Grade 1 (G1) and G2 tumors from patients with G3 and G4 tumors.7–10 More recently, however, significant differences in prognosis between patients with G3 RCC and patients with G4 RCC have been reported in studies involving larger patient cohorts.11, 12 Such variability has been ascribed primarily to the moderate interobserver reproducibility of nuclear grading and to grade migration caused by suboptimal tissue fixation.5, 13 Tumor histotype is another possible confounding factor. Most investigators have affirmed the prognostic validity of nuclear grading only for conventional and papillary RCC.4 Furthermore, a recent report showed that the specific survival probability for patients who had conventional RCC was significantly poorer than the corresponding probabilities for patients who had papillary tumors and patients who had chromophobe tumors.14 Therefore, both the reproducibility and the prognostic value of nuclear grading should be assessed only in large series of patients with the same tumor histotype. To our knowledge, only a limited number of recent studies have met these requirements.15, 16 In the current study, the reproducibility of Fuhrman nuclear grading after pathology slide review and the independent predictive value of both the original and the reviewed systems were evaluated in a series of patients with conventional RCC.

MATERIALS AND METHODS

From the Kidney Cancer Database of the Department of Urology at the University of Verona (Verona, Italy), we selected all 491 patients who had undergone surgical treatment for localized RCC between 1986 and 2000. In the current study, we investigated only the 388 patients who had conventional RCC (79%). Fifty-seven patients (11.6%) with papillary RCC, 25 patients (5.1%) with chromophobe RCC, 8 patients (1.6%) with collecting duct RCC, and 13 patients (2.6%) with unclassified tumors were excluded. For each patient, the following clinical variables were considered: age, gender, Eastern Cooperative Oncology Group (ECOG) performance status, mode of presentation, and presence of synchronous metastases. All tumors that were diagnosed during abdominal imaging studies for signs and symptoms that were unrelated to RCC were classified as incidental. Hematuria, flank pain, palpable flank masses, and systemic symptoms suggesting the presence of advanced-stage disease (weight loss, fever, paraneoplastic syndromes) were used to identify symptomatic patients.

In all cases, clinical staging included abdominal computed tomography (CT) scans and chest X-rays. Bone scans and brain CT scans were obtained only when indicated by signs and symptoms.

Two hundred ninety patients (74.7%) underwent radical nephrectomy, 73 patients (18.8%) underwent elective nephron-sparing surgery (NSS), and 25 patients (6.4%) underwent imperative NSS. For patients with normal contralateral kidneys, elective NSS was indicated only if single peripheral tumors measuring ≤ 4 cm in greatest dimension were present. Patients who had bilateral tumors or neoplasia that involved anatomically or functionally solitary kidneys underwent imperative NSS.

The following pathologic variables were examined: tumor size (cm), original Fuhrman nuclear grade, local extension of primary tumor, and regional lymph node involvement. Data on these variables were obtained from the surgical pathology reports in patients' clinical records. The original pathologic diagnoses were made by five different pathologists. Pathologic stage was reassigned according to the 2002 TNM system.17

The oncologic follow-up schedule included abdominal imaging twice a year (alternating ultrasound and CT scans) and chest X-rays once yearly for the first 5 years. Abdominal imaging studies and chest X-rays were scheduled once yearly thereafter.

Data on survival were obtained from the clinical files at our center and, when necessary, by contacting patients' general practitioners or relatives or by searching death records.

Pathology slides were reviewed by a single uropathologist who had experience dealing with RCC (G.M.). The pathologist reassigned Fuhrman nuclear grades and reassessed the presence of tumor necrosis. Nuclear grading was based on the highest-grade tumor area identified. The pathologist was blinded to both original pathologic diagnoses and follow-up data.

The Pearson chi-square test and analysis of variance were used to compare categoric and continuous variables, respectively. The κ statistic, which is a measure of agreement between observers that corrects for chance agreement, was used to evaluate the concordance between original and reviewed nuclear grades. Concordance was considered to be fair when κ values ranged from 0.00 to 0.20, moderate when κ values ranged from 0.21 to 0.45, substantial when κ values ranged from 0.46 to 0.75, and almost perfect when κ values ranged from 0.76 to 0.99.18

The survival interval was defined as the time elapsed between surgery and the last clinical evaluation or death. Survival curves were estimated using the Kaplan-Meier method. Patients who remained alive or who died of other causes were censored (disease-specific survival). The log-rank test was used for the comparison of survival curves and for univariate analysis. A Cox proportional hazards model was used for multivariate analysis. We developed three Cox proportional hazards models: Model A, which included all of the clinical and pathologic variables that were significant in the univariate analysis, except for Fuhrman nuclear grade; Model B, which included the variables from Model A together with the original Fuhrman nuclear grade; and Model C, which included the variables from Model A together with the reviewed Fuhrman nuclear grade. The log-likelihood ratio test was used to compare Cox proportional hazards models. In all statistical analyses, a 2-sided P value < 0.05 was considered significant. All data were analyzed using Statistical Package for the Social Sciences software (Version 10.1; SPSS Inc., Chicago, IL).

RESULTS

The mean patient age (± standard deviation) was 60.38 years ± 11.85 years (median, 61 years; range, 22–86 years). Two hundred eighty-seven patients (74%) were male, and 101 were (26%) female. ECOG performance status was 0 in 362 patients and ≥ 1 in 26 patients. Tumor presentation was incidental in 226 patients (58.2%) and symptomatic in 162 patients (41.8%). At the time of the initial diagnosis, 37 patients (9.5%) had distant metastases.

The mean pathologic tumor size (± standard deviation) was 5.8 cm ± 3.2 cm (median, 5 cm; range, 1–19 cm). One hundred thirty-seven tumors (35.3%) were classified as pathologic T1a (pT1a), 91 tumors (23.5%) were classified as pT1b, 50 tumors (12.9%) were classified as pT2, 30 tumors (7.7%) were classified as pT3a, 75 tumors (19.3%) were classified as pT3b–pT3c, and 5 tumors (1.3%) were classified as pT4. Regional lymph node involvement was observed in 16 patients (4.1%), and tumor necrosis was identified in 45 (11.6%).

Fuhrman nuclear grade was assessed after reviewing a median of 4 slides (range, 2–15 slides). The mean number of evaluated slides (± standard deviation) per tumor was 3.52 ± 1.47 for tumors measuring ≤ 4.0 cm; 4.52 ± 1.81 for tumors measuring 4.1–7.0 cm; and 6.14 ± 2.58 for tumors measuring > 7.0 cm (P < 0.001).

The original Fuhrman nuclear grade was G1 in 111 patients (28.6%), G2 in 141 patients (36.3%), G3 in 108 patients (27.8%), and G4 in 28 patients (7.3%). After pathology slide review, the nuclear grades were reassigned as follows: G1 in 49 patients (12.6%), G2 in 138 patients (35.6%), G3 in 150 patients (38.7%), and G4 in 51 patients (13.1%). Concordance was moderate (κ = 0.44; P < 0.001). Tumors that originally were classified as G1 were upgraded by 1 grade in 38.7% of cases, by 2 grades in 18.9% of cases, and by 3 grades in 2.7% of cases. Tumors originally classified as G2 were upgraded by 1 grade in 34% of cases and by 2 grades in 4.3% of cases. The grading of tumors originally classified as G3 and G4 remained unchanged in 73.1% and 89.3% of cases, respectively (Table 1).

Table 1. Concordance between Original and Reviewed Fuhrman Nuclear Gradesa
Original gradeReviewed grade: no. of patients (%)
Grade 1 (n = 49)Grade 2 (n = 138)Grade 3 (n = 150)Grade 4 (n = 51)
  • a

    κ = 0.44; P < 0.001.

Grade 1 (n = 111)44 (39.6)43 (38.7)21 (18.9) 3 (2.7)
Grade 2 (n = 141) 3 (2.1)84 (59.6)48 (34.0) 6 (4.3)
Grade 3 (n = 108) 2 (1.9)10 (9.3)79 (73.1)17 (15.7)
Grade 4 (n = 28) 0 (0.0) 1 (3.6) 2 (7.1)25 (89.3)

Both the original and reviewed Fuhrman nuclear grades were significantly correlated with mean pathologic tumor size, pathologic stage, venous involvement, and the presence of tumor necrosis (Table 2). Tables 3 and 4 show correlations between nuclear grade and both the pathologic stage and the pathologic size of the primary tumor.

Table 2. Correlation between Pathologic Variables and Original and Reviewed Fuhrman Nuclear Grades
 Pathologic size (cm)Pathologic stage > II (%)Venous involvement (%)Tumor necrosis (%)
Original grade    
 Grade 1 (n = 111)4.3511.79.04.5
 Grade 2 (n = 141)5.5124.813.58.5
 Grade 3 (n = 108)6.9751.927.816.7
 Grade 4 (n = 28)9.1775.057.135.7
 P value< 0.001< 0.001< 0.001< 0.001
Reviewed grade    
 Grade 1 (n = 49)4.136.16.14.1
 Grade 2 (n = 138)5.0620.310.93.6
 Grade 3 (n = 150)6.2042.024.713.3
 Grade 4 (n = 51)8.6060.839.235.3
 P value< 0.001< 0.001< 0.001< 0.001
Table 3. Correlation of Original and Reviewed Fuhrman Nuclear Grades with Local Extension of the Primary Tumor
 Local pathologic stage (pT, 2002): no. of patients (%)
pT1a (n = 137)pT1b (n = 91)pT2 (n = 50)pT3a (n = 30)pT3b,c (n = 75)pT4 (n = 5)
  • a

    P < 0.001.

Original gradea      
 Grade 1 (n = 111)63 (56.8)27 (24.3)8 (7.2)3 (2.7)10 (9.0)0 (0.0)
 Grade 2 (n = 141)50 (35.5)39 (27.7)22 (15.6)11 (7.8)19 (13.5)0 (0.0)
 Grade 3 (n = 108)23 (21.3)22 (20.4)17 (15.7)12 (11.1)30 (27.8)4 (3.7)
 Grade 4 (n = 28)1 (3.6)3 (10.7)3 (10.7)4 (14.3)16 (57.1)1 (3.6)
Reviewed gradea      
 Grade 1 (n = 49)35 (71.4)5 (10.2)6 (12.2)0 (0.0)3 (6.1)0 (0.0)
 Grade 2 (n = 138)60 (43.5)37 (26.8)18 (13.0)6 (4.3)15 (10.9)2 (1.4)
 Grade 3 (n = 150)35 (23.3)42 (28.0)16 (10.7)17 (11.3)37 (24.7)3 (2.0)
 Grade 4 (n = 51)7 (13.7)7 (13.7)10 (19.6)7 (13.7)20 (39.2)0 (0.0)
Table 4. Correlation of Original and Reviewed Fuhrman Nuclear Grades with Pathologic Tumor Size
 Pathologic tumor size: no. of patients (%)
≤ 4.0 cm (n = 153)4.1–7.0 cm (n = 133)> 7.0 cm (n = 102)
  • a

    P < 0.001.

Original gradea   
 Grade 1 (n = 111)67 (60.4)33 (29.7)11 (9.9)
 Grade 2 (n = 141)55 (39.0)54 (38.3)32 (22.7)
 Grade 3 (n = 108)30 (27.8)38 (35.2)40 (37.0)
 Grade 4 (n = 28)1 (3.6)8 (28.6)19 (67.9)
Reviewed gradea   
 Grade 1 (n = 49)35 (71.4)7 (14.3)7 (14.3)
 Grade 2 (n = 138)65 (47.1)49 (35.5)24 (17.4)
 Grade 3 (n = 150)46 (30.7)63 (42.0)41 (27.3)
 Grade 4 (n = 51)7 (13.7)14 (27.5)30 (58.8)

The mean follow-up duration (± standard deviation) was 64 months ± 46.2 months (median, 56 months; range, 6–205 months). On univariate analysis, the variables that were associated with disease-specific survival in patients with conventional RCC were ECOG performance status ≥ 1 (P = 0.007), symptomatic tumor presentation (P = 0.04), metastases at the time of diagnosis (P < 0.0001), local extension of the primary tumor (P < 0.0001), pathologic regional lymph node involvement (P < 0.0001), the presence of tumor necrosis (P < 0.0001), and both original (P < 0.0001) and reviewed (P < 0.0001) Fuhrman nuclear grade (Table 5). Univariate analysis allowed us to identify three prognostic categories defined by local extension of the primary tumor (pT1a vs. pT1b/pT2 vs. > pT2), as well as three prognostic categories defined by nuclear grade (G1–2 vs. G3 vs. G4) (Figs. 1, 2).

Table 5. Five-Year and 10-Year Disease-Specific Survival Probabilities: Univariate Analysis
VariableNo. of patientsDisease-specific survival (%)Log-rank P value
5 yr rate10 yr rate
  1. ECOG: Eastern Cooperative Oncology Group; pT: pathologic tumor status; pN: pathologic lymph node status; G: Grade.

Age (yrs)    
 20–402495.895.80.07
 41–6016084.670.9 
 61–8019676.866.6 
 > 80880.0 
Gender    
 Male28780.669.20.86
 Female10183.972.6 
ECOG performance status    
 036282.170.80.007
 ≥ 12671.459.5 
Mode of presentation    
 Incidental22684.273.40.04
 Symptomatic16277.566 
Distant metastases    
 Absent35186.676.3< 0.0001
 Present3731.59.8 
Primary tumor extension (pT, 2002)    
 pT1a137100.094.4< 0.0001
 pT1b9189.485.5 
 pT25082.265.5 
 pT3a3045.913.0 
 pT3b–c7554.038.9 
 pT4540.0 
Tumor necrosis    
 Absent34386.274.7< 0.0001
 Present4547.737.9 
Lymph node involvement (pN)    
 Absent44183.073.2< 0.0001
 Present5041.612.5 
Original nuclear grade    
 G111195.488.1< 0.0001
 G214190.776.6 
 G310865.353.3 
 G42840.0 
Reviewed nuclear grade    
 G149100.094.7< 0.0001
 G213890.184.2 
 G315077.160.9 
 G45154.745.9 
Figure 1.

Kaplan–Meier disease-specific survival according to original nuclear grade (log-rank test: P < 0.0001). Analysis of the disease-specific survival probabilities for patients with Grade 1 (G1) and G2 renal cell carcinoma (RCC) yielded statistically overlapping results (log-rank test: P = 0.06). Statistically significantly different disease-specific survival probabilities were observed between patients with G2 and G3 RCC (log-rank test: P < 0.001) and between patients with G3 and G4 RCC (log-rank test: P < 0.02).

Figure 2.

Kaplan–Meier disease-specific survival according to reviewed nuclear grade (log-rank test: P < 0.0001). Analysis of the disease-specific survival probabilities for patients with Grade 1 (G1) and G2 renal cell carcinoma (RCC) yielded statistically overlapping results (log-rank test: P = 0.051). Statistically significantly different disease-specific survival probabilities were observed between patients with G2 and G3 RCC (log-rank test: P < 0.004) and between patients with G3 and G4 RCC (log-rank test: P < 0.009).

Table 6 summarizes data regarding the three Cox proportional hazards models that were developed using clinicopathologic variables found to be significant on univariate analysis (Table 4). Both the original (Model B) and reviewed (Model C) Fuhrman nuclear grades were able to predict disease-specific survival independently in patients with conventional RCC. Specifically, Model B (chi-square test, 10.36; P = 0.001) and Model C (chi-square test, 10.88; P = 0.0009) both were significantly more predictive than Model A.

Table 6. Multivariate Analysis: Cox Proportional Hazards Models
Prognostic variableCategoryHR95% CIP value
  1. HR: hazard ratio; 95% CI: 95% confidence interval; pT: pathologic tumor status; pN: pathologic lymph node status; G: Grade.

Model A (log likelihood = 689.139)    
 Performance status0/≥ 11.8520.899–3.8170.095
 Mode of presentationIncidental/symptomatic0.8970.559–1.4390.652
 Tumor extension (pT)pT1a/pT1b–2/> pT23.6372.425–5.4550.000
 Lymph node involvementpN0/pN+2.5991.367–4.9420.004
 MetastasesM0/M+6.7844.004–11.4950.000
 Tumor necrosisAbsent/present3.1851.918–5.2890.000
Model B (log likelihood = 683.958)    
 Performance status0/≥ 11.7590.852–3.6320.127
 Mode of presentationIncidental/symptomatic0.9090.569–1.4540.692
 Tumor extension (pT)pT1a/pT1b–2/> pT23.2502.153–4.9060.000
 Lymph node involvementpN0/pN+2.2591.172–4.3540.015
 MetastasesM0/M+6.8544.053–11.5930.000
 Tumor necrosisAbsent/present2.5351.469–4.3720.001
 Original nuclear gradeG1–2/G3/G41.4861.057–2.0890.023
Model C (log likelihood = 683.699)    
 Performance status0/≥ 11.7190.831–3.5560.144
 Mode of presentationIncidental/symptomatic0.9320.579–1.5000.773
 Tumor extension (pT)pT1a/pT1b–2/> pT23.2702.181–4.9020.000
 Lymph node involvementpN0/pN+2.5511.337–4.8670.005
 MetastasesM0/M+7.1124.183–12.0920.00
 Tumor necrosisAbsent/present2.4701.420–4.2960.001
 Reviewed nuclear gradeG1–2/G3/G41.5461.070–2.2340.020

DISCUSSION

The Fuhrman nuclear grading system is used by most uropathologists in North America and Europe.19 Such widespread diffusion and popularity are attributable primarily to the system's simplicity and its proven correlation with other pathologic variables.20 This 4-tiered grading system is essentially based on nuclear size and morphology and on the presence or absence of nucleoli. G1 tumors consist of cells with small (approximately 10 μm), round, uniform nuclei with inconspicuous or absent nucleoli; G2 tumors have larger nuclei (approximately 15 μm) with irregular morphology and small nucleoli when examined under high power (×400 magnification); G3 tumors have even larger nuclei (approximately 20 μm) with irregular outlines and large, prominent nucleoli that are evident even at low power (×100 magnification); and G4 tumors differ from G3 lesions in that they contain bizarre, multilobed nuclei and heavy chromatin clumps.2

Several recent studies have shown that the Fuhrman nuclear grading system is capable of predicting disease-specific survival in patients with RCC, regardless of pathologic stage.12, 15, 21–23 However, the reported 5-year disease-specific survival probabilities are extremely variable; stratified by grade, they are 50–97% for patients with G1 tumors, 30–90% for patients with G2 tumors, 10–78% for patients with G3 tumors, and 9–66% for patients with G4 tumors (Table 7). Moreover, prognostic stratification has been demonstrated only after patients are clustered into two or three different categories. This is the most relevant drawback of the four-tiered nuclear grading system, as, in theory, each grade should identify patients who have significantly different tumor-related outcomes, and survival advantage cutoff points should form a continuum.6

Table 7. Disease-Specific Survival Probabilities Categorized by Nuclear Grade: Critical Literature Analysis of Studies Using Four-Tiered Nuclear Grading Systems
ReferenceNo. of patientsTNM stageHistotypeFive yr survival rate (%)Grade clustering
G1G2G3G4
  • G: Grade; NR: not reported.

  • a

    G1–2.

  • b

    G1–3.

  • c

    G3–4.

Skinner et al., 19711272I–IVAll75655626NR
Furhman et al., 19822103I–IVAll65303210G1 vs. G2–4
Medeiros et al., 19887121I–IVAll85702123G1–2 vs. G3–4
Grignon et al., 19898103I–IVAll88a8043G1–2 vs. G3–4
Green et al., 19892555IAll 91b9G1–3 vs. G3–4
Munichor et al., 1992979I–IVAll505510.514.3G1–2 vs. G3–4
Gelb et al., 19932682IAllNRNRNRNRG1–2 vs. G3–4
Bretheau et al., 199510190I–IVAll76725135G1–2 vs. G3–4
Usubuton et al., 199824165I–IVAll97837866G1–2 vs. G3–4
Tsui et al., 200022643I–IVAll896546cG1 vs. G2 vs. G3–4
Ficarra et al., 200111333I–IVAll94865931G1–2 vs. G3 vs. G4
Frank et al., 2002121801I–IVAllNRNRNRNRG1–2 vs. G3 vs. G4
Minervini et al., 200227213I–IVAll968760cG1–2 vs. G3–4
Amin et al., 200228405I–IVAll100948035G1–2 vs. G3 vs. G4
Mejean et al., 20031588I–IVPapillary91a51cG1–2 vs. G3–4
Cheville et al., 2003141985I–IVClear cellNRNRNRNRG1–2 vs. G3 vs. G4
Cheville et al., 200314270I–IVPapillaryNRNRNRNRG1–2 vs. G3 vs. G4

The earliest data reported by Fuhrman et al. highlighted the need to cluster patients who had G2, G3, and G4 tumors into a single group on the basis of their differing prognosis compared with patients who had G1 tumors.2 Most of the subsequently published articles reported significant survival differences only after separating patients with G1 and G2 tumors from patients with G3 and G4 tumors.7–10, 24 The small number of patients analyzed in those studies may explain this finding (Table 7). More recent retrospective analyses involving larger series have allowed identification of three categories of patients with different outcomes. Tsui et al. reported significant survival differences between patients with G1 tumors and patients with G2 tumors and also between patients with G2 tumors and patients with G3–4 tumors.22 In other series, even when the survival probabilities for patients with G1 and G2 RCC overlapped, significantly different outcomes were observed among patients with G2, G3, and G4 tumors.11, 12, 14, 21, 28

Such variability may be explained by the moderate interobserver reproducibility of Fuhrman nuclear grading, suboptimal tissue fixation, and the inclusion of a variety of tumor histotypes. Bretheau et al. reported high grading concordance among different pathologists.10 However, their data have not been confirmed in subsequent reports, which showed moderate interobserver agreement both in 4-tiered (κ = 0.29–0.33) and 2-tiered (κ = 0.45) systems.10, 20, 29 Interobserver variability may depend on the heterogeneity of the RCC, as areas of different grades often are found within a given tumor.6, 20 In the current study, nuclear grade reassignment after pathology slide review confirmed that the reproducibility of grading was moderate.

Nuclear grading should be based on the highest-grade area identified within the tumor. However, the minimum size required for such an area to be considered significant has not yet been standardized.4, 13, 20 Furthermore, the extent of tumor sampling may be a factor in the identification of the highest-grade area of a given lesion. In general, larger tumors should subjected to more extensive sampling. In fact, in the current series, we observed a significant correlation between mean tumor size and number of slides assessed.

The specimen fixation protocol used also can markedly affect the assignment of nuclear grades. Suboptimal fixation can make nuclear chromatin appear coarser and less hyperchromatic and can alter nucleolar morphology, thereby leading to grade migration.13

Another possible explanation for the variability of the reported disease-specific survival probabilities involves tumor histotype. The International Union Against Cancer and the American Joint Committee on Cancer recommend assignment of nuclear grades only for patients with conventional and papillary RCC.4 At present, only limited data are available regarding the validity of nuclear grading in chromophobe tumors.16 Furthermore, Cheville et al. recently found that patients who had conventional RCC had a significantly poorer prognosis compared with patients who had papillary or chromophobe RCC, regardless of pathologic tumor stage.14 Nonetheless, most investigators who have analyzed the prognostic relevance of nuclear grade have not considered stratification according to histotype. Recent results reported by Cheville et al. indicated that patients with G1–2 tumors had significantly better outcomes compared with patients who had G3 tumors and that patients with G3 tumors had a more favorable prognosis compared with patients who had G4 tumors; these findings were true for patients with conventional, papillary, and chromophobe RCC.14 Similar findings were reported by Mejean et al. in patients with papillary RCC.15 Our data confirm the need to cluster patients with G1 and G2 tumors into a single subgroup. Nuclear grading was found to be predictive of disease-specific survival, and its prognostic value was independent of other variables, such as tumor necrosis and pathologic stage, that were identified as significant on multivariate analysis.

Our multivariate analyses justify the inclusion of nuclear grade in predictive models, such as those proposed recently by teams at the University of California–Los Angeles (UCLA; Los Angeles, CA) and the Mayo Clinic (Rochester, MN).12, 30 On the basis of our findings, however, we hypothesize that the SSIGN (‘Stage, Size, Grade, and Necrosis’) scoring system may have been more suitable for patients in the current study. Our ability to use the UCLA Integrated Staging System in the current series was markedly limited by the lack of independent predictive value of ECOG performance status on multivariate analysis. This lack of predictive value was attributable to the characteristics of the patients in the study cohort, who, unlike patients in the UCLA cohort, primarily had nonmetastatic disease.22

In conclusion, in the current study, the concordance between original and reviewed Fuhrman nuclear grades was found to be moderate. In addition, both grading systems were independently predictive of disease-specific survival in patients with conventional RCC. On multivariate analysis, the pathologic stage of the primary tumor, lymph node status, distant metastases, and tumor necrosis also were identified as independent predictors of disease-specific survival.

Ancillary