Prognostic factors in renal cell cancer

Authors


Won-Ho Park, St Lukes Cancer Centre, Royal Surrey County Hospital, Egerton Road, Guildford, Surrey GU2 7XX, UK.
e-mail: wpark@doctors.org.uk

Abbreviations
IVC

inferior vena cava

NSS

nephron-sparing surgery

ECOG

Eastern Cooperative Oncology Group

VHL

Von Hippel-Lindau

HIF

hypoxia-inducible factor

CAIX

carbonic anhydrase IX.

INTRODUCTION

RCC accounts for 3% of male cancers and 2% of female cancers in the UK, with 6665 new cases diagnosed in 2002 [1]. If detected early RCC is a surgical and potentially curable disease, but 20–30% of patients present with metastatic disease, and metastases develop in 20–40% of patients who have had nephrectomy for localized disease [2]. Once metastases develop, the prognosis for long-term survival is much poorer, with 5-year survival rates of <10% in patients with stage IV disease [3]. RCC is a chemotherapy-resistant disease and the mainstay of treatment has until recently been immunotherapy with interferon-α and/or interleukin-2. These have been shown to produce responses of 10–20%, some durable [4]. Recently the multi-targeted tyrosine kinase inhibitors, e.g. sorafenib and sunitinib, have produced stabilization or shrinkage of disease in up to 80% of patients [5,6]. However, the intent of treatment remains palliation of disease.

The lack of curative systemic therapy coupled with the highly variable natural history of RCC highlights the need for identifying patient and disease factors that might be used to predict outcome. The development of a reliable prognostic tool could be used to tailor specific treatment strategies and follow-up regimens for patients, and provide them with valuable information and counselling about the prognosis. Prognostic factors must be considered when conducting trials, to ensure comparability of treatment populations and therefore the proper interpretation of outcome can be ascertained.

CURRENT SYSTEMS

PATHOLOGICAL FACTORS

TNM staging system

The TNM staging classification incorporates tumour size, node involvement and metastatic spread, as with other cancers, but also considers the prognostic impact of tumour thrombus and the extent of spread into the inferior vena cava (IVC). Using this tool the 5-year survival rate is reported to be 91%, 74%, 67% and 32% for stages I to IV [7]. However, there has been much debate about the prognostic accuracy of this system, in particular for the precise thresholds of tumour size and the significance of tumour thrombus, perirenal/sinus fat involvement, adrenal involvement and lymphadenopathy.

Tumour size

Several studies examined the T1 tumour size threshold; the revision of the TNM staging system in 2002 added a subdivision of T1 tumours into T1a and T1b, using a 4-cm threshold [8]. This has particular significance with the increasing prevalence of nephron-sparing surgery (NSS), where studies show the clinical effectiveness and safety of NSS in tumours of <4 cm [9].

More recently Leibovich et al.[10] retrospectively reviewed 91 patients treated with NSS and 841 treated with radical nephrectomy for 4–7 cm RCC. When adjusted for features known to be associated with RCC prognosis, e.g. perinephric fat, renal vein and regional lymph node involvement, nuclear grade, histological tumour necrosis and histological subtype, they found no statistical differences in outcome between patients treated with NSS or radical nephrectomy for 4–7 cm RCC. Emerging data suggest that the width of tumour margin bears no relationship to the risk of disease progression, as long as complete resection is achieved [11,12].

Adrenal involvement and tumour thrombus

Tsui et al.[7] reported the 5-year cancer-specific survival rate of stage T3 disease to be 42%, vs 83% and 57% for stages T1 and T2, respectively. However, recently several studies were reported that examined the prognostic impact of fat and adrenal invasion, and the significance of tumour thrombus level. These studies have called into question the prognostic accuracy of Stage T3 according to the 2002 TNM staging system.

Stage T3a is currently defined as tumour invasion into the adjacent adrenal gland and/or infiltration of the perinephric fatty tissue but not extending beyond Gerota’s fascia. Han et al.[13] investigated the clinical behaviour of pT3a tumours in patients after nephrectomy. They reported that there was significantly worse survival in 27 patients with adrenal involvement than in 187 similarly staged tumours with no adrenal involvement. There were no survivors at 5 years in patients with direct adrenal involvement, compared with 36% in patients with perinephric/renal sinus fat involvement. They also noted that the survival of patients with adrenal gland involvement was similar to those tumours with involvement of adjacent organs, currently staged as T4. These findings prompted the suggestion that direct adrenal gland invasion be staged as T4.

RCC invades the venous system in 4–10% of newly diagnosed cases [14,15]. The 2002 TNM classification stratifies disease with tumour thrombus into two groups, T3b being that with tumour thrombus below the diaphragm and T3c above. Moinzadeh and Libertino [15], in 2004, retrospectively reviewed 153 cases of RCC with vascular venous extension, to test the assertion that level of tumour thrombus affects survival. They found that patients with tumour thrombus in the IVC below the level of the diaphragm had significantly lower survival than those patients with tumour thrombus in the renal vein. However, extent of tumour thrombus in the IVC had no statistically significant effect on cancer-specific survival rates.

Kim et al.[14] compared 216 patients who had had nephrectomy and tumour thrombectomy with 653 patients who had nephrectomy with no thrombectomy. The 3-year recurrence-free rates for RCC were 78% for patients with no thrombus, 60% if there was renal vein involvement (T3b), 46% if there was IVC involvement below the diaphragm (T3b) and 34% if there was IVC involvement above the diaphragm (T3c). They found no significant survival difference between T3a and T3b tumours but worse survival rates for patients with T3c tumours (P = 0.009). Although the presence of venous thrombus was associated with a higher risk of recurrence and lower cancer survival rates, there was no significant difference when the data were corrected for clinical and pathological features known to be predictors of prognosis. The group therefore concluded that tumour grade, tumour stage and Eastern Cooperative Oncology Group (ECOG) performance status were better predictors of outcome than level of tumour thrombus involvement.

More recently, Leibovich et al.[16] reviewed 697 patients including pT3a, pT3b, pT3c and pT4 tumours. In agreement with Moinzadeh and Libertino [14], but in contrast to Kim et al.[15], they found that among patients with pT3b tumours, those with tumour thrombus restricted to the renal vein had significantly better survival than those with tumour thrombus extension into the IVC. Similarly they noted that patients with perinephric/renal sinus fat invasion had half the cancer-specific survival rate than patients with no fat invasion. They therefore proposed a further subclassification of T3 tumours based on tumour thrombus and fat invasion, and showed a significant improvement in prognostic accuracy for their patients.

In an interesting recent study of 227 patients with pT3–pT4 RCC [17], a univariate analysis of pathological stage according to the 2002 TNM staging criteria failed to show any statistically significant difference in survival. These authors reclassified tumour stage into three categories combining perirenal fat invasion, tumour thrombus level, adrenal gland and involvement of Gerota’s fascia as factors, and on multivariate analysis showed their new classification to be an independent predictive variable. The studies by Leibovich et al.[16] and Ficarra et al.[17] show that improvements can be made to current T3–T4 staging but further work is needed to optimize staging.

Lymphadenopathy

Pantuck et al.[18] reviewed lymphadenopathy in 900 patients and its impact on survival and response to immunotherapy. Their findings supported the assertion that lymph node-positive status (N+) is associated with larger, higher grade and more locally advanced tumours. In addition they noted that not only is N+ disease more likely to be associated with metastases, but also that patients with N+ disease and metastases had significantly worse prognosis than patients with metastatic disease alone. When reviewing the response rate to immunotherapy, patients with N+ disease had poorer response rates. Interestingly, 112 patients with N+ disease who had had lymph node dissection had a statistically significant survival advantage over 17 who had not had lymph node dissection. However, they did not identify any survival advantage with lymph node dissection in those patients who clinically had negative lymph nodes.

PATHOLOGICAL FACTORS NOT INCLUDED IN THE TNM STAGING SYSTEM

Tumour grade

The tumour grade of clear-cell RCC is a well recognized independent prognostic indicator, and the most commonly used remains that developed by Fuhrman et al. [19] This system was validated in a study by Tsui et al.[7], with 5-year cancer-specific survival rates of 89% (Grade 1), 65% (Grade 2) and 46% (Grades 3 and 4). Grading of papillary and chromophobe RCC remains controversial.

Histological classification

There are four main histological subtypes of RCC. The current classification system is based on the Heidelberg classification system, which subdivides RCC into clear cell, papillary, chromophobe, collecting duct and unclassified subtypes, which account for 70–80%, 10–20%, 5% and <1% of RCC, respectively [20]. In a retrospective study of 2385 patients treated with radical nephrectomy the 5-year cancer-specific survival rates in clear cell RCC were 69%, vs 87% for papillary and 87% for chromophobe subtypes, respectively [21]. A more recent multicentre study of 4063 patients showed a trend toward better survival rates with chromophobe tumours than with clear cell and papillary subtypes. However, after adjusting for TNM stage, Fuhrman grade and ECOG performance status, no significant survival difference remained among the three groups [22].

Sarcomatoid features can be associated with any of the histological subtypes and is recognized to be associated with poor response rates to immunotherapy and poor outcome [21].

Tumour necrosis

The presence of tumour necrosis was reported in several studies to be an adverse prognostic feature of RCC, conferring a two- to three-fold higher risk of death from RCC than in those patients with no tumour necrosis [21,23]. In a study of 2385 patients, Cheville et al.[21] found that although tumour necrosis was a significant predictor of poor prognosis in patients with clear cell RCC the same was not true of papillary cell RCC.

NON-PATHOLOGICAL FACTORS

Performance status

Performance status is a convenient measure of the overall impact of disease on a patient and was recognized very early as an independent predictor of survival. Since then the prognostic significance of performance status has been repeatedly shown in a series of analyses of large datasets [3,23–25].

Thrombocytosis

In a study of 804 patients with RCC thrombocytosis was found to be strongly correlated with T stage, Fuhrman grade, tumour size, nodal invasion and the presence of distant metastases [26]. Thrombocytosis has an impact on survival in both localized and metastatic tumours, and was retained as an independent prognostic factor after multivariate analysis. The 5-year survival rate was 70% for patients with platelet counts of 450 000/mm3, vs 37% when the platelet count was >450 000/mm3.

Neutrophil count

Atzpodien et al.[27] studied 425 patients with metastatic RCC and in a multivariate analysis found that, the largest and most significant association with a poor outcome was in patients with elevated neutrophil count (≥6500 cells/µL). Donskov et al.[28] corroborated this finding more recently in a study of 120 patients, but also noted that the presence of intratumoral neutrophils was another significant independent risk factor.

PROGNOSTIC MODELS

Many groups have endeavoured to identify clinical features that might predict survival and response to therapy. Elson et al.[24] in 1988 retrospectively reviewed 610 cases of recurrent and metastatic RCC; multivariate analysis showed that ECOG performance status, time from initial diagnosis, number of metastatic sites, previous cytotoxic chemotherapy and recent weight loss were important indicators of survival. Using a simple function of the number of these factors, they were able to divide the patients into five prognostic groups. The median survival for the five subgroups was 12.8, 7.7, 5.3, 3.4 and 2.1 months, respectively.

More recently various large analyses were reported that combine clinical, pathological and in some cases laboratory factors to generate a variety of algorithms. These algorithms have greatly advanced our ability to prognosticate for patients with RCC, and are being incorporated into clinical trial design.

RESECTED PRIMARY TUMOURS

Kattan et al.[25] developed a nomogram to predict the likelihood of recurrence for up to 5 years after surgery for RCC. The predictor variables used were patient symptoms, histology, tumour size and pathological stage. The nomogram thereby developed had an area under the receiver operating characteristic curve of 0.74 (where a measure of 0.5 represents no discriminating ability and 1.0 represents perfect discrimination). Of the variables on the nomogram only tumour size and histology were statistically significant at 5% accuracy. Statistically insignificant predictor variables were not omitted from the final model, due to the bias this would have on the remaining predictor variables and subsequent effect on predictor accuracy. Pathological stage was not found to be statistically discriminating but this study was done before the 2002 revision of the TNM staging system, and it would be interesting to see what impact this change would have on the predictive strength of this nomogram.

Frank et al.[23] conducted a study of 1801 patients with unilateral clear cell RCC treated with previous nephrectomy, and developed an outcome prediction model in RCC based on 1997 TNM cancer stage, tumour size, nuclear grade and histological tumour necrosis. These factors were combined into the SSIGN (stage, size, grade and necrosis) algorithm with an increasing SSIGN score correlating with poorer 5-year survival. Pathological features such as nuclear grade and tumour necrosis were important factors in this prognostic model.

Leibovich et al.[29] conducted a study of 727 patients with metastatic clear cell RCC and formulated a scoring algorithm based on number and site of metastases, time from nephrectomy, tumour thrombus, nuclear grade and tumour necrosis. The most striking difference of this algorithm from others developed is the inclusion of features specifically related to metastatic disease, site and number. Patients with the lowest scores had 1-year cancer-specific survival rates of 85%, vs 25% for those with the highest scores. ECOG performance status was not significantly associated with death from cancer in this study, but almost all patients within the study group had an ECOG performance status of 0.

FIRST-LINE SYSTEMIC TREATMENT FOR METASTATIC RCC

Motzer et al.[3] studied 670 patients with advanced RCC treated with immunotherapy or chemotherapy. Under multivariate analysis the five pretreatment features most significantly associated with shorter survival were low Karnofsky performance status, high lactate dehydrogenase level, low haemoglobin level, high corrected serum calcium level and absence of nephrectomy. Using these factors they stratified patients into three groups (favourable-, intermediate- and poor-risk groups). The median survival time in patients with no risk factors was 20 months, whereas patients at poor risk with three or more risk factors had a median survival of only 4 months. Motzer et al.[30] also applied this model to compare survival between cytokine therapy and chemotherapy for the three risk groups, to determine whether cytokine therapy benefits patients by risk group in metastatic disease. They found that patients receiving cytokine therapy did indeed have better survival than with chemotherapy, but the greatest mean difference in survival was in the group with few risk factors, while those with poor-risk features had poor survival (<6 months), irrespective of the therapy received.

A similar multivariate analysis of prognostic criteria was applied to patients treated with interferon-α as first-line systemic therapy, and the prognostic model remained the same, except that time from diagnosis to treatment with interferon-α of <1 year was substituted for absence of previous nephrectomy [30].

Zisman et al. developed the UCLA Integrated Staging System (UISS) for both non-metastatic and metastatic disease, which combined the 1997 TNM stage, ECOG performance status and Fuhrman grade [32]. This algorithm is particularly useful, as it is simple to apply and relies on variables that would be readily available at any modern medical practice. In an international multicentre study this algorithm was found to be an accurate predictor of survival in those patients with localized disease, but was found to be less accurate for metastatic RCC [33]. In two of eight centres, the system did not discriminate well for metastatic disease, perhaps reflecting the limitations of staging, interobserver variability in grading and assigning ECOG performance status in patients with extensive disease.

SECOND-LINE SYSTEMIC TREATMENT FOR METASTATIC RCC

Interferon-α and interleukin-2 are standard therapies for metastatic RCC but the response rates are low. It is therefore important to identify prognostic factors that can be used to predict outcome in the second-line setting. Bou Merhi et al.[34] studied 85 patients and in a multivariate analysis identified low haemoglobin (below the laboratories lower limit), raised alkaline phosphatase (>120 U/L) and raised corrected calcium (>10 mg/dL) as independent predictors of survival. In a similar study of 251 patients, Motzer et al.[35] also identified low haemoglobin and high serum calcium levels as independent risk factors for survival, in addition to low Karnofsky performance status (<80%), but alkaline phosphatase was not found to be an independent risk factor. Patients were then stratified into three groups according to their number of risk factors. There was a 17% 3-year survival rate for patients with no risk factors, 7% with one risk factor and 2% with two or more risk factors.

THE FUTURE

THE INTERNATIONAL KIDNEY CANCER WORKING GROUP

Although there are many reports analysing prognostic factors in metastatic RCC, interpreting the data as a whole can be difficult. The factors evaluated, coding of factors and populations treated differ among studies. This is particularly relevant when considering the changing diagnostic capabilities and management in kidney cancer over time, and how that might change patient prognosis. Certainly, survival has improved for patients treated in recent years (in the 1990s) compared with those treated earlier [35]. How much of this improvement is due to the introduction of cytokine therapy remains a moot point.

The need for a comprehensive, well characterized and well validated set of prognostic factors in metastatic RCC is clear. The International Kidney Cancer Working Group was convened in 2002 by the Kidney Cancer Association, and is establishing a database of >4000 patients in first-line trials for metastatic disease. The goal is to develop a single validated prognostic model. In addition, this database might be used as an investigational resource to study the natural history of metastatic RCC, and as an aid in the design, analysis and interpretation of clinical studies.

MOLECULAR MARKERS

While current staging systems are based on pathological and clinical criteria, the incorporation of biomarkers could radically change our approach to diagnosis, treatment selection and prognosis in RCC. The development of techniques that have enabled measurement of the expression of thousands of genes in a single experiment has allowed the rapid analysis of many molecular markers which might be correlated with clinical data.

Von Hippel-Lindau (VHL) gene mutation or methylation occurs in ≈ 80% of sporadic clear cell RCCs [36]. VHL protein regulates hypoxia-inducible factor (HIF)-1 and HIF-2 by binding them and precipitating their rapid degradation. VHL mutations are scattered throughout this very large gene and have different effects on the activity of VHL protein and the regulation of downstream genes.

Of the genes under the control of HIF-1, much attention has been paid to carbonic anhydrase IX (CAIX), which is induced constitutively in certain tumour types but is absent in most normal tissues, except for epithelial cells of the gastric mucosa [37]. Bui et al.[38] reported the expression of CAIX in 94% of 321 tumour samples. Low CAIX staining was an independent predictor of poor survival in patients with metastatic RCC. Survival-tree analysis identified 85% CAIX staining as the most accurate threshold for predicting survival. Furthermore, complete responses to interleukin-2 therapy and patient survival was correlated with CAIX expression in two recent studies [38,39].

Kim et al.[40] used tissue microarrays to develop a prognostic profiling system based on eight tumour markers that were previously reported as being linked to the development of malignancies. Immunohistochemical analysis of Ki-67, p53, gelsolin, CAIX, CAXII, PTEN, EpCAM and vimentin was performed on a custom tissue microarray using clear cell RCC from 318 patients, representing all stages of localized and metastatic disease. In a multivariate analysis CAIX, vimentin and p53 were statistically significant independent predictors of survival.

CONCLUSION

The existing staging and prognostic models have significantly improved the prognostic accuracy in the management of patients with RCC. However, there is a need for a universally accepted prognostic model to aid clinical trial design, analysis and interpretation, and treatment selection. Future iterations of these prognostic models will very probably include molecular characteristics. These new models will need to be sufficiently flexible to take into account widening treatment options and our improved understanding of the molecular pathology of renal carcinoma.

CONFLICT OF INTEREST

Tim Eisen is a consultant to, and study investigator, for Bayer/Pfizer and Wyeth.

Ancillary