How can we improve prognostic models in renal cell carcinoma?

Abstract

The therapeutic improvements in renal cell carcinoma brought about by the transition from the ‘cytokine era’ to the ‘targeted agents era’, have not affected the peculiar prognostic heterogeneity of the disease, nor have they diminished the importance of risk group classification based on easily assessable and commonly available laboratory and clinical variables. In the landmark study conducted by Motzer et al. before biological agents were available, the median survival of patients in the good prognosis group was 20 months, while the patients in the poor-risk group had a median survival time of only 4 months. With the introduction of anti-VEGF agents, overall survival has approximately doubled in all risk classes. In a population-based analysis of 670 patients treated with anti-VEGF agents, either in the first-line setting or in the second-line setting after cytokines, stratification according to the Database Consortium model showed that patients in the favorable risk group had a median overall survival of 43.2 months, while patients in the poor-risk group had a median overall survival of 7.8 months.

Keywords:

• biological therapy
• concordance index
• kidney cancer
• prognostic models

The therapeutic improvements brought in renal cell carcinoma by the transition from the ‘cytokine era’ to the ‘targeted agents era’ [1] have not affected the peculiar prognostic heterogeneity of the disease, nor have they diminished the importance of risk group classification based on easily assessable and commonly available laboratory and clinical variables. In the landmark study conducted by Motzer et al. [2]. before biological agents were available, the median survival of patients in the good prognosis group was 20 months, while the patients in the poor-risk group had a median survival time of only 4 months. With the introduction of anti-VEGF agents, overall survival has approximately doubled in all risk classes. In a population-based analysis of 670 patients treated with anti-VEGF agents either in the first-line setting or in the second-line setting after cytokines, stratification according to the Database Consortium model showed that patients in the favorable risk group had a median overall survival of 43.2 months, while patients in the poor-risk group had a median overall survival of 7.8 months [3]. Of note, the Database Consortium model has shown similar discriminatory ability compared to other four existing models (the Memorial Sloan-Kettering Cancer Center, the Cleveland Clinic Foundation, the International Kidney Cancer Working Group models and the updated French model adapted to the Avastin and Roferon in renal cell carcinoma trial), with their respective concordance indexes falling in the range of 64 – 68% [3]. These models have been developed in patients naïve to and treated with biological agents, and have included 13 variables overall [3]. A simplified model, based on three variables only, that is hemoglobin levels, performance status and calcemia, has proven to be useful for risk assessment in patients treated with systemic therapy in the second-line setting [4]. In patients receiving everolimus after failure of at least one anti-VEGF agent, Motzer et al. showed that the good, intermediate and poor prognosis groups established using these three variables presented a 1-year survival rate of 70, 56, and 26%, respectively [5]. In this same therapeutic setting, a similar separation in the survival curves was achieved without the use of laboratory findings in the retrospective study by Wong et al. published on this issue of EOP [6]. The prognostic model by Wong et al includes five variables with a similar impact on prognosis, which were clear cell histology (hazard ratio [HR] = 2.9), Karnofsky performance status score (< 80%; HR = 2.9), duration of metastatic renal cell carcinoma (< 1 year; HR = 2.7), progression on first-line tyrosine kinase inhibitor (HR = 2.2), and liver metastasis (HR = 1.9). The three risk classes obtained were respectively associated with 1-year overall survival rates of 84% for patients with 0 – 2 risk factors, 63% for patients with 3 risk factors, and 22% for patients with 4 – 5 risk factors. The use of a prognostic model based on risk groups has intrinsic limitations, as each of the variables employed are categorized and are assigned an equal prognostic value. A prognostic model based on a nomogram can instead incorporate continuous variables and account for different hazard ratios of the variables considered, which can result in a better discriminatory ability. Furthermore, the biological mechanism underlying the prognostic value of the several clinical, radiologic and laboratory variables identified have not been elucidated. Whether some conditions may depend on the timing of the diagnosis and may develop as the burden of the disease increases (lead-time bias), or may be truly indicative of a more biologically aggressive disease is difficult to establish on a clinical basis only. In this regard, valuable information could be provided by assessment of serum cytokines and angiogenic factors, as shown by the work by Tran et al. [7]. In a cohort of 344 patients enrolled in a Phase III clinical study of pazopanib versus placebo, high versus low levels of IL-6, IL-8, hepatocyte growth factor (HGF), tissue inhibitor of metalloproteinases (TIMP)-1 and VEGF were associated to a worse prognosis, both in the pazopanib arm and in the placebo arm. Unlike currently available risk models based on clinical and commonly available laboratory factors, serum cytokines and angiogenic factors may also have important therapeutic implications. In this work, patients with high levels of IL-6 appeared to derive greater benefit from pazopanib (HR for progression, 0.31) with respect to patients with low levels (HR for progression, 0.51) (p for interaction < 0.05). Furthermore, treatment with pazopanib translated into a survival advantage in patients with high levels, but not in patients with low levels of IL-6, IL-8, HGF, TIMP-1, and VEGF, as if these patients had been less affected by the treatment. Such a finding has not been reported by using the available prognostic models discussed before, and may allow identifying patients with a favorable prognosis who could safely defer treatment, thus avoiding therapy side effects, and gaining possible positive effects on quality of life. As IL-6 is known to be directly secreted by renal cell carcinoma cells and cause an autocrine stimulatory [8] and a proinflammatory systemic effect [9,10], this cytokine may at least partially explain and mediate the negative prognostic value of findings such as anemia, neutrophilia and thrombocytosis. Similarly, hypercalcemia might be related to high VEGF levels [11]. These cytokines and angiogenic factors could be combined with the growing number of prognostic/predictive clinical and biochemical variables, including the neutrophil/lymphocyte ratio and the use of angiotensin converting enzyme inhibitors [12,13], in order to develop a unified model. The accuracy of such a model should be explored in different datasets of patients, especially in those enrolled in comparative Phase III trials on targeted agents [14,15]. Cytokines and angiogenic factors may prove to be a powerful tool to detect truly aggressive disease and provide insights into tumor biology at an individual patient level, with potential prognostic and therapeutic implications.

Declaration of interest

The authors have no relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript. This includes employment, consultancies, honoraria, stock ownership or options, expert testimony, grants or patents received or pending, or royalties.