Therapeutic Landscape in Kidney Cancer

Sandy Srinivas, from Stanford University, gave a comprehensive talk on the Therapeutic Landscape in Kidney Cancer. Dr. Srinivas started her talk by describing the history of one of her patients who, still alive today, had a nephrectomy for kidney cancer in April 2001. He was treated for his first metastasis in October 2004 and went on to receive treatment with six different targeted therapies, one of them repeated, switching from one to another due to progression (liver and gastric metastases) or adverse events. Her point was that although there are several options available today to treat mRCC and survival has increased, selection of one therapy versus another is not standardized, and the therapies have toxicities. For example, Dr. Srinivas astutely pointed out that if one were to consult the NCCN (National Comprehensive Cancer Network) Guidelines (  for clear cell mRCC patients, it offers any one of six drugs for first-line therapy, pushing only everolimus into the second-line category.  Dr. Srinivas stated that their choices were reasonable since there is currently little direct trial result evidence that compares one therapy against another either for efficacy or tolerability. However, some face-to-face trials are now being conducted, e.g. pazopanib vs sunitinib.

Dr. Srinivas reviewed the trial history of the available therapies. Since similar data was presented in Elisabeth Heath’s talk, see, it won’t be repeated here. One difference, however, is that Dr. Srinivas discussed the viability of IL-2 as an option for a select group, specifically, young, clear cell patients with good organ function and good risk. 

Dr. Srinivas went on to delineate the factors that should be taken into account in order to make a decision among the available therapies: patient characteristics, tumor characteristics, and drug characteristics. For example, Dr. Srinivas feels that a VEGF inhibitor would be appropriate for someone with good performance status, a need for rapid response due to being symptomatic or for pre-surgical reasons, and with no co-morbidities such as refractory diabetes or uncontrollable high cholesterol and triglycerides. On the other hand, one might choose an mTOR inhibitor for patients with poor risk, or who have had prior VEGF therapy, or have co-morbidities of congestive heart failure or uncontrollable hypertension. 


To rationalize the choices, Dr. Srinivas would prefer to rely on biomarker development, which would assist the oncologist to “predict recurrence, choose the appropriate therapy, predict resistance, and avoid toxicity”. Unfortunately, there are not yet any confirmed biomarkers in kidney cancer. This has led to a scramble among RCC researchers to examine both tumor and patient characteristics, such as what genes are mutated. According to Dr. Srinivas, biomarkers could: “help predict recurrence, choose the appropriate therapy for first-line treatment, predict resistance to aid in choosing second-line therapy, and avoid toxicity”. 

Predict Recurrence

This is an important area in that someone with a high risk of recurrence would be wise to take an adjuvant therapy (which are in trials but not yet available), and someone with a low risk could reduce the number of follow-up visits and scans. Current factors being used for prognostic purposes are: tumor size, tumor spread (renal vein, nodes, collecting system), Fuhrman grade, presence of tumor necrosis or sarcomatoid features, and performance status of the individual. Some factors are more accurate predictors than others, e.g. sarcomatoid features, yes, but tumor size less so. And then the question remains as to how to put them together? 

In order to rationalize their treatment recommendations, oncologists are really looking for a test, preferably genetic, that will give them a score with prognostic value. To this end, a company called Genomic Health. Inc. is developing such a test for kidney cancer that is similar to the one they did for breast cancer. The project was reported on at last year’s ASCO conference by Brian Rini of the Cleveland Clinic. It uses the same technology as Oncotype DX, which is based on the genetic profile of breast cancer patients. For kidney cancer, they analyzed over 700 genes from 942 patients and found 16 of them significant for recurrence of RCC. These genes are now being validated. Dr. Srinivas also referred to the work of Kim Rathmell of the University of North Carolina, whose work is slightly further behind that of Oncotype DX. 

[Ed. Note: For breast cancer, Genomic Health, Inc. developed Oncotype DX, by creating three risk categories to predict recurrence of breast cancer within 10 years. Their prospective study yielded risk of recurrence as: low score (7%), intermediate score (14%), high score (31%). The test was validated against age at surgery and tumor size, which basically had no significant predictive value. As a result, estrogen-receptor positive women who took the test and went on to receive tamoxifen as adjuvant therapy were found to have derived, for the low and intermediate risk groups, little or no benefit from the addition of chemotherapy to the regimen, while high risk patients received substantial benefit far outweighing the side effects of the therapy.] 

Dr. Srinivas listed a number of potential genetic biomarkers that might predict tumor recurrence in kidney cancer including the growth factor gene IMP-3 for which high expression predicts a poor 5-year progression-free survival (PFS) (33% vs. 89%); CXCR3, a chemokine, which is a subset of the immune proteins called cytokines, for which a low expression predicts poor PFS (57% vs. 82%); survivin, a gene that encodes a cell-death inhibitor, for which a high expression predicts poor survival (59% vs. 87%); and B7-H1, an immune system inhibitor, for which high expression again predicts for poor PFS (57% vs 86%). [Ed. Note: While researching these agents, we found a reference that showed that high expression of B7-H1 and survivin together predicted low 5-year cancer specific survival (16.2%) whereas high B7-H1 and low survivin yielded 70.0% survival! At this point oncologists are not rushing to test their patients for any of these indicators, probably because they haven’t been validated in large-scale prospective trials, plus there are no approved adjuvant therapies for kidney cancer as yet.] 

Dr. Srinivas went on to list the current trials for adjuvant therapy, which are reproduced below with the editor’s addition of Data Date (the date when data are available to perform analysis of the trial results), Location, and Status. We’ve also referred to, and if there was a discrepancy, we have used the latter’s data with an asterisk. We also added an everolimus adjuvant trial, which commenced in April. Girentuximab is a monoclonal antibody that is administered via IV infusion. Although there are no data available yet, it is known that many of the participants have discontinued therapy due to intolerable toxicities.

Choose the Appropriate Therapy

Like most other researchers, Dr. Srinivas uses the two predictive models in use today that help oncologists classify patients in clinical trials as to their survival risk status, namely, the MSKCC model, developed by Robert Motzer of Sloan-Kettering, and the model developed by Daniel Heng of the Tom Baker Cancer Centre in Calgary, AB. This is an esoteric area of RCC so we’re only covering it to point up some anomalies. Dr. Motzer first developed his model in 1999, defining good, intermediate, and poor risk categories for survival. He developed five variables that independently predicted for poor survival: low performance status (KPS), high serum lactate dehydrogenase (LDH), low hemoglobin (anemia), hypercalcemia, and absence of prior nephrectomy (later changed to interval between diagnosis, Dx, and treatment, Tx, less than 1 year). Dr. Heng’s model for favorable, intermediate, and poor risk for survival has six variables, the four from MSKCC including Dx to Tx < 1 year, and adding neutrophilia (high white blood cell count pointing to inflammation), and thrombocytosis (high platelet count). Dr. Motzer modified his model in 2004, reducing the number of independent variables to three: low KPS, low hemoglobin, and hypercalcemia. 

If you’ve slogged through these model descriptions, you’ve reached the interesting part. MSKCC (1999 and 2004) are not prognostic models, which predict survival for all types of therapy. They are predictive models based on the therapy that the patients took while being studied, namely cytokines (interferon-alpha (INF) and Interleukin-2).  Unfortunately, fewer and fewer patients  try IL-2 and the only ones on INF are those who cannot afford targeted therapies So, Dr. Heng came along and validated his model on patients undergoing targeted therapy. But oncologists who present trial results still use the MSKCC model, the 1999 version. It’s hard to change old habits. The models are not only used for comparisons of patient risk and response to therapy, but sometimes also in therapy selection, e.g. temsirolimus versus a VEGF therapy for poor risk patients. Finally, the good, intermediate, and poor risk categories are defined by the number of risk factors a patient has. For specifics, see the original source material. For MSKCC: JCO Vol17 Issue 8 Aug 1, 1999 and JCO Vol 22 Number 3 Feb 1, 2004; for Heng: JCO Vol 27 Issue 34 Dec 1, 2009.

Except perhaps for temsirolimus, these models don’t really help in the selection of therapies. Dr. Srinivas went on to quote from a number of genetic and clinical studies that predict for response in a single therapy. But there were no large scale studies of multiple therapies using biomarkers to predict comparative benefit of one therapy over another. One interesting point was a reference to a retrospective study by Brian Rini of the Cleveland Clinic that showed, for targeted therapies, an association between increased hypertension during treatment and increased tumor response, longer PFS, and increased overall survival. Controlling the increased blood pressure did not diminish the response. So, Dr. Srinivas remarked, does one escalate the dosage to effectuate a higher grade of hypertension? Let’s see if someone takes up her challenge. 

Dr. Srinivas also wondered who are the 6-8% of patients who have a complete, durable response to Interleukin-2? The SELECT Trial, reported on at last year’s ASCO Conference, was supposed to have verified the long held opinion that a high expression of carbonic anhydrase IX predicts for response to IL-2. Much to the investigators’ dismay, their hypothesis was proved wrong. 

Predict Toxicity

The following table of selected adverse events was presented by Dr. Srinivas. Note the color coding of the pluses: red signifies a class action, i.e., the adverse event (AE) is present only for a class of therapies (VEGF inhibitors or mTOR inhibitors but not both); orange signifies an AE that is a by-product of all targeted therapies.

If axitinib were included in the above table, it would have the following approximate profile[1]:

There don’t seem to be many studies in this area so it is a nascent field for research. Dr. Srinivas only gave one example where gene polymorphisms (minor mutations) could predict susceptibility to specific toxicities such as, in this case, hand-foot syndrome for sunitinib patients. 

Dr. Srinivas ended her talk by saying that “there are enormous challenges in the biomarker discovery field” and drug development for kidney cancer has been “way ahead of biomarker development”. 

This was a comprehensive survey of the therapeutic regimens for kidney cancer with an emphasis on biomarkers. Now that there are seven FDA-approved drug protocols for kidney cancer, oncologists are looking for standards as to which therapy would better fit the patient in terms of efficacy and toxicity. It is clear that no standard now exists but judging from the number of papers delivered at ASCO on biomarkers, we may see something tangible in a couple of years. Nevertheless, we must remember that none of the targeted therapies are curative with built-up resistance requiring a change within a year (median). Kidney cancer therapy needs a new paradigm.

[1] Per discussion with Dr. Srinivas.