Search
Prognostic Factors for Overall Survival
Daniel Heng, from the Tom Baker Cancer Center in Calgary, Alberta, presented a multi-center, retrospective study in a poster session, delineating the prognostic factors for overall survival in patients with metastatic renal cell carcinoma who are treated with VEGF-targeted agents, i.e. Sutent, Nexavar, or Avastin. 645 patients at seven sites were included in the study, which has now (August 2009) been expanded to 11 international sites. This study was undertaken to update the widely used risk factors that are currently used by kidney cancer oncologists, the MSKCC risk factors, which were developed by Dr. Robert Motzer at Memorial Sloan-Kettering, and the CCF model developed at the Cleveland Clinic. Both of these models were developed in the era of immunotherapies (these two models are summarized at the end of this article). Dr. Heng wanted to update these factors in a large study where targeted agents were used.
The objective of having a prognostic model is for oncologists to better plan treatment of metastatic patients (some therapies are shown to work well on high risk patients, for example) and the ancillary benefit of designing trials segregating patients by low, intermediate, and poor risk for survival and gauging the effect of therapies on each specific population. Having a better idea of the patient's prognosis also assists the oncologist in patient counseling.
Patients in Heng's study were not segregated by histology, and patients who had received prior treatment by immunotherapy (one-third of the patients) were included. For the other two-thirds of the patients, the targeted therapy was their first treatment. The time from diagnosis to targeted therapy was 1.4 years with a median age of 60 years old. The breakdown in therapies was:
| Therapy | Percentage |
| Sutent | 61.4% |
| Nexavar | 31.0% |
| Avastin | 7.6% |
Univariate Analysis
On the way to developing their prognostic factors through multivariate analysis, the investigators performed univariate analysis on a number of factors, some of which provided significant results and others not. Note that many of the univariate variables, even if statistically significant, were not independent predictors of overall survival so were not included in the final model.
Univariate Analysis
| Variable | Median Overall Survival Months |
| Nephrectomy vs None | 27.0 vs 10.9 |
| Anemia vs Not | 16.9 vs 39.4 |
| Hypercalcemia vs Not | 8.8 vs 26.8 |
| Neutrophilia vs Not | 5.9 vs 27.0 |
| Thrombocytosis vs Not | 10.4 vs 27.4 |
| KPS < 80 vs > 80 | 9.4 vs 31.6 |
| Time from diagnosis to treatment < 1 yr vs > 1 yr | 15.8 vs 30.9 |
| Elevated LDH vs Not | 15.8 vs 26.3 |
| Non-clear cell histology vs clear cell histology | 16.8 vs 22.7 |
| Sarcomatoid features vs None | 8.6 vs 22.7 |
| Neutrophilia is an over-abundance of neutrophils, a type of white blood cells, used to fight infections. Thrombocytosis is an abnormally high number of platelets in the blood, which could indicate inflammation. KPS is Karnofksky Performance Status ranging from 0-100%, with KPS = 80 being "Normal activity with effort; some signs or symptoms of disease" LDH is lactate dehydrogenase, an enzyme, which is found to be elevated in most cancers, heart failure, anemia, etc. | |
Multivariate Analysis
From the multivariate analysis, the investigators extracted their prognostic factors. The results of that analysis are as follows.
Significant Variables - Multivariate Analysis
| Clinical Variables | Hazard Ratio |
| Karnofsky Performance Status < 80 | 2.51 |
| Diagnosis to treatment < 1yr | 1.42 |
| Lab Variables | |
| Neutrophil count > ULN | 2.42 |
| Calcium > ULN | 1.81 |
| Hemoglobin < LLN | 1.72 |
| Platelet count > ULN | 1.49 |
| LLN = lower limit of normal; ULN = upper limit of normal | |
For explanations of Karnofsky Performance Status (KPS)[1] and Hazard Ratio[2], see footnotes below.
Given that a low hemoglobin count is defined as anemia and a high platelet count is thrombocytosis, the six factors constituting a poor prognosis are:
Dr. Heng tested his model for first-line pts only, second-line pts only, Sutent pts only, and Nexavar pts only, and it performed weel in all of these sub-population. For statisticians, the c-index = .73, which validates the model.
Heng now developed his forecast:
The investigators established risk groups yielding the following data for the study population.
| Risk Group | Median Overall Survival |
| Favorable risk | 37 months |
| Intermediate risk | 28.5 months |
| Poor risk | 9.4 months |
Note that these survival times are longer than those developed by Dr. Motzer in his revised 2004 MSKCC model, which generated median survival of 22.1 months, 11.9 months, and 5.4 months for favorable, intermediate, and poor risk categories. This improvement could be due to the broader use of scanning, which leads to earlier diagnosis, the availability of second-line therapy, and the use of targeted therapy, which first became available in December 2005.
As mentioned, Dr. Heng, who already has the participation of the Cleveland Clinic and Dana Farber in Boston, is expanding his model to include additional institutions and his model will undergo external validation, but one must remember that this is a retrospective study not a prospective one, which wold hold more validity. He will present his findings at ASCO 2010. A new model is clearly needed since the 2002 MSKCC model is based on cytokine therapy, which is no longer the standard of care for metastatic kidney cancer. Dr. Motzer, from Memorial Sloan-Kettering, did not participate in Dr. Heng's study, but he is not standing still. In 2008, he developed an 11-variable nomogram for progression-free survival for Sutent therapy - see below. One difference in methodology between the Heng and Motzer models is that Motzer's sample population comprises only trial patients while Heng's includes community patients, who represent a broader spectrum of the metastatic population. For example, patients with brain metastasis are usually restricted from clinical trial participation. About 8% of Heng's sample have brain mets so poorer risk patients are accounted for. Since there is no official body that certifies kidney cancer prognostic models, we may be seeing a struggle between competing models in the future.
Addendum - Other Prognostic Models
The point of reference for Heng's study were earlier studies by Robert Motzer of Memorial Sloan-Kettering Cancer Center in 2002, 2004, and 2008, and one by a group at the Cleveland Clinic in 2005. These models, excepting the 2008 one, were based on the fact that patients were to receive cytokine therapy, i.e. Interferon-alpha or Interleukin-2. Motzer's 5-factor model, http://jco.ascopubs.org/cgi/reprint/20/1/289, is the most commonly one used by oncologists conducting clinical trials when they want to stratify the patient population by risk category. It is a modification of his earlier 1999 model substituting "diagnosis to treatment" for "prior nephrectomy". The model develops MSKCC risk categories with the risk factors as follows.
MSKCC Risk Factors (2002)
| Variable | Hazard Ratio |
| Lactate dehydrogenase > 1.5 x ULN | 3.23 |
| Corrected calcium > 10 mg/dL | 1.93 |
| Hemoglobin < LLN | 1.53 |
| Karnofsky Performance Status < 80 | 1.52 |
| Diagnosis to treatment* < 1yr | 1.48 |
| * Treatment is with Interferon-alpha LLN = lower limit of normal, ULN = upper limit of normal | |
The MSKCC risk categories derived from this model are: favorable (0 risks), intermediate (1-2 risks), and poor (3-5 risks). Note that in 2004, http://jco.ascopubs.org/cgi/reprint/22/3/454, Dr. Motzer refined his model dropping two variables that were lacking statistical significance.
MSKCC Risk Factors (2004)
| Variable | Hazard Ratio |
| Karnofsky Performance Status < 80 | 3.5 |
| Corrected calcium > 10 mg/dL | 3.4 |
| Hemoglobin < LLN | 1.7 |
The therapy in this study was either Interferon-alpha or Interleukin-2. The MSKCC risk categories derived from these factors are: favorable (0 risks), intermediate (1 risk), and poor (2-3 risks). Despite being updated in 2004, most oncologists still use the 2002 model.
In October 2008, Dr. Motzer published the results of a Phase III trial in first-line patients randomized to either Sutent or Interferon-alpha (Cancer. 2008 Oct 1;113(7):1552-8). The investigators developed a nomogram (a graphic representation yielding probabilities) that predicts the probability of progression-free survival for patients receiving Sutent therapy. The nomogram contains 11 variables: corrected calcium level, the number of metastatic sites, hemoglobin level, prior nephrectomy, the presence of lung and liver metastases, thrombocytosis, performance status, time from diagnosis to treatment, and serum levels of alkaline phosphatase and lactate dehydrogenase.
The second reference for Heng is the Cleveland Clinic study, http://jco.ascopubs.org/cgi/reprint/23/4/832. Their study verified four of the five MSKCC factors with the Karnofsky Performance Status being dropped since their study population did not contain sufficient variance in this indicator to include it in the model.
Back to ASCO 2009 Table of Contents
[1] Karnofsky Performance Status is a score between 0 (death) and 100 (normal) that indicates the condition of the patient. A score of 80 indicates ‘normal activity with effort with some signs or symptoms of disease'. A 70, on the other hand, indicates that the patient ‘can care for himself buy is not able to carry on normal daily activity including work'.
[2] Hazard Ratio is the relative risk of dying first in one group versus another. So where HR = 2.51 where KPS < 80, the relative risk of dying first is 2½ greater than for patients with a KPS ≥ 80. Expressing it as a probability, where p = HR/(1 + HR), the probability of dying first is 70%.