The Surveillance, Epidemiology, and End Results Cancer Survival Calculator SEER*CSC: Validation in a Managed Care Setting

Statistical reports from the Surveillance, Epidemiology, and End Results (SEER) Program are frequently consulted for information on patient survival (1). Although informative, these reports generally provide information by stage, age group, and/or calendar year, and the statistic provided is relative survival, which does not account for competing causes of death. Relative survival (2) is considered a “net” survival measure that is a policy-relevant cancer progress measure recorded over time and reflects only the chance of dying of cancer and is “uncontaminated” by the changing risk of death from other causes. It does not, however, have a direct bearing on an individual patient’s prognosis. The SEER Cancer Survival Calculator (SEER*CSC), previously described and identified as the Cancer Survival Query System (3), is a set of web-based tools or nomograms that are being developed by the Surveillance Research Program of the National Cancer Institute with the focus of providing information on prognosis that is made as specific as possible to an individual patient within the limitations of population-based data. These estimates differ from those provided by Howlader et al. (4) in that the latter compare crude (survival in the presence of competing risks) and net survival for broad patient groupings by age, stage, and comorbidity, in contrast with SEER*CSC, which provides crude estimates that are individualized to the extent possible. The SEER*CSC has been described by Feuer et al. (3) including the methodology used to create the nomograms based on SEER data, Medicare claims data linked to SEER data (SMC), and Medicare claims data for noncancer patients in SEER areas. Currently, the SEER*CSC includes nomograms for colorectal and prostate cancer, with nomograms for breast and head and neck cancers under development. Although there are an increasing number of cancer nomograms developed for predicting a variety of clinical end-points including death, for example, those available on the Memorial Sloan Kettering Cancer Web site (5) and Adjuvant Online (6), the SEER*CSC has some unique features. First, the SEER*CSC provides equal focus on the chance of dying of cancer and of other causes, and the factors that contribute to each. This is important so that the physician and patient can consider the full spectrum of medical interventions, rather than isolating the cancer and its treatment. As cancers are increasingly diagnosed in early stages, many cancer patients might be surprised to learn that they have a higher chance of dying from conditions other than their cancer. Second, the SEER*CSC uses a patient’s comorbid conditions to compute their “health-adjusted age” (a measure of the patient’s life expectancy independent of their cancer), which is incorporated into the calculations. Finally, because the SEER Program maintains a large diverse population-based database, the estimates presented in this study are more reliable than those that come from any clinical cohort or trial. Estimates are also available for groups that are not usually included in trials (eg, the elderly or those with significant comorbidities). The SEER*CSC, as an adjunct to standard SEER survival statistics, was designed to provide useful information to both physicians and patients by providing crude estimates (ie, estimates in the presence of competing causes of death) of the risks of cancer-specific and other causes of death. This tool aims to give the health professional and patient the best available population-based estimates for short- and long-term survival, as one of many pieces of information to weigh in helping patients and their families make difficult treatment and personal decisions. Examples of the SEER*CSC web interface are provided in Figure 1, which shows the data input page for an African American prostate cancer patient, Figure 2, which shows what comorbidities are included in the system to adjust other cause survival, and Figure 3, which shows one of the options for presenting survival predictions from an inquiry. Figure 1. Surveillance, Epidemiology, and End Results Cancer Survival Calculator SEER*CSC web interface; cancer and demographic data input. Figure 2. Surveillance, Epidemiology, and End Results Cancer Survival Calculator SEER*CSC web interface; comorbidity data input. Figure 3. Surveillance, Epidemiology, and End Results Cancer Survival Calculator SEER*CSC web interface; survival statistics output. The two nomograms included in the SEER*CSC were internally validated using the SEER data (7). The objective here is to externally validate the prostate and colorectal nomograms using patients from Kaiser Permanente Colorado (KPCO) based on methods that have been described by DeLong et al. (8) and Kattan et al. (9). This expands the validation of the nomograms to a broader community population from an integrated health-care delivery system. Another paper in this monograph will use the same cohorts of patients to characterize patterns of care both before and after their diagnosis of cancer, as a function of their predicted chances of death from cancer and other causes (10).