Yan Yan, M.D., M.H.S., Ph.D.
Jay Piccirillo, M.D., F.A.C.S.
Project Overview:
Prostate cancer is the most common cancer among men in the U.S. and the second leading cause of cancer death. As in other cancers, the majority of men diagnosed with prostate cancer are elderly. While some prostate cancer patients die from this disease, most of them die from other causes due to both a slow progression of this disease and a high prevalence of other comorbid conditions.
Death from prostate cancer and from other causes is usually associated with different risk factors and has different implications for health care needs. The purposes of current project are: (1) to identify the risk factors related to different causes of death, and (2) to predict individualized probabilities of death from prostate cancer or from other causes using risk factor information. Our study cohort consists of all prostate cancer patients diagnosed between 1991-1999 in the SEER program and enrolled in the Medicare program. The data for this project come from newly merged SEER-Medicare datasets, which contain information on patient demographics, tumor grade and stage, primary treatment options, comorbidity, and vital status, etc. We will use the part of our two sandwiching method as a statistical tool. Our two sandwiching approach has several advantages. First, it models death from other causes as a competing event not as censored information, thus the probability estimates are unbiased. Second, it avoids the Markovian assumption for successive stage modeling. Third, it is flexible to accommodate both parametric and non-parametric modeling. In addition, it provides a framework for estimating many important clinical and epidemiological quantities.
With the statistical tool we have previously developed, we will estimate several types of individualized probabilities: the absolute cause-specific death probability, Pj (t | Z); the conditional cause-specific death probability, Pj (m | t, Z); the relative risks of death from prostate cancer versus from other causes, RR (t | Z) and RR (m | t, Z). These probabilities have a wide range of applications. Clinicians can use the absolute cause-specific death probability to consult individual patients about optimal treatment options before treatment, and can use the conditional cause-specific death probability with the relative risk estimates to formulate more effective prevention and treatment strategies for reducing death risks from prostate cancer and from other causes. Health care policy makers can use the absolute cause-specific death probability to make the long-term plan about future health care needs, and can use the conditional cause-specific death probability with the relative risk estimates to adjust their long-term plan for possible changes. In addition, these probabilities enable prostate cancer patients to better understand their prognoses, therefore, to more actively participate in the joint-decision making.
Final Report:
Prostate cancer is the most common cancer among men in the US and the second leading cause of cancer death. As in other cancers, the majority of men diagnosed with prostate cancer are elderly. While some prostate cancer patients die from this disease, most of them die from other causes because prostate tumors usually grow slowly and prostate cancer patients are usually old with many other health conditions.
