Wednesday, October 22, 2008
Columbus A-C (Hyatt Regency Penns Landing)
Purpose: Prostate cancer (PCa) is the most common solid tumor in American men and is screened for using prostate specific antigen (PSA). The imperfect sensitivity and specificity of PSA create uncertainty in the threshold at which to recommend prostate biopsy.
Methods: We report on a Partially Observable Markov Decision Process (POMDP) model. Underlying our model is a discrete time Markov process that represents the progression of patients through health states of no PCa, PCa not detected, PCa detected and treated and death. The first two living states are not directly observable and must be inferred from a PSA test and subsequent biopsy. We use 5 discrete PSA ranges to represent possible PSA states of the patient. The false negative probability of biopsy is estimated from secondary sources. We estimate the conditional probability for each health state of the patient using a population-based dataset of PSA testing results done from 6803 men from Olmstead County, MN with 3 or more PSA tests from 1988 to 2006. Transition probabilities are inferred from SEER data and the Mayo Clinic Radical Prostatectomy Registry. We assume that at each annual decision epoch the decision to biopsy or not is based on the most recent PSA test. The objective in our model is to maximize expected quality adjusted life years. For each year of life the patient realizes quality adjusted life years estimated using utility values for health states, biospy, and treatment outcomes drawn from secondary sources.
Results: Solving the POMDP model yields an age and PSA state dependent threshold for when to recommend biospy. We consider a decision horizon from age 35 to 85. Results illustrate a time varying PSA threshold of 4.0 for ages 35-57, 7.0 for ages 58-64, and 10.0 for ages 65-66. For ages > 66 the optimal policy indicates no biopsy at any PSA state, and implies discontinuation of standard PSA screening at age 66.
Conclusions: Partially Observable Markov Decision Processes offer a means for quantifying the value of information provided by PSA tests and subsequent biopsy. Optimal thresholds for recommending biopsy are highly age dependent, and the optimal policy indicates discontinuing PSA testing at age 66.
Methods: We report on a Partially Observable Markov Decision Process (POMDP) model. Underlying our model is a discrete time Markov process that represents the progression of patients through health states of no PCa, PCa not detected, PCa detected and treated and death. The first two living states are not directly observable and must be inferred from a PSA test and subsequent biopsy. We use 5 discrete PSA ranges to represent possible PSA states of the patient. The false negative probability of biopsy is estimated from secondary sources. We estimate the conditional probability for each health state of the patient using a population-based dataset of PSA testing results done from 6803 men from Olmstead County, MN with 3 or more PSA tests from 1988 to 2006. Transition probabilities are inferred from SEER data and the Mayo Clinic Radical Prostatectomy Registry. We assume that at each annual decision epoch the decision to biopsy or not is based on the most recent PSA test. The objective in our model is to maximize expected quality adjusted life years. For each year of life the patient realizes quality adjusted life years estimated using utility values for health states, biospy, and treatment outcomes drawn from secondary sources.
Results: Solving the POMDP model yields an age and PSA state dependent threshold for when to recommend biospy. We consider a decision horizon from age 35 to 85. Results illustrate a time varying PSA threshold of 4.0 for ages 35-57, 7.0 for ages 58-64, and 10.0 for ages 65-66. For ages > 66 the optimal policy indicates no biopsy at any PSA state, and implies discontinuation of standard PSA screening at age 66.
Conclusions: Partially Observable Markov Decision Processes offer a means for quantifying the value of information provided by PSA tests and subsequent biopsy. Optimal thresholds for recommending biopsy are highly age dependent, and the optimal policy indicates discontinuing PSA testing at age 66.
See more of: Poster Session V
See more of: 30th Annual Meeting of the Society for Medical Decision Making (October 19-22, 2008)