Purpose: Prostate cancer (PCa) is the most common solid tumor in American men. There are many risk factors and different guidelines concerning PCa screening and biopsy decision policies. We investigate the optimal use of PSA history to determine whether to refer a patient for biopsy.
Method: We report on a Partially Observable Markov Decision Process (POMDP) model. Underlying this 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. Before PCa is confirmed, a belief state is used to define the patient’s state relative to the presence of prostate cancer. The belief state defines the probability the patient is in the PCa not detected state. The belief state is updated based on the results of annual PSA tests using Bayesian updating. We compute the biopsy referral policy that maximizes expected quality adjusted life years (QALYs). Parameters such as PSA test results, conditional probabilities for PCa, are estimated using a dataset of 11,872 men from Olmstead County, MN from 1988 to 2006. Other parameters such as utility values for health states, biospy, and treatment outcomes are drawn from secondary sources in the medical literature.
Result: We evaluate a base case that assumes a patient at age 40 with no PCa. Solving the POMDP model yields an age and belief state dependent threshold for biospy referral. Sensitivity analysis illustrates that the optimal referral policy can improve the expected QALYs by as much as 0.30 across the general population, and as much as 1.34 for patients eventually diagnosed with PCa compared to no screening.
Conclusion: Depending on the utility decrements associated with prostate biopsy and treatment, optimal use of PSA testing as part of an annual screening program can have a significant benefit to the patient. Our POMDP model offers a means for obtaining the optimal screening guideline and quantifying the effect of following the optimal guideline.
Candidate for the Lee B. Lusted Student Prize Competition