| To Register SMDM Homepage | |
|
|
|
||||
Purpose: Clinicians recognize that there seem to be variable levels of aggressiveness in many cancers. We performed a decision analysis of the effectiveness of prostate cancer screening with the assumption that there were two levels of aggressiveness: one high and one low.
Methods: We modified the prostate cancer screening model of Cantor SB, et al. (J Fam Pract 1995; 41:33-41). We considered annual screening beginning at age 50, using digital rectal exam (DRE) and Prostate Specific Antigen (PSA) tests.
The sensitivity and specificity of the screening tests were held the same, but we allowed the transition rates from the initial and subsequent stages of cancer to vary in two subgroups: one slowly developing and one rapidly developing. The proportion of rapidly developing cancers was set at 25%, 50%, and 75% and the transition probabilities were recomputed so that (1) the mean time a slow progressing disease spent in a state was twice the mean time of the fast progressing disease, and (2) the average transition probabilities remained the same. We computed effectiveness using the same utilities for health states as in the original paper.
Results: The original analysis, based on a single level of disease progression, indicated that screening resulted in decreased effectiveness by 0.67 quality-adjusted life years (QALYs). In the new analysis with two levels of disease aggressiveness, we found that if 25% of the cancers were fast progressing, then no screening is still preferred, but by only 0.24 QALYs. With 50% and 75% of the cancers fast progressing, screening has slightly higher effectiveness by 0.12 and 0.42 QALYs, respectively.
Conclusions: When more detailed modeling was incorporated in the prostate cancer decision analysis, the net benefits of screening were enhanced. Our analysis showed that relaxing the assumption of a single set of probabilities for transitions between health states can change the optimal recommendations. Often a decision analysis relies on many assumptions about the disease or other conditions which may or may not be true. Our analysis highlights one assumption that can be critical to the outcome of the analysis.
See more of Poster Session - CEA: Methods and Applications; Health Services Research
See more of The 26th Annual Meeting of the Society for Medical Decision Making (October 17-20, 2004)