|
Activity Number:
|
328
|
|
Type:
|
Topic Contributed
|
|
Date/Time:
|
Wednesday, August 11, 2004 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
Section on Health Policy Statistics
|
| Abstract - #300864 |
|
Title:
|
Estimating Adenoma Risk for Microsimulation Modeling
|
|
Author(s):
|
Onchee Yu*+ and Carolyn M. Rutter and Diana Miglioretti and Margaret Mandelson
|
|
Companies:
|
Group Health Cooperative Center for Health Studies and Group Health Cooperative Center for Health Studies and Group Health Cooperative Center for Health Studies and Group Health Cooperative Center for Health Studies
|
|
Address:
|
1730 Minor Ave., Seattle, WA, 98101-1466,
|
|
Keywords:
|
meta-analysis ; nonhomogenous Poisson ; extra-Poisson variability ; MCMC
|
|
Abstract:
|
The risk of developing an adenoma is a key input for microsimulation models of colorectal cancer. Most colorectal cancers arise from adenomas and discovery of adenomas influences future screening. We estimate adenoma risk by combining information from autopsy studies using Bayesian meta-analysis. We use a multinomial distribution for counts with bin probabilities based on a nonhomogeneous Poisson model for individual risk. This allows us to combine information across studies that use different groupings to summarize adenoma counts. Risk is modeled as a function of age and sex and included pseudo-individual-level random effects to incorporate extra-Poisson variability. Models were estimated using WinBugs software. We assessed model fit by comparing posterior predicted values to observed data and validated our model by comparing results from colonoscopy studies to model-based estimates. We estimate that the overall probability of at least one adenoma increases from 0.24 at age 50 to 0.30 at age 60 and 0.36 at age 70. Men were more likely have an adenoma than women (e.g., prevalence at age 60 of 0.36 vs 0.25).
|
