Friday February 24, 2012, 12:00-1:00pm, LD 030
A predictive model of serial, patient-specific colonoscopy results [flyer]
Eric A. Sherer, PhD (CTSI
Disease and Therapeutic Response Modelling Program at IUPUI, and
Roudebush Veterans Affairs Medical Center in Indianapolis)
Abstract:
The use of regular colonoscopy has increased considerably since
Medicare began covering screening colonoscopy in 2001 for average-risk
patients and the objective of our research is improve the efficiency
and effectiveness of post-colonoscopy care by tailoring recommendations
for a specific patient. In particular, the model was devel-oped to
capture the clinical result that a patient’s risk for colorectal
neoplasia varies most strongly with his base-line colonoscopy findings.
The model framework relies on a neoplasia natural history model to
describe progres-sion through four neoplasia development states with
patient age. We use multiple possible combinations of nat-ural history
model parameter sets, whose a priori likelihoods are a function of
patient sex, to model the overall neoplasia formation and, after a
colonoscopy, the parameter set combination likelihoods adjust in a
Bayesian manner based on the results and conditions of the colonoscopy
and their model predictions can. The adjustment of model predictions
operationalizes, and is necessary to capture, the clinical results that
multiple or advanced neoplasia at baseline colonoscopy is an
independent predictor of multiple or advanced neoplasia at follow-up
colonoscopy, and vice versa for negative colonoscopies, and the
adjustment of parameter set combination likeli-hoods accounts for the
possibility that patients may have different neoplasia development
rates. To support model identification, observational longitudinal
colonoscopy results, procedure details, and patient characteris-tics
were collected for 4,084 patients. We found that at least two parameter
sets specific to each sex with model adjustments was required to
capture the longitunidal colonoscopy data and inclusion of multiple
possible param-eter set combinations, which account for random
variations within the population, was necessary to accurately predict
the second-time colonoscopy findings for patients with a history of
advanced adenomas. Application of this model to predict CRC risks for
patients adhering to guideline recommended follow-up colonoscopy
intervals found that there are significant differences in risk with
patient age, gender, and preparation quality and demon-strates the need
for a more rigorous investigation into these follow-up recommendations.