Jaywant Philip Parmar MD, Presenter: Nothing to Disclose

Talissa A. Altes MD, Abstract Co-Author: Nothing to Disclose

George Stukenborg, Abstract Co-Author: Nothing to Disclose

Bruce Jay Hillman MD, Abstract Co-Author: Nothing to Disclose

Emerging non-lumenographic anatomic, functional and molecular vascular imaging may better identify patients at risk of stroke who will benefit from CEA. A Bayesian decision model is presented elaborating exactly how an alternative imaging technology should perform to be better than current lumenographic methods.

A computerized decision tree model was constructed. Expected utility of 4 alternatives are simultaneously calculated: 1. the current NASCET strategy, 2. a possible future image based strategy, 3. a random chance decision, and, 4. a perfect information decision. Fixed parameters included utilities of the 8 possible outcomes and population probabilities of strokes avoidable and unavoidable by CEA. According to NASCET observational data, a baseline 2-year stroke rate of 26% with 9% unavoidable was considered. Variable parameters included the imaging test dependent sensitivities, specificities, positive predictive values, negative predictive values, and positive imaging sign dependent probabilities of CEA avoidable stroke.

For example, under the baseline population conditions, the future technique becomes the best choice if its positive observation has a 90% probability of identifying an avoidable stroke with very good sensitivity and specificity (80%) as compared to the randomized trial demonstrated 65% stroke rate reduction achievable through the positive observation of 70-99% stenosis when measured with near perfect sensitivity and specificity (98%). As interventions become less risky (false positive utility discount < 32%), the random chance decision dominates. Sensitivity analysis of the baseline stroke rate within in the range of clinical uncertainty (20-30% 2 year CEA avoidable stroke rate) was found to be particularly complex, possibly supporting the random chance, NASCET and future strategies within reasonable ranges of the remaining estimable parameters.

The model as developed can be used to assess novel imaging diagnostics using preliminary research results. Important effects of the test population treatable disease rate and CEA treatment risk on the decision to use the new imaging method are demonstrated and quantified.

Parmar, J, Altes, T, Stukenborg, G, Hillman, B, A Value of Information Decision Analysis Comparing the Utility of Current and Future Imaging Diagnostics for Carotid Endarterectomy Patient Selection.  Radiological Society of North America 2005 Scientific Assembly and Annual Meeting, November 27 - December 2, 2005 ,Chicago IL. http://archive.rsna.org/2005/4419402.html