Wednesday, October 22, 2014
Poster Board # PS4-34

Candidate for the Lee B. Lusted Student Prize Competition

Philip Akude, MSc, Mike Paulden, MA., MSc. and Christopher McCabe, PhD, University of Alberta, Edmonton, AB, Canada
Purpose:  Optimising the value of monitoring tests is currently under-researched. Conventionally, tests are repeated at fixed intervals using a fixed test cut-point value. Theoretically, the performance of a diagnostic test changes with repeated use in any given population. We therefore hypothesize that conventional practice may be sub-optimal.

   In this study, a simulation model is built to: investigate the test performance and optimal test cut-point value for a monitoring test; optimize the test cut-point value and retest interval in a dynamic equilibrium to maximize population health.

Method: Two simulated experiments are carried out on a hypothetical population of 10,000 patients who are being monitored for disease progression or remission for a period of 6 months.

   In the first experiment, a dichotomous test is applied on the monitoring population each month for the entire monitoring period. After each test, patients are repeatedly stratified into diagnostic sub-populations of positives or negatives using a cut-point value between 0 and 1, in increments of 0.05. The test performance (including sensitivity and specificity) and optimal cut-point in each sub-population is measured after each repeat test.

   In the second experiment, patients are allowed to undergo only two tests within a monitoring period. The interval between these two tests and the test cut-point value used for each test are simultaneously determined. The population net health benefit (in QALYs) is estimated for all possible intervals and cut point values, and an optimal retest interval and test cut point value is obtained. 

Result: The test performance and optimal cut-point changes for each repeat use of the test. Preliminary results from our hypothetical simulation suggests that, patients should be initially diagnosed using a cut-point of 0.25. Patients with a positive diagnosis should be re-tested after 4 months using a cut-point of 0.55, while those with a negative diagnosis should be re-tested after 2 months using a cut-point of 0.20. The most efficient conventional strategy is associated with an expected health loss of 50 QALYs across the population compared to the optimal strategy.

Conclusion: Maximising the value of a monitoring test requires a dynamic test cut-off with respect to population characteristics based on previous test results, and the re-test interval.