Automated Selection of Patients for Clinical Trials Eugene Fink Lawrence O. Hall Dmitry B. Goldgof Bhavesh D. Goswami Matthew Boonstra Jeffrey P. Krischer
A clinical trial is an evaluation of a new treatment procedure. Clinical trials When physicians conduct a trial, they recruit patients with matching health problems and medical histories.
Selection of patients The selection of trial participants is a manual procedure, and physicians may miss eligible patients. Gotay [1991] demonstrated that physicians choose only 39% of the eligible patients Fallowfield et al. [1997] showed that they choose less than 50% of the eligible patients
Expert system We have developed a system that automatically selects prospective trial participants.
Outline Knowledge base Selection results Cost reduction
Knowledge base Tests and questions Eligibility criteria
Tests and questions Example Mammogram, $150 What is the cancer stage? Does the patient have invasive cancer? Biopsy, $400 How many lymph nodes have tumor cells? What is the greatest tumor diameter? The knowledge base includes a list of medical tests. A test description includes a dollar cost and related questions about a patient’s health.
Eligibility criteria The knowledge base also includes logical expressions that represent eligibility for the available trials. AND cancer-stage {II, III} OR invasive-cancer = NO lymph-nodes 3 tumor-diameter 2.5 Example
Selection process The system collects data until it can determine whether the eligibility expressions are TRUE or FALSE. If patient records do not provide enough data, the system identifies the required medical tests.
Outline Knowledge base Selection results Cost reduction
Experiments We have used fifteen breast-cancer trials at the Moffitt Cancer Center. Past data from 187 patients Current data from 169 patients
Finding eligible patients Past data for 187 patients Clinical Trial Parti- cipants Other Eligible
Finding eligible patients Current data for 169 patients Clinical Trial Parti- cipants Other Eligible
Participation in other trials Current data for 169 patients Clinical Trial Incom- patible Compa- tilbe No Other Trial
Summary The results suggest that the system can increase the number of trial participants by a factor of three.
Outline Knowledge base Selection results Cost reduction
Medical tests The selection of trial participants may require medical tests. The total cost of tests may depend on their ordering Finding the right ordering is often a complex problem
Ordering of tests The system chooses the ordering of tests that reduces their expected total cost. The ordering is based on the test costs, number of trials that require each test, and structure of eligibility expressions. After getting the results of the first test, it revises the ordering of the remaining tests.
Cost reduction Past data for 187 patients Clinical Trial Mean Cost W/O Test Reordering With Test Reordering $70. $209. $35. $11. $60. $19.
Cost reduction Past data for 187 patients $0 $250 $200 $150 $100 $ W/O Reordering With Reordering
Cost reduction Current data for 169 patients Clinical Trial Mean Cost W/O Test Reordering With Test Reordering $192. $36. $107. $192. $3. $107.
Cost reduction Current data for 169 patients $0 $250 $200 $150 $100 $ W/O Reordering With Reordering
Conclusions We have developed a system that selects prospective trial participants. Helps physicians to identify eligible patients Can increase the number of trial participants Can reduce the cost of the selection process
Future work Deploy the system at the Moffitt Cancer Center Evaluate the satisfaction of physicians and nurses Add probabilistic reasoning