The NINDS rt-PA Stroke Study demonstrated that intravenous thrombolytic therapy (TT) improves outcomes in acute ischemic stroke (AIS), when delivered within 3 hours of symptom-onset, despite a significant risk of thrombolytic-related intracranial hemorrhage (ICH). However, multiple trials including patients beyond this time window have failed to find any benefit for TT in AIS. In part because of the narrow time window of therapeutic opportunity, less than 5% of all AIS patients are currently treated with TT. In order for TT to have a major-impact-stroke outcome, it is critical to find patients who might benefit when treated beyond this time window. We hypothesize that the risk of thrombolytic-related ICH in AIS, and the likelihood of benefit from TT, are significantly influenced by individual patient characteristics. Based on our experience using similar methods analyzing outcomes in acute myocardial infarction, and based on our preliminary analyses of AIS trials, we propose to demonstrate that: 1) A substantial subgroup of """"""""low risk/high benefit"""""""" patients treated beyond the 3-hour time window can benefit from TT, and 2) this """"""""low risk/high benefit"""""""" subgroup of patients can be identified on tile basis of easily obtainable pre-treatment clinical information. To accomplish this, we will undertake the following: 1) Create a combined database from several major randomized clinical trials testing thrombolytic therapy for AIS. 2) Develop statistical models that predict the risks and the benefits of TT in patients with AIS. 3) Use the above multivariate models to identify patients with a highly favorable risk- benefit profile likely to benefit from TT for AIS, even when treated more than 3 hours after the onset of stroke symptoms. 4) Plan and conduct a pilot study to evaluate the feasibility of """"""""real-time assisted multi-dimensional patient selection"""""""" for a clinical trial testing TT in AIS more than 3 hours from symptom-onset. In order to accomplish the above Aims, I will concurrently undertake a structured training program to: 1) Refine and consolidate my research skills in logistic-regression predictive modeling, including extending methods I have employed in one clinical domain (acute myocardial infarction) to another (AIS), 2) become familiar with new modeling techniques (such as neural networks and classification trees), 3) become more familiar with the development and technology transfer of predictive instruments for real-world clinical use. ? ?
Showing the most recent 10 out of 18 publications
