The great heterogeneity of the acute ischemic stroke population makes it difficult to detect treatment effects in clinical trials. An appropriately validated predictive risk model, using data available at study entry, will improve clinical trial design by 1) allowing more rational entry criteria which exclude patients who are very unlikely to benefit from treatment, 2) permitting randomization stratified by predicted outcome thereby reducing the impact of imbalance in small trials, and 3) allowing analysis of treatment effect adjusting for baseline risk which provides greater power to detect small but clinically important effects. In addition, a surrogate measure of long term outcome based on an updated risk prediction, obtained in the early post-stroke period may prove to be highly predictive of 3 month clinical outcome. A reliable surrogate measure may reduce costs, required sample sizes and missing outcome data in clinical trials. The goal of this application is to investigate these new methodologic tools aimed at reducing the impact of this heterogeneity on clinical stroke trials. We will address the following hypotheses. Hypothesis 1: a multivariable predictive risk model including 6 acute clinical variables and 1 acute imaging variable will predict 3 month outcome in acute ischemic stroke patients. Hypothesis 2: the model is a valid predictor of 3 month outcome in 2 independent data sets. Hypothesis 3: adjustment for predicted risk will 'increase the magnitude of the identified treatment effect of tPA in the NINDS tPA trial. Hypothesis 4: an updated risk prediction based on information available at day 5 post-stroke can reliably predict 3 month outcome in acute ischemic stroke patients. To test hypotheses 1 and 2, we will use multivariable logistic regression to establish the validity of the model in a prospective cohort of 360 acute stroke patients and 2 external data sets. We will evaluate model performance (discrimination, calibration) using standard methods. Hypothesis 3 will be tested by evaluating the effect of tPA on stroke outcomes adjusted for baseline risk in the NINDS tPA trial. We will test hypothesis 4 by determining; the risk of 3-month stroke outcomes based on information available at day 5 after symptom onset. The results of this effort should facilitate study design and increase the ability to detect treatment effects in acute stroke trials. This will be valuable to patients with stroke, providers, payers and developers of new drug therapies.
Showing the most recent 10 out of 16 publications
