Utility of the stroke-thrombolytic predictive instrument

¹ University Medical Center Groningen, Netherlands

^* To whom correspondence should be addressed. E-mail: m.uyttenboogaart{at}neuro.umcg.nl.

Accepted 1 April 2008

	Abstract

Objective: The goal of this study was to evaluate the utility of the stroke thrombolytic predictive instrument (s-TPI) in predicting clinical outcome in acute ischaemic stroke patients treated with intravenous tissue plasminogen activator (t-PA).

Methods: We assessed the external validity of the s-TPI in 301 consecutive stroke patients treated with intravenous t-PA. Clinical outcome was measured with the modified Rankin scale (mRs) at 3 months. We used the s-TPI to calculate probabilities of a good outcome (mRs 0-1) and a very poor outcome (mRs 5-6). We compared these probabilities with the observed outcome using receiver-operator characteristics (ROC) curves and calibration curves. Subgroup analyses for different onset to treatment time windows were performed.

Results: According to the s-TPI, the mean predicted probability of a good and a very poor outcome in the validation cohort were 0.45 and 0.17. The area under the ROC curves were 0.80 (4.5 hour time window), 0.82 (3 hour time window) and 0.77 (3-4.5 hours time window) for predicting good outcome, and 0.78 (4.5 hours), 0.80 (3 hours) and 0.74 (3-4.5 hours) for predicting very poor outcome. Calibration curves revealed a slight overestimation of probabilities of a good outcome and underestimation of probabilities of a very poor outcome.

Conclusions: The s-TPI appears to be reasonably valid for predicting outcome after t-PA treatment in daily practice, although a slight overestimation of a good and underestimation of a very poor outcome was observed.