Using change in the National Institutes of Health Stroke Scale to measure treatment effect in acute stroke trials.

BACKGROUND AND PURPOSE Outcome measures in acute stroke trials are being refined. Changes in neurological deficits might be useful outcome measures because they can measure the entire spectrum of deficits. METHODS We analyzed data from the acute stroke treatment trial Trial of Org 10172 in Acute Stroke Treatment (TOAST). Using logistic regression analysis, we modeled the probability of the TOAST predefined very favorable outcome (VFO; both Glasgow Outcome Scale 1 and modified Barthel Index 19 to 20) at 3 months. Within-subject changes (baseline-3 months) on the National Institutes of Health Stroke Scale (NIHSS) was the main predictor of interest. RESULTS The baseline median NIHSS for the entire TOAST cohort was 7, and it improved by 4 points (interquartile range 3 to 6) among 603 patient with VFO and by 2 points (interquartile range -1 to 5) among 638 patients without a VFO (P<0.001). The odds for VFO increased by 2.29 (95% CI, 2.06 to 2.54; P<0.001) for each 1-point improvement on the NIHSS. In receiver operating characteristic analysis, final NIHSS < or =2 was a good predictor of VFO, but no single NIHSS change cut point was a good predictor of VFO. CONCLUSIONS NIHSS change appears to be a useful outcome measure for acute stroke trials and is not fully comparable to dichotomized functional outcomes.