Boosting the chances to improve stroke treatment

Area of Science:

• Neurology
• Clinical Trials
• Biostatistics

Background:

• Stroke treatment trials often rely on simplified outcome measures.
• Dichotomizing ordinal scales leads to a loss of critical patient recovery data.
• Existing alternative analyses include using full ordinal scales or combining multiple scales.

Purpose of the Study:

• To evaluate information loss from collapsed ordinal scales in stroke trials.
• To compare information content across multiple stroke outcome scales.
• To determine the potential for pooling stroke outcome scales for enhanced analysis.

Main Methods:

• Reanalyzed three common stroke recovery scales: modified Rankin Scale, NIH Stroke Scale, and Barthel Index.
• Utilized multivariate analysis on data from 1652 patients (Citicoline pooling data analysis).
• Assessed common and unique information across scales by analyzing rank distances.

Main Results:

• Collapsing scales into binary outcomes (recovered/not recovered) results in significant information loss.
• The analyzed scales shared 90.7% of their information, indicating minimal unique contributions.
• Individual scale values reflect nuanced differences in patient status.

Conclusions:

• Full ordinal scales provide a more comprehensive estimate of patient outcomes in stroke.
• Stroke outcome scales are highly correlated, supporting their pooling into a single estimate.
• Optimizing analysis by using full or pooled scales can increase the power of stroke clinical trials.