Improving power to detect disease progression in multiple sclerosis through alternative analysis strategies
- First Online:
- Cite this article as:
- Healy, B., Chitnis, T. & Engler, D. J Neurol (2011) 258: 1812. doi:10.1007/s00415-011-6021-1
- 145 Downloads
In patients with multiple sclerosis, investigation of a treatment effect on disease progression in clinical trials and observational studies often uses sustained progression on the expanded disability status scale (EDSS) as an outcome. It is not clear whether this outcome is the most powerful to detect a treatment effect on clinical disease progression. Assessment of EDSS modeling choice on the detection of treatment effect was of interest. This assessment was separately conducted under three potential treatment effects: treatment reducing the chance of higher future EDSS, treatment increasing the chance of lower future EDSS, and treatment leading to both effects. To assess the effect of modeling choice, nine modeling strategies were applied to the data to determine the most powerful approach. EDSS measurements were simulated at 6 month intervals for 24 months. Each patient’s initial EDSS value ranged between 0 and 3, and probabilities of transitioning from one EDSS state to another were based on the empirical probabilities of transition obtained from available clinical data. Modeling approaches based on sustained progression had less power than approaches which modeled the EDSS score directly, regardless of treatment effect. This difference was especially pronounced when the treatment effect corresponded to an increase in the probability of improvement. Sustained progression on the EDSS is a less powerful outcome measure for clinical progression than approaches based on the actual EDSS values.