Baseline-adjusted proportional odds models for the quantification of treatment effects in trials with ordinal sum score outcomes

Researchers often use sum scores of ordinal outcomes in clinical trials. Common methods lack power or the ability to incorporate baseline information. This paper introduces baseline-adjusted proportional odds logistic regression models to address these limitations. The method's validation focuses on ordinal sum score outcomes from neurological clinical trials, such as the upper extremity motor score (UEMS) and the spinal cord independence measure (SCIM). The study compares the statistical power of the novel models to conventional approaches. The simulation study demonstrated that the statistical power of the new method was greater than that of conventional methods. The proposed models allow for direct interpretation of the global treatment effect and have superior statistical power.

• 1
  The simulation study demonstrated that the statistical power of the novel method was greater than that of conventional methods.
• 2
  Baseline adjustments were more suited for the analysis of the upper extremity motor score compared to the spinal cord independence measure and its self-care subscore.
• 3
  The newly introduced permutated ePolr test outperformed conventional analysis methods in every simulation setup.