An efficacy evaluation method for non-normal outcomes in randomized controlled trials.

Area of Science:

• Biostatistics
• Clinical Trials
• Medical Data Analysis

Background:

• Randomized controlled trials (RCTs) commonly use linear regression for efficacy evaluation.
• Ordinary least squares (OLS) regression is inefficient when residuals are not normally distributed, impacting parameter estimation accuracy.
• Confounding variables and non-normal residuals pose challenges in clinical efficacy studies.

Purpose of the Study:

• To introduce a novel exponential squared loss (ESL) model for robust treatment effect evaluation in clinical trials.
• To address limitations of OLS regression in the presence of non-normal residuals and confounding variables.
• To enhance estimation efficiency in efficacy studies with non-normal data.

Main Methods:

• Development of an exponential squared loss (ESL) model for linear efficacy evaluation.
• Application of the ESL model to analyze Alzheimer's Disease Assessment Scale-cognitive (ADAS-cog) scores in mild cognitive impairment (MCI) studies.
• Comparative analysis with ordinary least squares (OLS) regression using simulation and real-world data.

Main Results:

• The ESL model provides robust estimation for non-normal data, outperforming OLS regression with contaminated data.
• Simulations demonstrated superior performance of ESL over OLS, particularly when non-normal contamination reached 30%.
• Application to an MCI study showed results consistent with existing medical literature, validating the ESL model's efficacy.

Conclusions:

• The proposed ESL model effectively overcomes limitations of confounding variables and non-normal residuals in RCT efficacy studies.
• ESL offers improved estimation efficiency compared to OLS in challenging data scenarios, making it suitable for real-world clinical research.
• This method enhances the reliability and accuracy of treatment effect evaluation in diverse clinical trial settings.