Investigating drug plasma levels and clinical response using random regression models

Area of Science:

• Psychiatry
• Biostatistics
• Longitudinal Data Analysis

Background:

• Longitudinal psychiatric studies often face challenges with missing data.
• Traditional statistical methods like repeated measures MANOVA can be limited by incomplete datasets.
• The Riesby dataset provides a valuable resource for evaluating analytical approaches in psychiatric research.

Purpose of the Study:

• To reanalyze the Riesby dataset using a random regression model.
• To compare the effectiveness of random regression with traditional methods (MANOVA) for longitudinal psychiatric data.
• To investigate the effects of DMI and IMI plasma measurements on Hamilton Depression Rating Scale (HAM-D) scores.

Main Methods:

• Reanalysis of the Riesby dataset using a random regression model.
• Comparison with repeated measures MANOVA.
• Utilizing HAM-D scores and HAM-D change from baseline scores as dependent measures.
• Assessment of DMI, IMI, and endogenous effects, as well as autocorrelation.

Main Results:

• Random regression showed a significant effect of DMI plasma measurements on HAM-D scores, particularly when using HAM-D change from baseline.
• No significant effect of IMI was found.
• Marginally significant endogenous effect observed with actual HAM-D scores.
• Evidence of marginally significant autocorrelation in residuals.
• MANOVA failed to detect a significant DMI effect due to data exclusion.

Conclusions:

• Random regression models offer a robust alternative for analyzing longitudinal psychiatric data with missing values.
• DMI plasma measurements are significantly associated with depression symptom changes.
• The random regression approach is superior to MANOVA when dealing with incomplete longitudinal data in psychiatric research.