Unequal intra-group variance in trajectory classification.

Area of Science:

• Biostatistics
• Clinical Research Methodology
• Data Science

Background:

• Longitudinal patient classification, or trajectory classification, is common in clinical research.
• Existing models often assume equal variance within patient groups, which may not reflect disease heterogeneity.
• Unequal variances between healthy and diseased subjects can lead to inaccurate patient stratification.

Purpose of the Study:

• To develop and evaluate a novel trajectory classification model that accommodates unequal intra-group variances.
• To assess the impact of this new model on classification accuracy compared to traditional models.
• To demonstrate the model's utility in a clinical setting.

Main Methods:

• Developed a classification Expectation Maximization (EM) algorithm to estimate trajectories and group classifications.
• Incorporated unequal intra-group variance into the classification model.
• Evaluated model performance using simulations and a clinical study on cyclosporine A dosage post-cardiac transplantation.

Main Results:

• Simulations revealed high misclassification rates (up to 50%) with equal variance assumptions when real variances differed.
• The new model with unequal variances significantly reduced misclassification rates.
• In a clinical trial, the unequal variance model yielded more meaningful patient groups and highlighted low-dose cyclosporine A benefits.

Conclusions:

• The proposed trajectory classification model effectively handles unequal intra-group variances, crucial for heterogeneous clinical data.
• This approach enhances patient group meaningfulness and accuracy, particularly in disease research.
• The model is recommended for trajectory classification, especially when sample sizes and repeated measurements are adequate.