Adjusting for informative cluster size in pseudo-value based regression approaches with clustered time to event data

Stat Med, 2023 · DOI: 10.1002/sim.9716 · Published: June 15, 2023

Bioinformatics Research Methodology & Design

Simple Explanation

This paper addresses how to analyze clustered time-to-event data using pseudo-value regression when the size of the clusters is informative. Informative cluster size (ICS) means there's a relationship between the number of participants in a cluster and the outcome being measured. The authors explore different strategies for adjusting for ICS by reweighting, in the context of pseudo-value regression, which is used to model the effect of covariates on the progression of a disease. The paper includes theoretical arguments and simulation experiments to determine the most accurate strategy for adjusting for ICS. The methods are demonstrated using real-world datasets from a periodontal study and a study of spinal cord injury patients undergoing locomotor-training rehabilitation.

Study Duration

Not specified

Participants

Spinal cord injury: 497 patients. Periodontitis: 99 patients.

Evidence Level

Not specified

Key Findings

1
The study identifies appropriate strategies for adjusting for informative cluster size (ICS) in pseudo-value regression analysis.
2
Simulation experiments show that correct inference for covariate effects is achieved by adjusting for ICS in the estimating equations, especially using cluster-weighted GEE (CWGEE).
3
The CWGEE method consistently outperforms the standard GEE method in the presence of ICS, providing less biased estimates and better coverage probabilities.

Research Summary

The paper addresses the problem of informative cluster size (ICS) in the context of pseudo-value regression for clustered time-to-event data. It explores different strategies for ICS adjustment, including reweighting methods, to improve the accuracy of inferences. Through theoretical arguments and simulation experiments, the study identifies the correct strategy for adjusting for ICS in pseudo-value regression, which involves appropriate reweighting in the estimating equations. The proposed methods are demonstrated in real-world applications, including a periodontal study and a study of spinal cord injury patients, showcasing the practical relevance and effectiveness of the developed techniques.

Practical Implications

Improved Statistical Inference

Using the correct ICS adjustment strategy in pseudo-value regression leads to more accurate and reliable statistical inferences about covariate effects.

Enhanced Data Analysis

The proposed methods enable researchers to effectively analyze clustered time-to-event data, particularly in situations where cluster size is informative.

Broad Applicability

The findings have broad implications for medical research and other fields where clustered data and informative cluster size are common, such as dental studies and rehabilitation research.

Study Limitations

1
The pseudo-value regression model is only an approximate model for the true underlying model that generates the transition times
2
For a small number of clusters, the corresponding p-values of the estimated coefficients from the pseudo-value regression may be unreliable.
3
Additional complications may arise due to the interaction of the various weights used for adjustments and the impact of the covariates on censoring, ICS, and transition times.

Your Feedback

Was this summary helpful?

Back to Bioinformatics