Semiparametric marginal promotion time cure model for clustered survival data

2025-05-14

AI Summary

This study addresses clustered right-censored survival data with a cured subgroup, proposing a semiparametric marginal promotion-time cure model. To overcome inefficiency and nonrobust inference arising from misspecification of the within-cluster correlation structure—common in conventional approaches—we introduce, for the first time, a unified framework integrating generalized estimating equations (GEE) with quadratic inference functions (QIF). This enables consistent, asymptotically normal, and highly efficient robust estimation of regression coefficients. Simulation studies demonstrate that the proposed method substantially outperforms existing alternatives across diverse correlation structures. Applied to a periodontal disease cohort, it successfully identifies several novel clinical risk factors and reveals their heterogeneous effects on cure probability. The methodology provides a theoretically rigorous yet practically applicable tool for cure-rate modeling and clustered survival analysis.

Technology Category

• Reasoning under Uncertainty: Relational Probabilistic Models
• Machine Learning: Calibration & Uncertainty Quantification
• Cognitive Modeling & Cognitive Systems: Conceptual Inference and Reasoning

Application Category

• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Web Mining and Content Analysis: Robustness and generalizability of Web mining methods

Abstract

Modeling clustered/correlated failure time data has been becoming increasingly important in clinical trials and epidemiology studies. In this paper, we consider a semiparametric marginal promotion time cure model for clustered right-censored survival data with a cure fraction. We propose two estimation methods based on the generalized estimating equations and the quadratic inference functions and prove that the regression estimates from the two proposed methods are consistent and asymptotic normal and that the estimates from the quadratic inference functions are optimal. The simulation study shows that the estimates from both methods are more efficient than those from the existing method no matter whether the correlation structure is correctly specified. The estimates based on the quadratic inference functions achieve higher efficiency compared with those based on the generalized estimating equations under the same working correlation structure. An application of the proposed methods is demonstrated with periodontal disease data and new findings are revealed in the analysis.

Problem

• Modeling clustered survival data with cure fraction
• Proposing efficient semiparametric estimation methods
• Improving accuracy in clinical and epidemiological studies

Innovation

• Semiparametric marginal promotion time cure model
• Generalized estimating equations for estimation
• Quadratic inference functions for optimal efficiency