Time-dependent predictors in clinical research, performance of a novel method

Area of Science:

• Biostatistics
• Survival Analysis
• Epidemiology

Background:

• Patient survival predictors can change due to lifestyle modifications over time.
• Standard statistical methods, like traditional Cox regression, assume constant risk ratios, limiting their ability to analyze time-dependent factors.
• Novel time-dependent factor analysis methods have emerged to address these limitations in survival studies.

Purpose of the Study:

• To assess the performance of time-dependent factor analysis in survival studies.
• To demonstrate the inadequacy of standard Cox regression for time-varying predictors.
• To highlight the benefits of time-dependent and segmented time-dependent Cox regression models.

Main Methods:

• Application of time-dependent Cox regression models to survival data.
• Comparison of time-dependent Cox regression with the usual Cox regression.
• Utilizing SPSS statistical software for analysis.
• Employing segmented time-dependent Cox regression to adjust for peak predictor values.

Main Results:

• Time-dependent Cox regression provided better precision than usual Cox regression (P = 0.117 vs. 0.0001).
• Blood pressure, insignificant in overall analysis, became a significant survival predictor (P = 0.04) after adjustment for periods of highest blood pressure using segmented time-dependent Cox regression.
• Inclusion of time-dependent low-density lipoprotein cholesterol improved the precision of treatment modality as a survival predictor (P = 0.02 to 0.0001).

Conclusions:

• Predictors of survival, such as smoking, cholesterol, and blood pressure, can change over time.
• Time-dependent and segmented time-dependent predictors are effective for survival analysis when predictors evolve.
• Advanced analytical models are needed and available for survival analysis that adjust for time-varying predictor effects.