Inference in randomized studies with informative censoring and discrete time-to-event endpoints

Area of Science:

• Biostatistics
• Clinical Epidemiology
• Psychiatric Research

Background:

• Analyzing time-to-event data in clinical trials is crucial for understanding treatment efficacy.
• Right-censored data is common in longitudinal studies, posing analytical challenges.
• Informative censoring, where censoring depends on prognostic factors, can bias results.

Purpose of the Study:

• To present a novel statistical method for estimating and comparing treatment-specific distributions of discrete time-to-event variables.
• To enable adjustment for informative censoring based on measured prognostic factors.
• To quantify the sensitivity of inferences to unmeasured factors causing residual dependence between event time and censoring.

Main Methods:

• Development of a statistical methodology for discrete time-to-event data analysis.
• Incorporation of adjustment for measured covariates influencing both event time and censoring.
• Quantification of sensitivity to unmeasured confounders using a defined approach.
• Validation through a simulation study to assess practical performance.

Main Results:

• The proposed method effectively estimates treatment-specific time-to-event distributions.
• Adjustment for informative censoring improved the accuracy of the estimates.
• Sensitivity analyses provided insights into the potential impact of unmeasured factors.
• Simulation results demonstrated the practical utility and performance of the methodology.

Conclusions:

• The developed method offers a robust approach for analyzing right-censored discrete time-to-event data in clinical trials.
• It addresses key challenges of informative censoring and unmeasured confounding.
• The methodology is applicable to chronic schizophrenia trials and other similar research settings.