Pitfalls in time-to-event analysis of registry data: a tutorial based on simulated and real cases

Area of Science:

• Biostatistics
• Epidemiology
• Medical Research Methodology

Background:

• Survival analysis (time-to-event analysis) is crucial for understanding time to specific outcomes.
• Methodological errors in survival analysis can lead to biased results.
• Rare disease registries, like CEREDIH, offer valuable data for such analyses.

Purpose of the Study:

• To highlight essential elements for correct initial steps in survival analysis.
• To demonstrate the impact of methodological errors using simulations and real-world data.
• To provide clinicians and healthcare professionals with knowledge on time-to-event data analysis.

Main Methods:

• Utilized non-parametric methods for survival analysis.
• Focused on handling right censoring, left truncation, competing risks, and recurrent events.
• Employed simulations and real-life data from the French national registry of primary immunodeficiencies (CEREDIH).

Main Results:

• Simulations confirmed that ignoring censoring, truncation, competing risks, and recurrent events induces significant bias.
• Correct application of non-parametric methods is shown to yield unbiased results.
• Demonstrated practical application of these methods using the CEREDIH registry data.

Conclusions:

• Appropriate statistical techniques are vital for accurate survival analysis, especially in rare disease research.
• Non-parametric methods effectively address complexities like censoring and competing risks.
• This tutorial enhances the understanding and application of time-to-event analysis for healthcare professionals.