Left truncation in linked data: A practical guide to understanding left truncation and applying it using SAS and R

Area of Science:

• Biostatistics
• Survival Analysis
• Medical Informatics

Background:

• Time-to-event data, including survival time, are vital in medical research and drug development.
• Right censoring is a common feature of such data, with established analytical methods like Kaplan-Meier and Cox models.
• Left truncation, where subjects are excluded due to early event occurrence, introduces selection bias and immortal time bias, often overestimating survival.

Purpose of the Study:

• To provide a tutorial on analyzing time-to-event data with both left truncation and right censoring without bias.
• To illustrate potential biases, such as immortal time bias, arising from left truncation.
• To demonstrate practical implementation using SAS and R.

Main Methods:

• Utilized a nationwide electronic health record-derived de-identified database.
• Applied statistical methods designed to handle left-truncated and right-censored data.
• Provided example code in SAS and R for reproducible analysis.

Main Results:

• Demonstrated that left truncation can lead to biased results, specifically overestimation of survival time.
• Showcased methods to correct for left truncation and right censoring, yielding unbiased estimates.
• The tutorial provides practical code for implementing these unbiased analytical techniques.

Conclusions:

• Accurate analysis of time-to-event data requires addressing both left truncation and right censoring.
• Ignoring left truncation can lead to significant biases in survival estimates, impacting therapeutic evaluation.
• The presented methods and code facilitate unbiased survival data analysis in medical research.