Challenges of guarantee-time bias.

Area of Science:

• Biostatistics
• Clinical Epidemiology
• Survival Analysis

Background:

• Guarantee-time bias (GTB), also known as immortal time bias, is a significant challenge in survival analyses.
• It arises when comparing groups defined by an event occurring during follow-up, potentially distorting results for outcomes like disease-free survival (DFS).
• Identifying and mitigating GTB is crucial for accurate interpretation of clinical trial data.

Purpose of the Study:

• To define guarantee-time bias (GTB) and its implications in survival analysis.
• To present and evaluate three analytical techniques for removing GTB: conditional landmark analysis, extended Cox model, and inverse probability weighting.
• To illustrate the impact of GTB using bisphosphonate use and DFS in the BIG 1-98 trial.

Main Methods:

• The study defines GTB and reviews its occurrence in published literature.
• It discusses three statistical methods to correct for GTB: conditional landmark analysis, extended Cox models, and inverse probability weighting.
• The BIG 1-98 trial data was re-analyzed to compare naive approaches with GTB-corrected methods.

Main Results:

• A naive analysis of bisphosphonate use on DFS in the BIG 1-98 trial suggested a substantial benefit.
• However, analyses employing conditional landmark analysis, extended Cox models, and inverse probability weighting found no statistically significant reduction in DFS events with bisphosphonate therapy.
• These findings highlight the potential for GTB to inflate perceived treatment effects.

Conclusions:

• Guarantee-time bias can lead to erroneous conclusions in survival analyses, particularly when treatment assignment is based on events during follow-up.
• Conditional landmark analysis, extended Cox models, and inverse probability weighting are effective methods for removing GTB.
• Accurate assessment of treatment efficacy requires careful consideration and application of bias-correction techniques in survival data analysis.