Simultaneously Dealing With Immortal Time Bias and Residual Confounding: A Case Study of a High-Dimensional...

Area of Science:

• Epidemiology
• Biostatistics
• Pharmacovigilance

Background:

• Observational studies of time-dependent treatments are prone to immortal time bias and residual confounding, hindering accurate treatment effect estimation.
• These biases can distort findings regarding the effectiveness of therapies like disease-modifying drugs (DMDs) in chronic conditions such as multiple sclerosis (MS).

Purpose of the Study:

• To implement a novel analytical framework combining a nested case-control (NCC) design with high-dimensional propensity score (hdPS) analysis.
• To simultaneously address immortal time bias and residual confounding in the estimation of treatment effects.
• To evaluate the association between DMDs and all-cause mortality in individuals with MS.

Main Methods:

• A retrospective cohort of 19,360 individuals with MS in British Columbia was analyzed.
• A 1:4 NCC analysis was employed to mitigate immortal time bias.
• High-dimensional propensity score (hdPS) analysis was utilized to control for residual confounding.
• Sensitivity analyses were conducted using various hdPS parameters and matching strategies to ensure robustness.

Main Results:

• The NCC-hdPS analysis included 3,209 cases and 12,293 controls.
• Exposure to DMDs was associated with a significant 28% reduction in mortality risk (HR: 0.72, 95% CI: 0.62-0.84).
• Sensitivity analyses confirmed consistent findings, with hazard ratios ranging from 0.70 to 0.77 across different methodological variations.

Conclusions:

• The combined NCC and hdPS approach provides a robust framework for unbiased estimation of treatment effects in real-world observational studies.
• This methodology effectively addresses both immortal time bias and residual confounding, enhancing the validity of research findings.
• Reproducible R code is provided to encourage the adoption of this advanced analytical technique by researchers.