Instrumental variable estimation of early treatment effect in randomized screening trials

Area of Science:

• Biostatistics
• Epidemiology
• Clinical Trials

Background:

• Randomized screening trials for cancer primarily use the intention-to-screen principle.
• Cancer mortality reductions in trials are influenced by screening, technology, and early treatment effects.
• There's a need to separately evaluate the impact of early treatment on mortality within the screening-detectable subgroup.

Purpose of the Study:

• To define and estimate the causal effect of early versus delayed treatment on cancer mortality among individuals detected through screening.
• To develop novel statistical methods for analyzing screening trial data to isolate treatment effects.

Main Methods:

• Formulation of a simplified structural multi-state model for screening trials.
• Development of two estimators (estimating equation and likelihood-based) for cause-specific hazard ratios.
• Extension of instrumental variable methods to handle time-dependent intermediate variables.
• Simulation studies to evaluate estimator performance using a data-generating mechanism based on the multi-state model.

Main Results:

• The proposed methods extend existing instrumental variable approaches for time-to-event and competing risks data.
• Simulation studies demonstrate the performance of the new estimators.
• The methods are illustrated using data from the US National Lung Screening Trial (NLST) for CT screening for lung cancer.

Conclusions:

• The developed methods provide a way to estimate the causal effect of early treatment on cancer mortality in screening-detectable populations.
• These approaches can disentangle the effects of screening from the effects of early treatment.
• The application to lung cancer screening trial data demonstrates practical utility.