Bounding the per-protocol effect in randomized trials: an application to colorectal cancer screening

Area of Science:

• Epidemiology
• Biostatistics
• Preventive Medicine

Background:

• Per-protocol effect estimation in randomized trials often relies on unverifiable assumptions.
• Partial identification methods offer a promising approach to estimate bounds for the per-protocol effect under more plausible assumptions.

Purpose of the Study:

• To estimate bounds for the per-protocol effect of colorectal cancer screening using data from the Norwegian Colorectal Cancer Prevention trial.
• To assess the range of effectiveness and potential harm of sigmoidoscopy screening.

Main Methods:

• Utilized data from a randomized trial of 98,792 individuals aged 50-64 years.
• Estimated bounds for colorectal cancer incidence, mortality, and all-cause mortality over 10 years.
• Employed partial identification strategies under the assumption that randomization influenced outcomes only through screening.

Main Results:

• The 10-year risk difference for colorectal cancer was bounded between -0.6% and 37.0%.
• Bounds for colorectal cancer mortality ( -0.2% to 37.4%) and all-cause mortality (-5.1% to 32.6%) were also estimated.
• The bounds quantified maximum possible effectiveness but could not rule out harm.

Conclusions:

• Bounding the per-protocol effect clarifies reliance on unverifiable assumptions.
• This approach highlights the range of effect sizes with higher confidence.
• It can inform expectations about differential benefits or harms in subpopulations.