Randomized controlled trials of biomarker targets

Area of Science:

• Clinical research methodology
• Biostatistics
• Causal inference

Background:

• Randomized controlled trials (RCTs) are standard for estimating treatment effects.
• Biomarker targets are increasingly used in RCTs for conditions like hypertension.
• Conflicting recommendations and inconclusive meta-analyses have emerged from trials of biomarker targets.

Purpose of the Study:

• To explain, using causal reasoning, why RCTs of biomarker targets can produce misleading conclusions.
• To identify key threats to the validity of trials that use biomarker targets.

Main Methods:

• Causal reasoning applied to analyze potential biases in biomarker target trials.
• Examination of four key threats: intention-to-treat with lack of blinding, incomparability across trials, time-varying adaptive treatment strategies, and Goodhart's law.
• Illustration using 15 RCTs of blood pressure targets for hypertension management.

Main Results:

• Intention-to-treat analysis can be misleading without blinding.
• Significant variation exists in patient populations, therapies, and treatment strategies across trials.
• Lack of accounting for time-varying treatment strategies and potential off-target effects complicates interpretation.

Conclusions:

• Meta-analyses of target trials require critical assessment for therapy variations and off-target effects.
• Future RCTs should compare treatment algorithms incorporating biomarkers, not just targets.
• Causal inference methods adjusting for time-varying confounding are recommended for observational studies.