Using Evidence Factors to Clarify Exposure Biomarkers

Area of Science:

• Epidemiology
• Biostatistics

Background:

• Observational studies face challenges with bias, particularly confounding.
• Independent evidence factors can improve the reliability of treatment effect inferences.

Purpose of the Study:

• To introduce and illustrate a novel method for combining two statistically independent evidence factors from the same dataset.
• To enhance the robustness of observational study findings against bias.

Main Methods:

• Define a study with two evidence factors, where each factor is immune to specific biases invalidating the other.
• Apply meta-analysis techniques to combine evidence from statistically independent factors, even when derived from the same data.
• Illustrate the method using an example investigating the effect of cigarette smoking on homocysteine levels, employing self-reported smoking and cotinine biomarker data.

Main Results:

• The proposed method allows for statistically independent inferences from a single dataset.
• Evidence factors can be combined using meta-analysis, increasing statistical power and reducing bias.
• The joint sensitivity of self-reported smoking and cotinine levels to confounding bias was examined.

Conclusions:

• This novel approach of using two independent evidence factors strengthens observational research.
• Combining evidence from multiple, bias-immune factors improves the validity of treatment effect estimations.
• The method offers a powerful tool for analyzing complex relationships in observational data, such as the impact of smoking on health markers.