Odds are the sign is right

Area of Science:

• Biostatistics
• Epidemiology
• Statistical Modeling

Background:

• Sensitivity analysis is crucial for evaluating the robustness of study findings.
• Existing methods often rely on assumptions about unmeasured covariates, which may be mediating.
• A need exists for methods applicable when unmeasured covariates are confounding.

Purpose of the Study:

• Introduce a novel odds ratio condition for sensitivity analysis.
• Provide a method that can accommodate unmeasured dichotomous confounding variables.
• Demonstrate the reliability and applicability of the proposed condition.

Main Methods:

• Developed a new condition for sensitivity analysis based on odds ratios.
• Conducted statistical simulations to compare the proposed condition with existing methods.
• Applied the odds ratio condition to analyze the association between Zika virus and birth defects.

Main Results:

• The odds ratio condition demonstrated comparable reliability to other common sensitivity analysis conditions.
• The proposed method is effective when the unmeasured covariate acts as a confounder.
• The analysis of Zika virus and birth defects illustrated the practical application of the odds ratio condition.

Conclusions:

• The odds ratio condition offers a valuable tool for sensitivity analysis, particularly in the presence of unmeasured confounding.
• This approach enhances the rigor of statistical inference in observational studies.
• The method provides a reliable way to interpret associations, such as the link between Zika virus and adverse birth outcomes.