Which factors confound or modify the relationship between body weight and mortality?

Area of Science:

• Epidemiologic research
• Biostatistics
• Public health

Background:

• Covariate adjustment is standard for confounding in epidemiologic research.
• Exclusions are analytic strategies for interactions altering exposure-outcome relationships.
• The BMI-mortality relationship analysis involves both methods, often due to reverse causality concerns.

Purpose of the Study:

• To clarify the appropriate use of covariate adjustment versus exclusions in BMI-mortality research.
• To address concerns of reverse causality, particularly regarding subgroups like smokers.
• To provide guidance on analytical strategies for the BMI-mortality relationship.

Main Methods:

• Review of analytical strategies in epidemiologic research, focusing on covariate adjustment and exclusions.
• Statistical framework for testing findings after large-scale exclusions to differentiate from chance results.
• Examination of evidence regarding interactions in BMI-mortality relationships across subgroups.

Main Results:

• Weight of evidence suggests interactions are generally absent in BMI-mortality analyses.
• Exclusion of subgroups (e.g., smokers, age, sex) typically leaves the mortality curve's shape unchanged.
• While exclusions may alter absolute risk levels, they do not fundamentally change the exposure-outcome relationship's shape.

Conclusions:

• Advises against large-scale exclusions for common traits in BMI-mortality analyses unless specific subgroup questions are posed.
• Emphasizes the importance of using representative study samples with measured exposure variables.
• Cites evidence from a large social experiment on population-wide BMI shifts.