Confounder adjustment in observational studies investigating multiple risk factors: a methodological study

Area of Science:

• Epidemiology
• Biostatistics
• Observational Studies

Background:

• Accurate causal inference in observational studies relies on appropriate confounder adjustment.
• Methods for confounder adjustment in studies with multiple, non-mutually confounded risk factors are often overlooked.
• This study reviews confounder adjustment methods and issues in multiple risk factor investigations.

Purpose of the Study:

• To summarize confounder adjustment methods used in studies of multiple risk factors.
• To identify and categorize common and recommended adjustment strategies.
• To highlight potential biases arising from inappropriate confounder adjustment.

Main Methods:

• Methodological review of cohort and case-control studies published between January 2018 and March 2023.
• Searched PubMed for studies on multiple risk factors for cardiovascular disease, diabetes, and dementia.
• Classified confounder adjustment methods into six categories based on study objectives and adjustment strategies.

Main Results:

• 162 studies were included, focusing on either broad risk factor exploration (54.3%) or specific risk factors (45.7%).
• Confounder adjustment was often unsatisfactory, with only 6.2% using the recommended method (adjusting each risk factor separately for confounders).
• Mutual adjustment (including all risk factors in one model) was used in over 70% of studies, while other methods or unclear approaches were also prevalent.

Conclusions:

• Significant variation exists in confounder adjustment methods for multiple risk factors.
• The prevalent use of mutual adjustment may lead to overadjustment bias and inaccurate effect estimates.
• Future research should adopt more appropriate adjustment strategies, avoiding indiscriminate multivariable modeling.