Epidemiology methods
Bias, confounding, and effect modification: three different explanations for a result
Bias is systematic error, confounding distorts an association through another factor, and effect modification describes genuinely different effects across groups or contexts. They require different design, analysis, and reporting responses.
Educational content only. Not medical advice.
Bias systematically moves a result away from the target
Selection bias arises when entry, retention, or analysis creates groups that do not support the intended comparison. Information bias arises when exposure, outcome, or covariate measurement differs systematically or is misclassified. Increasing sample size narrows random error but does not eliminate systematic bias; design and measurement protections are needed.
Confounding mixes the effect of another factor into the association
A confounder is associated with the exposure and independently related to the outcome without being a consequence of the exposure pathway being estimated. Randomization can balance known and unknown confounders in expectation. Observational studies use restriction, matching, stratification, weighting, or modeling, but adjustment only addresses measured variables and correct model assumptions.
Effect modification is a finding to describe
Effect modification means the magnitude or direction of an association differs across levels of another variable. Unlike confounding, it is not necessarily a distortion to remove. The relevant effect scale matters: variation can appear on an absolute scale but not a relative scale, or vice versa. Subgroup comparisons require prespecification, power, interaction tests, and biological context.
Adjustment does not make an estimate automatically causal
A multivariable model can reduce measured confounding while introducing errors through overadjustment, collider conditioning, poor functional form, missing data, or measurement error. Readers should inspect the causal rationale for variables, compare crude and adjusted estimates, examine sensitivity analyses, and separate exploratory subgroup patterns from replicated effect modification.
Evidence limits
- Unmeasured and poorly measured confounding can remain after statistical adjustment.
- Bias direction and magnitude are often uncertain and may involve several mechanisms.
- Subgroup effects are vulnerable to multiplicity and low power unless planned and confirmed.
Sources and further reading
These sources ground the definitions and evidence boundaries on this page. A citation is a route for verification, not an endorsement of a product or personal use.
CDC Field Epidemiology Manual
Analyzing and Interpreting Data
Official epidemiology guidance on bias, confounding, effect modification, chance, and interpretation.
Open sourceCochrane Handbook for Systematic Reviews of Interventions
Assessing Risk of Bias in a Randomized Trial
Authoritative methods guidance on randomization, allocation concealment, deviations, missing outcomes, measurement, and selective reporting.
Open sourceU.S. Food and Drug Administration / ICH
E9(R1) Statistical Principles for Clinical Trials: Estimands and Sensitivity Analysis
Official framework connecting trial objectives, estimands, intercurrent events, analysis, and interpretation.
Open sourceCommon questions
Is confounding a type of random error?
No. It is a systematic mixing of effects that can distort an association.
Should effect modification be adjusted away?
Usually it should be described on the relevant scale because it may represent real heterogeneity.
Does regression remove all confounding?
No. It depends on measured variables, causal structure, model form, and data quality.
Continue with context
Keep building your evidence-reading skills
Explore the public PeptideSchool research library for more source-backed methods, glossaries, and evidence maps.