Causal Interaction in Clinical Epidemiology: Concepts, Measurement, and Interpretation

Mayta
Mar 17
4 min read

Updated: Mar 25

Introduction

In clinical epidemiology, most diseases arise from the combined influence of multiple exposures rather than a single cause. Understanding how these exposures act together is essential for identifying causal mechanisms and designing effective prevention strategies.

Causal interaction refers to a situation in which two exposures jointly influence an outcome in a way that differs from what would be expected based on their individual effects. This concept is central to causal inference, where the goal is to determine whether observed associations reflect true causal relationships rather than confounding, bias, or chance.

A general causal framework can be expressed as:

Y=f(X1,X2,X1×X2∣confounders)

where X1 and X2 are exposures and X1×X2 represents their joint effect.

Definition of Causal Interaction

Causal interaction refers to a situation in which two exposures jointly influence an outcome in a way that cannot be explained by simply combining their individual effects.

Consider two exposures:

X₁ = Exposure 1 (e.g., diabetes)
X₂ = Exposure 2 (e.g., hypertension)
Y = Outcome (e.g., stroke)

If the combined effect of diabetes and hypertension on stroke risk is greater (or smaller) than what we would expect based on their independent effects, then interaction is present.

In epidemiologic analysis, this joint effect is often referred to as the joint exposure effect or joint effect. Understanding these relationships helps reveal how multiple determinants operate together to influence disease risk.

Statistical Framework for Interaction

Epidemiologists typically assess interaction using two conceptual scales:

Multiplicative interaction
Additive interaction

These two approaches capture different perspectives on how risk factors combine.

Multiplicative Interaction

Multiplicative interaction evaluates whether the combined effect of two exposures is greater or smaller than the product of their individual effects.

This type of interaction is usually assessed using regression models, such as:

Logistic regression
Poisson regression
Cox proportional hazards regression

Effect measures used in these models include:

Risk Ratio (RR)
Odds Ratio (OR)
Prevalence Ratio (PR)

Example

Suppose:

RR for exposure X₁ = 2
RR for exposure X₂ = 3

Under a multiplicative model, the expected joint effect would be:

If the observed joint effect is RR = 10, this suggests positive multiplicative interaction, meaning the combined effect is stronger than expected.

Multiplicative interaction is commonly detected through interaction terms in regression models.

Additive Interaction

Additive interaction evaluates whether the combined absolute risk from two exposures exceeds the sum of the risks associated with each exposure individually.

This perspective is particularly important for public health and prevention, because it focuses on the excess number of cases attributable to the interaction between exposures.

Three commonly used measures quantify additive interaction:

1. Relative Excess Risk due to Interaction (RERI)

Measures the excess risk attributable to interaction beyond the sum of individual risks.

2. Attributable Proportion (AP)

Represents the proportion of disease among individuals exposed to both factors that is attributable to their interaction.

3. Synergy Index (SI)

Measures the strength of the synergistic interaction between exposures.

Why Additive Interaction Matters in Public Health

While multiplicative interaction is widely used in statistical modeling, additive interaction is often more relevant for clinical and public health decision-making.

Additive interaction answers practical questions such as:

How many cases could be prevented if one exposure were eliminated?
Which combinations of risk factors contribute most to disease burden?

Because additive interaction quantifies excess disease risk, it directly informs:

Prevention strategies
Risk stratification
Resource allocation in healthcare systems

The Interaction Continuum for Causal Interpretation

When evaluating causal synergy between exposures, researchers often consider evidence from both interaction scales.

A general interpretation framework is as follows:

Demonstrating additive interaction is often considered stronger evidence of meaningful causal synergy.

However, detecting additive interaction can be difficult because it usually requires very large sample sizes to achieve sufficient statistical power.

Interaction vs Effect Modification

The concept of interaction is sometimes confused with effect modification, but these ideas address different questions.

Interaction

Interaction evaluates the combined effect of two exposures on an outcome.

Example:

Smoking
Asbestos exposure

Together, they dramatically increase lung cancer risk beyond their independent effects.

Effect Modification

Effect modification occurs when the effect of a single exposure differs across subgroups of another variable.

Example:

The effect of a treatment may differ between:

Men and women
Younger and older patients

In this case, the modifying variable changes the strength of the exposure–outcome relationship, rather than acting as a second exposure contributing jointly to risk.

Conclusion

Causal interaction is a key concept in clinical epidemiology that helps explain how multiple exposures jointly influence disease outcomes. By examining both multiplicative and additive interaction, researchers can better understand the mechanisms underlying disease risk and identify opportunities for effective prevention.

Although multiplicative interaction is commonly assessed in regression models, additive interaction often provides greater insight for clinical decision-making and public health planning, because it directly reflects the number of cases attributable to interacting risk factors.

Understanding the distinction between interaction and effect modification is also essential for correctly interpreting epidemiological analyses and designing studies that address meaningful clinical questions.