Confounding and interactions

Question 1

Define a confounder

Answer 1

• Confounding is confusion or mixing of effects:
• The factor must be independently associated with the outcome being investigated
• The factor must be associated with the exposure being investigated
• Not be in the causal pathway between exposure and outcome

Question 2

How can you evaluate confounding

Answer 2

• Determine if potential confounder (e.g., age) is associated with the disease outcome and
• Determine if frequency of potential confounder differs by exposure of interest

alternatively

• Stratify on confounder and compute the stratum-specific odds ratios (ORs):
  
  • If OR has same direction and magnitude in each stratum but differs from overall unadjusted OR–> confounding
  • If OR differs by stratum of confounder –> effect modification (interaction)
• Stratify on confounder and compute the stratum-specific odds ratios (ORs):
  
  • If OR has same direction and magnitude in each stratum but differs from overall unadjusted OR –> confounding
    
    • Typically a 10% change from crude to adjusted OR (Mantel Haenszel procedure) indicates presence of confounding
    • Be sure to report threshold
• If OR differs by stratum of confounder –> effect modification (interaction)
  
  • Effect modification/interaction occurs when the measure of effect varies by a third variable (test for homogeneity)
  • Joint effects of the risk factor and the third variable differs from that expected on the basis of their independent effects

Question 3

What are approaches to handling confounding

Answer 3

• In designing study:
  
  • Restriction of eligibility criteria, e.g., only outdoor workers or only narrow age range so as to be homogeneous for disease incidence
  • Individual or group matching
  • Randomization
• In analysis of data:
  
  • Stratification, using Mantel-Haenszel for a weighted summary
  • Adjustment: direct or indirect standardization, multivariable regression

Question 4

What is effect modification

Answer 4

• The magnitude or direction of an association varies according to levels of a third variable
  
  • When the incidence rate of disease in the presence of two or more risk factors differs from the incidence rate expected from their individual effects
• Commonly referred to as:
  
  • Effect measure modification
  • Interaction
  • Unlike confounding, effect modification should be described and reported, rather than controlled for
• Keep in mind that the presence of effect modification depends on which measure of effect is evaluated (e.g. risk difference, risk ratio, etc.)
• Should always specify which scale was used in the analysis
• The risk difference (attributable risk) is on an additive scale
• The risk ratio is on a multiplicative scale

Question 5

Evaluating interactions and confounding by stratification

Answer 5

• First calculate the crude OR (not stratified by the possible effect modifier or confounder)
• Stratify by the possible effect modifier/confounder:
  
  • If a the factor is a confounder, the stratum-specific ORs should be of the same direction and magnitude, and different from the crude OR
  • If the factor is an effect modifier, then the stratum-specific ORs will be different from each other as well as from the crude OR