Effect Modification

Effect modification occurs when the magnitude of the effect of the primary exposure on an outcome (i.e., the association) differs depending on the level of a third variable. In this situation, computing an overall estimate of association is misleading. One common way of dealing with effect modification is examine the association separately for each level of the third variable. For example, suppose a clinical trial is conducted and the drug is shown to result in a statistically significant reduction in total cholesterol. However, suppose that with closer scrutiny of the data, the investigators find that the drug is only effective in subjects with a specific genetic marker and that there is no effect in persons who do not possess the marker. The effect of the treatment is different depending on the presence or absence of the genetic marker. This is an example of effect modification or "interaction".

Unlike confounding, effect modification is a biological phenomenon in which the exposure has a different impact in different circumstances. Another good example is the effect of smoking on risk of lung cancer. Smoking and exposure to asbestos are both risk factors for lung cancer. Non-smokers exposed to asbestos have a 3-4 fold increased risk of lung cancer, and most studies suggest that smoking increases the risk of lung cancer about 20 times. However, shipyard workers who chronically inhaled asbestos fibers and also smoked had about a 64-fold increased risk of lung cancer. In other words, the effects of smoking and asbestos were not just additive – they were multiplicative. This suggests synergism or interaction, i.e., that the effect of smoking is somehow magnified in people who have also been exposed to asbestos. Multivariable methods can also be used to assess effect modification.

A stratified analysis provides a way to identify effect modification. Recall that on the previous page we used a stratified analysis to identify confounding. When there is just confounding, the measures of association in the subgroups will differ from the crude measure of association, but the measures of association across the subgroups will be similar. In contrast, when there is effect modification, the measures of association in the subgroups differ from one another.

For a more complete discussion of the phenomena of confounding and effect modification, please see the online module on these topics for the core course in epidemiology "Confounding and Effect Modification."

Evaluation of a Drug to Increase HDL Cholesterol

Consider the following clinical trial conducted to evaluate the efficacy of a new drug to increase HDL cholesterol (the "good" cholesterol). One hundred patients are enrolled in the trial and randomized to receive either the new drug or a placebo. Background characteristics (e.g., age, sex, educational level, income) and clinical characteristics (e.g., height, weight, blood pressure, total and HDL cholesterol levels) are measured at baseline, and they are found to be comparable in the two comparison groups. Subjects are instructed to take the assigned medication for 8 weeks, at which time their HDL cholesterol is measured again. The results are shown in the table below.   

 

Sample Size

Mean HDL

Standard Deviation of HDL

New Drug

50

40.16

4.46

Placebo

50

39.21

3.91

On average, the mean HDL levels are 0.95 units higher in patients treated with the new medication. A two sample test to compare mean HDL levels between treatments has a test statistic of Z = -1.13 which is not statistically significant at α=0.05.  

Based on their preliminary studies, the investigators had expected a statistically significant increase in HDL cholesterol in the group treated with the new drug, and they wondered whether another variable might be masking the effect of the treatment. Other studies had, if fact, suggested that the effectiveness of a similar drug was différèrent in men and women. In this study, there are 19 men and 81 women.  The table below shows the number and percent of men assigned to each treatment.

 

Sample Size

Number (%) of Men

New Drug

50

10 (20%)

Placebo

50

9 (18%)

There is no meaningful difference in the proportions of men assigned to receive the new drug or the placebo, so sex cannot be a a confounder here, since it does not differed in the treatment groups. However, when the data are stratified by sex, they find the following:

 

 WOMEN

Sample Size

Mean HDL

Standard Deviation of HDL

New Drug

40

38.88

3.97

Placebo

41

39.24

4.21

 

 

 

 

MEN

 

 

 

New Drug

10

45.25

1.89

Placebo

9

39.06

2.22

 

On average, the mean HDL levels are very similar in treated and untreated women, but the mean HDL levels are 6.19 units higher in men treated with the new drug. This is an example of effect modification by sex, i.e., the effect of the drug on HDL cholesterol is is different for men and women. In this case there is no apparent effect in women, but there appears to be a moderately large effect in men. (Note, however, that the comparison in men is based on a very small sample size, so this difference should be interpreted cautiously, since it could be the result of random error or confounding.

It is also important to note, that in the clinical trials setting, the only analyses that should be conducted are those that are planned a priori and specified in the study protocol.   In contrast, in epidemiologic studies there are often exploratory analyses that are conducted to fully understand associations (or lack thereof).   However, these analyses should also be restricted to those that are biologically sensible.

LISA: [ When you say "epidemiologic studies" here, do you really mean observational studies, i.e., case-control studies and cohort studies as opposed to randomized clinical trials? I'm also not sure why this statement is here at all. It some out of place, and perhaps should be included elsewhere.]

When there is effect modification, analysis of the pooled data can be misleading.   In this example, the pooled data (men and women combined), shows no effect of treatment. Because there is effect modification by sex, it is important to look at the differences in HDL levels among men and women, considered separately.   In stratified analyses, however, investigators must be careful to ensure that the sample size is adequate to provide a meaningful analysis.