# Effect modification. Flashcards

1
Q

When does Effect Modification occur?

A

When the strength of association varies over different categories of a third variable in an association study.

2
Q

Name four examples of effect modifiers?

A
1. Sex.
2. Age.
3. Environmental exposure.
4. Genetic effect.
3
Q

Do you modify for effect modification?

A

No. It is not an error.

4
Q

What OR’s do you look at when looking at effect modification?

A

Stratum specific OR’s.

5
Q

How do you evaluate stratum specific OR’s?

A
1. Calculate P values on homogeneity tests.

2. Evaluate overlap of CI’s on stratum specific OR’s.

6
Q

What homogeneity test do you use to calculate the p values for effect modification?

A

Breslow-day test.

7
Q

What is the conceptual definition of an effect modifier?

A

The effect of the exposure and potential modifier differ from the expected joint effect.

8
Q

What is synergism?

A

Effect modifier extenuates the effect of an exposure.

9
Q

Antagonism?

A

Effect modifier diminishes/eliminates the effect of the exposure.

10
Q

For something to be an effect modifier what must the OR be after stratifying?

A

Statistically different.

11
Q

When will an effect modifier be described as being in the ‘grey zone’ what is overlapping?

A

The confidence intervals.

12
Q

Apart from using the Mantel-Haenzsel OR what other way can you adjust for confounding?

A

Report confounders separately.

13
Q

What is the equation for the null hypothesis for the Mantel-Haenzsel test?

A

H0= OR1=OR2

14
Q

What is the alternative hypothesis for the Mantel-Haenzsel test?

A

H1= OR1 is not equal to OR2.

15
Q

What distribution is used in the Mantel-Haenzsel test?

A

Chi squared to 1 dof.

16
Q

In words what is the null hypothesis for the Mantel-Haenzsel test?

A

Stratum specific OR’s are homogenous.

17
Q

What is not considered in the Breslow-Day test?

A

Ordering in the strata.

18
Q

Why is it difficult to stratify by multiple variables in the Breslow-Day test?

A

Data becomes sparse.

19
Q

Can you control for additional covariates in the Breslow-day test?

A

No.

20
Q

What is the equation for Q?

A

Sumk (weightk x log (logORk- LogORW )^1/2

21
Q

What is the equation for LogORw (weight log odds ratio).

A

sum( logORk x weightk) / sum weightk

22
Q

What is the reference category in the multiplicative model when determining if something is an effect modifier?

A

No exposure/ No effect.

23
Q

What are you determining in the multiplicative model?

A

Marginal association of effect/ exposure on outcome to see if there is a multiplicative interaction between the effect modifier and the exposure.

24
Q

What is H0 in the multiplicative model?

A

OR++= OR-+ x OR+-

I=1.

25
Q

What is the equation for the interaction effect in the multiplicative model?

A

OR++/(OR+- x OR-+) = 1

26
Q

If I>1 what type of interaction is present?

A

Synergism.

27
Q

If I<1 what type of interaction is present?

A

Antagonism.

28
Q

What type of studies do additive models measure the effect of an effect modifier in?

A

Prospective studies.

29
Q

What is the null hypothesis for an additive model on a cohort study?

A

RD++ = RD+- + RD-+

30
Q

What is the null hypothesis for an additive model on a case-control study?

A

OR++ = OR-+ + OR+-

31
Q

Why is it difficult to assess additive effects in a case control study?

A
1. Disease risk (incidence) is not available as there is no time point.
2. Can not directly test with a logistic regression model (as this is a multiplicative model).
32
Q

What sort of regression models are used to look at additive effects?

A

Linear, Poisson.

33
Q

What sort of regression models are used to look at multiplicative effects?

A

Logistic, poisson, survival (COX) analysis.

34
Q

What does the side of an interaction effect influence?

A

The size of the sample needed to detect its effect.

35
Q

What is another name for effect modification?

A

Biological modification.

36
Q

What needs to be present for an effect modification to be properly assessed?

A
1. A clear biological mechanism.

2. All available data.

37
Q

What does a statistical modification imply?

A

That a modification has been observed in a few studies but there is no sufficient supporting evidence.

38
Q

Interactions can be seen to occur when they do not actually exist. Why? (3 things).

A
1. Chance.
2. Systematic error.
3. Varying exposure doses between strata.
39
Q

When does qualitative effect modification occur instead of quantitative? What does this mean?

A

When measures of effect in different strata have different directions. This means that effect modification occurs on both the multiplicative and additive scale.

40
Q

If effect modification occurs on both scales you will always observe this in a single study. True or false?

A

False.

41
Q

Explain the phenomenon of reciprocity in terms of effect modifier Z, exposure E and outcome D.

A

If Z modifies the effect of E on D than Z also modifies the effect of E on D.

42
Q

Does the following statement describe an effect modifier or a confounder?

Belongs to a specific study.

A

Confounder.

43
Q

Does the following statement describe an effect modifier or a confounder?

Exists in nature.

A

Effect Modifier.

44
Q

Does the following statement describe an effect modifier or a confounder?

Different effects in different strata.

A

Effect modifier.

45
Q

Does the following statement describe an effect modifier or a confounder?

Increases biological knowledge of mechanism.

A

Effect modifier.

46
Q

Does the following statement describe an effect modifier or a confounder?

Prevented in the design stage, controlled for in analysis stage.

A

Confounder.

47
Q

Does the following statement describe an effect modifier or a confounder?

Creates confusion in the data.

A

Confounder.

48
Q

Does the following statement describe an effect modifier or a confounder?

Allows targeting of public health outcome.

A

Effect modifier.