Alternative Explanations For An Association

Question 1

What could you look at to help assess the role of chance?

Answer 1

• Confidence interval around measure of effect
• P-value
• Sample size (needs to be large enough)

Question 2

What is reverse causation? What is difficult to establish?

Answer 2

Reverse causation is the concept by which instead of the exposure causing the outcome, the outcome actually caused the exposure

Timing of onset of disease in relation to exposure is often difficult to establish

Question 3

When is reverse causation likely to be a problem?

Answer 3

More likely to be an issue in studies that take a snapshot of the population:

• case reports/series
• cross-sectional
• case-control

In these study types it is harder to establish if the exposure really did happen before the disease

Question 4

How is reverse causation overcome?

Answer 4

By ensuring the exposure measurements relate to the period before the disease onset, allowing for the fact that there may be a preclinical stage of the disease and that the person may experience symptoms prior to a diagnosis being made

Question 5

What is confounding and when does it occur?

Answer 5

Confounding is when you see an association between exposure and disease/outcome that is not a true association

This occurs when the association is distorted by some other exposure which is independently related to both the outcome and exposure of interest

Question 6

Give an example of confounding

Answer 6

In a study of occupation and lung cancer, miners were seen to be more likely to get lung cancer than others

Rather than coal dust causing lung cancer, the observed association may actually be due to the fact that miners smoke more than the general population and that smoking is the true cause of the increased risk

In this example, coal dust is the exposure, lung cancer is the outcome and smoking is the possible confounder

Question 7

When can a variable NOT be a confounder?

Answer 7

When it is on the causal pathway

E.g. Blood pressure would not be considered a confounder in the relationship between exercise and heart attack

Question 8

If confounder still are known, what should we do?

Answer 8

Collect data on them so we can then include them in our analysis

Question 9

What are the 3 main ways in which confounders distort an association?

Answer 9

Confounders can work in any direction:

• They can overestimate an association whereby the relation is deceased when the confounder is accounted for
• They can mask a true relation between exposure and disease (initially there is no apparent relation but after allowing for the confounder an association appears)
• They can change the direction of an association completely (an apparent positive association may become negative when the confounder is allowed for, and vice versa)

Question 10

How can we deal with confounding?

Answer 10

1. Study design (randomisation/restriction/matching)

2. Analysis (stratification/multivariate)

Question 11

What is randomisation and how can it combat confounding?

Answer 11

Individuals are assigned to the exposed and unexposed groups in an entirely random manner

This means that all confounding issues (known/Unknown) are distributed evenly across both conditions I.e. It breaks the relationship between confounder and exposure

However, this won’t work for all study designs, but is good for randomised controlled trials (e.g. Placebo vs. Drug)

Question 12

What is restriction and what is an issue with this way of combatting confounding?

Answer 12

Here, the study group is restricted to only one level of the confounding variable (e.g. Just focus on non-smoking miners)

However, this limits the number of participants in the study and also limits how generalisable the findings are to the general population

Question 13

What is matching in the context of confounding?

Answer 13

This is where a control is matched to each case based on potential confounders (e.g. Same age and gender)

Special statistics are required to analyse the data

Question 14

Is confounding more commonly controlled for by study design or analysis?

Answer 14

Analysis

Question 15

What is stratification and what does it rely on?

Answer 15

Stratification is when you carry out analysis and compute a measure of effect separately for each level of the potential confounder

It relies on the relevant information on the confounding factors being collected accurately

Question 16

What is the difference between restriction and stratification?

Answer 16

Restriction takes place at the study design stage, whereas stratification allows all the data to be collect and then the analysis is restricted/stratified

Question 17

What do multivariate analysis techniques effectively do in the context of confounding?

Answer 17

They adjust the estimated measure of effect, e.g. OR, to control for the confounders

Question 18

Name 6 common confounders

Answer 18

• Age
• Gender
• Socioeconomic status
• Diet
• Ethnicity
• Smoking habit

Question 19

What are the 5 possible explanations for an association?

Answer 19

1. Truth
2. Chance (sampling error)
3. Reverse causation
4. Confounding
5. Bias

Question 20

What is bias?

Answer 20

Systematic error in an estimate arising from problems in the study design/execution

It results in a measure of effect which can be either above or below the true value

Question 21

What is the key difference between bias and confounding?

Answer 21

You can correct for confounding in the analysis - you cannot correct for bias

Question 22

What are the two main types of bias?

Answer 22

Selection bias

Information bias

Question 23

What is selection bias?

Answer 23

Arises from defects in study design relating to how people are chosen to be, or end up in the study

Bias can be introduced if the criteria for inclusion relate to the exposure of interest

Particular issue in case-control studies

Question 24

What are three types of selection bias?

Answer 24

1. Non-response bias (particular issue with questionnaire studies)
2. Loss to follow-up bias (problem in cohort studies)
3. Healthy worker effect (issue with occupational exposures - workers generally healthier than general population)

Question 25

What is information bias?

Answer 25

Occurs when info is collected incorrectly or inaccurately I.e. When study subjects are misclassified according to their disease status, their exposure status or both

Question 26

Give some examples of reasons why patients may be misclassified?

Answer 26

• Error in instrument
• Subjects give incorrect info
• researcher records info incorrectly

Question 27

What are three common types of information bias?

Answer 27

1. Recall bias
2. Reporting bias
3. Observer (or interviewer) bias

Question 28

When can recall bias be a problem?

Answer 28

In case-control studies when cases may recall their exposure more completely than controls

Question 29

When does reporting bias occur?

Answer 29

When individuals with a disease are more likely to report exposure

Question 30

What is observer bias?

Answer 30

When the interviewer/researcher knows who the diseased and non-diseased people are and collects exposure info differently for them (or vice versa)

Question 31

Dealing with bias requires getting the study design right - how would you achieve this?

Answer 31

• Select an appropriate study population
• Maximise response rates
• Minimise loss to follow up
• Use objective rather than self-reported measures of exposure/outcome
• Follow standardised procedures in data collection
• Blind interviewer