Observational Studies in CVD

Question 1

What is a cross-sectional study?

Answer 1

They are a sample of population that are selected and about whom information is obtained at one point/period in time. Large studies can take place over years but each subject only contributes data once, thus, there is no follow up of subjects.

Question 2

How is data collected in a C-S study?

Answer 2

Data is collected via:
	- questionnaires
	- examinations
	- investigations
The information is mostly descriptive in nature and especially often includes prevalence. (ie CVD among Australians)

Question 3

What are the benefits of C-S studies?

Answer 3

C-S Studies are relatively cheap and easy and are often needed for a representative sample. They can explore associations among variables but no explicit data is obtained on temporal relationships, thus they are weak evidence for causality.

Question 4

Give an example of a C-S study?

Answer 4

Example: Australian diabetes, obesity and lifestyle study (began as a cohort study but is now considered a C-S study)

Question 5

What is a Case Control Study?

Answer 5

They are comparisons of previous exposure status between subjects with the outcome of interest (case) and subjects without the outcome of interest (controls). Controls and cases are matched 1:1, and confounders such as age and gender are often matched as well.

Question 6

What are the C-C study processing steps?

Answer 6

C-C Studies require some processing steps:

Step 1: Define and recruit cases; recruit controls by matching to cases (outcome ascertainment 1st)
Step 2: Determine previous exposure among subjects.

Question 7

Explain the principles surrounding odds with respect to C-C studies.

Answer 7

In these studies, we know that the exposure preceded the outcome in the end. This is because they work backwards, finding those with the outcome and looking for the exposure as a consequence.

In C-C Studies, explicit knowledge is obtained about temporal relationship between exposure and outcome. The are useful for studying rare outcomes. The key output of these studies is odds ratio and approximation of relative risk of outcome conferred by exposure. Because the C-C studies are not longitudinal, the RR can ony be approximated; this is reffered to as the odds ratio.

Question 8

What is a Cohort Study?

Answer 8

These are longitudinal studies with follow up of subjects in which incidence data is collected and a comparison is made regarding the outcomes between/among subgroups. (ie. those exposed vs not exposed to risk factor) From these studies relative risks can be derived.

Question 9

What is the difference between a prospective cohort study and a retrospective cohort study?

Answer 9

A prospective cohort study has the key target to gain explicit knowledge about the temporal relationship between exposure and outcome.

A retrospective cohort study shares this target despite the difference in approach.
The researcher has come in after the outcome has already occurred. The follow up is still taking place however it is followed up that which has already occurred. Thus the cohort study is maintained.

C Studies are explicit (often-detailed) in their knowledge about temporal relationship between exposure and outcome and can include multiple exposures and outcomes. Research Hypotheses can be addressed post hoc in established cohorts. (Other hypotheses can be addressed after the study ahs taken place; it just becomes a retrospective cohort study.)

Question 10

What are some disadvantages of Cohort studies?

Answer 10

Rare outcomes are difficult to study using this type of study and the method is neither cheap nor easy. But cohorts can be established as part of routine clinical care (ie. RMH stroke service) and this can be a very valuable resource.

Question 11

Give an example of a cohort study.

Answer 11

Example: Framingham Heart Study looking for incidence of CVD and stroke. This study has characterised the outcomes of CHD and stroke with certain risk factors. (Smoking, BP, male gender etc.)

Question 12

What is outcome ascertainment?

Answer 12

Outcomes are captured by ‘active’ and/or ‘passive’ follow up.

Active: explicit surveillance for outcomes, at ‘coal-face’
Passive: database linkage, mostly retrospective

There is a need for clear, objective definition criteria.

Question 13

What is Bias?

Answer 13

A bias is an error (unintentional) that leads to a systematic difference between/among groups which leads to under/overestimation of true results.

There are two main types of bias:

- Selection bias
- Information (measurement) bias

Question 14

Explain selection bias.

Answer 14

1. Systematic difference in characteristics of people selected for study and those not selected - specifically, people whose data were used for analysis and people whose data were not.

Example: the ‘worried well’ - the people who respond and partake in the studies are not always representative of the type of subject population involved in the outcome. (Perhaps only those worried about diabetes will partake in the diabetes study despite not holding risk factors or past exposure)

The observed result may not reflect the true situation, and/or may not be generalisable.

2. Systematic difference(s) in characteristics of subjects within groups being compared; these differences are partly responsible for the observed study results.

Example: systematic differences between cases and controls in the INTERHEART study OR systematic differences between subjects who drop out and don’t drop out in the F’ham study. (The people who stay in one area for 50 years are going to be very different from those who dropped out and left the region.)

Question 15

How is selection bias minimised?

Answer 15

• Careful recruitment (representative sample, cases and controls from the same source)
• maximise response
• minimise lost to follow-up

Question 16

What is information bias?

Answer 16

Systemic difference(s) in the way information is collected between among groups being compared. Differences are partly responsible for the observed study results. This type of bias arises when there is variability (especially subjectivity) in methods for collecting information.

Example: In the INTERHART study, differential recall of smoking status between subjects with MI (cases) and subjects without (controls) - recall bias (People are likely to remember that they smoked if they have CVD because there is a known association)

Differential measurement of risk factors between the elderly and non-elderly in the F’ham study OR Differential measurement of CVD (outcome) between males and females in the F’ham study. (Women appear to have more disease however they are more likely to attend doctors appointments and so their disease is more frequently seen than men)

Question 17

How is information bias minimised?

Answer 17

The key focus should be to ensure that methods for collecting information are uniform between/among groups being compared.

Examples:

- standardised tools
- objective (vs. subjective) assessment

Question 18

What is confounding?

Answer 18

Confounding is a specific mechanism that may cause bias.

Some universal confounders include:

1. Age
   - association between use of blue hair rinse (exposure) and bowel cancer (outcome)
2. Sex
   - association with baldness (exposure) and CVD (outcome)

Question 19

How do you minimise confounding?

Answer 19

In the design and execution stages of a study:
- matching by confounder
- restriction
In the analysis stage of a study:
- restriction
- stratification (analysis by sub-group)
- multivariate analyses

Observational studies are far more susceptible to confounding and bias than interventional studies and clinical trials.