HS300: Study Design II

Why conduct studies?

The valid and precise measurement of the association between an exposure and a disease or between a treatment and an outcome

What can case control studies tell us?

• Better understand the causes of disease by establishing risk factors for a disease

• Or in clinical trial to establish whether a treatment is effective

Logical flow chart

• Who was exposed to the disease and who was not?

• Compare them to those who do not have the risk factor

Sequence in case control

• Initial classification of subjects is Disease Status

• Disease is always determined before Exposure/Risk factor

• Individual Cases (with disease) and Controls (without the disease) are always selected first, at the start of the study

• Exposure is determined later

Case control studies

• Analytical – basic premise: disease does not occur randomly, but rather in describable pattern that reflect an underlying etiology/cause.

• Retrospective Exposure ← Disease

2 groups of subjects

• One group in which everyone has the disease (Cases), and

• Second group in which everyone is free of the disease (Controls)

Characteristics of case-control studies

• Initial classification of subjects by disease status.

• A single point of observation.

• Exposure is determined retrospectively.

Selection of cases

• 2 broad steps:

• Identifying a case operationally – “all true cases have equal opportunity of entering study, and no false cases”

• Avoid misclassification of subjects

Case definitions

• Defining a case/disease – broad(inclusive) versus narrow (restrictive)

• Example…cases – myocardial infarction or arthritis?

• Better to be restrictive → avoid ‘misclassification’

Sources of cases

• Need to define a case conceptually.

• Ideally, identify and enroll all new (incident) cases in a defined population in a specified time period.

• A tumor registry may provide a complete listing of all cases.

• Medical facilities also may be a source of cases, but not always incident cases.

Why incident cases?

• Prevalent disease is a function of disease incidence and disease duration (P ~ I X D)

• Select incident cases because we are interested in factors that lead to developing the disease rather than factors that affect the duration (survival factors) of the disease

• e.g. Diabetes, heart attack

Prevalent vs. incident cases

• Prevalent cases make it difficult to separate

  • characteristics that are causal versus

  • characteristics that are consequential

Case definition is influenced by:

1. Standard diagnostic criteria

2. Stage or Severity of the disease

3. Whether the diagnostic criteria are subjective or objective

• Example: ADHD versus breast cancer

• Concern is: Misclassification (see below)

Goal for selection of cases

• ensure that all true cases have an equal opportunity of entering the study and that no false cases enter

• Ideally new cases – but not always possible – use prevalent cases

Controls

• A sample of the population that produced the cases, but they do not have the disease

• i.e. same base population for cases and controls

Selection of controls

• Controls should be the same as the cases in every respect except the outcome (disease) under study

• Controls must have the potential to become case

Selection of controls example

• How many control subjects? => statistical power

  • Case : Control ratio

  • 1:1 -> 1:4 (statistical power max at 4:1)

• There is no statistical advantage in exceeding 4:1 ratio

Sources of controls

1. Population

2. Hospital/Clinic

3. Family/Friend

Population controls

• Population-based controls (e.g drivers license) – same community

• Advantages:

  • same source population

• Disadvantages:

  • time & $$$

  • low participation rates

Relatives or associates based controls

• Relatives of cases?? → ???

• Friends of cases → ???

• Advantages

  • Complex SES variable (income, race, education)

  • First degree relative—genetic

Hospital based controls

• Patients from the same hospital as the cases

• Who can be included?

  • Decide which disease events among controls are appropriate to allow as controls

Hospital based controls rules

1. Illness of control must be unrelated to the exposure under study

2. Controls should have the potential to become a case

Hospital based controls advantages and disadvantages

• Advantages:

  • cheap, quick

  • same source population

  • good recall

• Disadvantages:

  • sick

Measures of association

Strength of association between the exposure and the disease (outcome)

Is there an association between exposure and a specific disease? Is the exposure disproportionately distributed between the cases and the controls?

The exposure may be a

1. risk factor for the disease …or

2. not associated …or a

3. protective factor

Case Control / Odds Ratio

(odds of exposure among disease group (cases))/(odds of exposure among control group)

Case control odds ratio table

Guiding principle of case control studies

to determine how much more likely or less likely the cases are to be exposed compared with the controls

Expression of the odds ratio

Start your statement with “the odds of…” and “Subjects with disease X”

Interpreting an odds ratio

• O.R. = 1:

  • The odds of exposure are equal among cases and controls

• O.R. = 2:

  • The cases were twice as likely as controls to be exposed

• Not all risk factors increase risk.

• O.R. = 0.5:

  • The cases were half as likely as controls to be exposed (lower risk, protective factor).

Caution in interpretation of odds ratio

• Retrospective

• Only one period of observation

• Not a direct determination of risk because of the way the groups are assembled. Groups do not represent population exposed and non exposed to the risk factor (no appropriate denominator for population at risk).

An OR provides a good approximation of risk when

• Controls are representative of a target population.

• Cases are representative of all cases.

Strengths: case control studies

1. Small study size

2. Quick, Inexpensive

3. Most efficient design to study etiology of rare diseases

4. Study diseases with long induction period

5. Evaluate multiple exposures - useful when little is known about the disease

Limitations of case control studies

1. Provide indirect estimate of risk.

2. Timing of exposure-disease relationship difficult to determine.

3. Subject to misclassifaction of exposure data inaccuracy because retrospective (misclassification)

4. Greater chance of bias than cohort study (because
   retrospective)

5. Not efficient for rare exposures (needs large # cases and controls to detect rare exposures)