Skip to content

Latest commit

 

History

History
130 lines (114 loc) · 7.84 KB

CaseControl_Design.md

File metadata and controls

130 lines (114 loc) · 7.84 KB

Overview of a case-control study

This retrospective study starts with the final health outcome and then traces back to identify potential causes that could have led to this outcome

Important historical case-control studies

  • 1950's: cigarette smoking and lung cancer
  • 1970's: post-memopausal estrogens and endometrial cancer
  • 1980's: Aspirin and Reyes syndrome; Tampon use and toxic shock syndrome; AIDS and sexual practices
  • 1990's: vaccine effectiveness; diet and cancer

Features of a case-control study

  • An observational study
  • A retrospective study that depends on recalling and recording events that happened before the selection of cases and controls
  • Use odds ratio to measure the association between the exposure and disease

Basic idea

  • We hypothesize that the exposure to a factor increases an individual's risk of developing the disease
  • We select the individuals with the disease as cases, and the individuals without the disease who are comparable to the cases as controls
  • We compare the odds of exposure in cases and controls to derive the odds ratio

Selection of cases

  • Define cases: 1) homogeneity of disease entity; 2) strict diagnostic criteria; 3) prevalent or incident cases
  • Source of cases: hospitals; disease registries; medical practices; general population
  • We typically take all the cases from the given source at the specified time

Selection of controls - selection bias

  • The contols are from the same source population as the cases, such that they would have been identified and included as cases had they developed the disease
  • Define controls: 1) free of the disease in question; 2) at risk of the disease; 3) other eligibility requirements as needed
  • Source of controls: hospitals; medical practices; population rosters; workforce; coomunity organizations; relatives of cases; friends or neighbors of cases
  • We typicall sample among potential controls
  • Who is the best control?
    • If cases are a random sample of all cases in the population, controls should be a random sample of all non-cases in the population sampled at the same time
    • If study cases are not a random sample of all cases, it is unlikely that a random sample of the population of non-cases will make up a good control population; in other words, controls must be selected so as to mirror the same biases that entered into the selection of cases
    • Comparability is more important than representativeness in the selection of controls; the control must be at risk of developing the disease; the control should resemble the case in all respects except for the presence of disease
    • ! Do not use convenient controls; use controls who are well defined

Source of data - exposure - recall bias

  • Interviews with cases and controls
  • Interviews with surrogates (relatives)
  • Medical records
  • Employment records
  • Biologic specimens: genetic characteristics; cholesterol; infection; markers of environmental exposures
  • Determine the accuracy of the exposure information since it is retrospective

Study base

  • It is composed of a population at risk of exposure over a period of risk of exposure
  • Both cases and controls should emerge from the same study base
  • If cases are selected exclusively from hospitalized patients, controls must also be selected from hospitalized patients
  • If cases must have gone through a certain ascertainment process (e.g., mammogram-detected screening), controls must have also
  • If cases must have reached a certain age before they can become cases, so must controls
  • If the exposure of interest is cumulative over time, the controls and cases must each have the same opportunity to be exposed to that exposure

Issues in matching controls

  • Identify the pool from which controls may come. This pool is likely to reflect the way cases and controls were ascertained (e.g., hospital, screening test, telephone survey)
  • Control selection is typically through matching
    • Define matching variables (e.g., age) and criteria (e.g., control must be within the same 5 year age group) in advance
  • Controls can be individually matched or frequency matched
    • Individual matching: select one or more controls who meet the matching criteria for each case
    • Frequency matching: select a population of controls such that the overall characteristics of the group match the overall characteristics of the cases. e.g., if 20% of cases are female, 20% of controls are female, as well.
  • Avoid over-matching: match only on factors known to be causes of the disease
  • Obtain power: match more than one control per case. Generally, the number of controls should be <= 5, since there is no further gain of power above 4 controls per case
  • Obtain generalizability: match more than one type of control

Issues with matching

  • It is difficult to find a good control if there are matching criteria on many variables
  • Once the variables are matched, we cannot explore the association of the disease with these matched variables

Measure the strength of association: odds ratio

Exposure Have disease No disease
Yes A B
No C D
  • Odds of exposure in the cases = A/C
  • Odds of exposure in the controls = B/D
  • Odds ratio (OR) = (A/C)/(B/D) = AD/BC
  • OR > 1: positive association (increased odds); 0 < OR < 1: negative association (decreased odds); OR = 1.0: no association

Interpreting the OR

  • What does the OR of age (1.02) mean in a multivariable logistic regression model? 
    Patients with one-year older are 2% more likely to have the disease, after controlling for other variables.
  • What does the OR of age (0.94) mean in a multivariable logistic regression model? 
    Patients with one-year older are 6% less likely to have the disease, after controlling for other variables.
  • We cannot calculate the incidence rate directly in a case-control study, because we do not have the population at risk. Instead, we calculate the odds of having the exposure in the cases and controls
  • The OR approximates relative risk (RR) when 1) controls are representative of the general population; 2) cases are representative of all diseased; 3) the disease is rare

Advantages of case-control studies

  • Well suited to the study of rare diseases or those with long latency
  • Allows study of multiple potential causes of a disease
  • Exisitng records can occasionally be used
  • Relatively efficient to organize and conduct
  • Relatively inexpensive compared to cohort and clinical trial
  • Relatively short time commitment from respondents compared to cohort and clinical trial
  • Investigators can obtain answers quickly

Disadvantages of case-control studies

  • Limitations in recall: sometimes, validation of exposure information is difficult or impossible
  • Control of confounding variables may not be complete
  • Restricted to a single outcome
  • Difficult to study rare exposures

Compare case-control and cohort studies

  • Case-control study
    • Only estimate the odds ratio
    • Potentially weaker causal investigation
    • Inexpensive
    • Short-term study
    • Can be powerful with small sample of cases
    • Efficient design for rare disease
    • Good for multiple exposures
    • Suffer recall bias
    • Less likely to suffer loss of follow-up
    • The results may not be generalizable
    • Cannot examine the natural course of disease, survival
  • Cohort study
    • Can calculate the incidence rate, risk, and relative risk
    • Potentially greater causal investigation
    • Expensive
    • Long-term study
    • Large sample size required
    • Efficient design for rare exposure
    • Good for multiple outcomes
    • Less likely to suffer recall bias
    • More likely to suffer loss of follow-up
    • The results may be generalizable
    • Can examine the natural course of disease, survival