Case Control Studies
- Compares cases and controls with respect to their level of exposure to a risk factor of interest
- Calculate the Odds of disease among the exposed/unexposed
- calculate the OR
When is a Case Control Study Desirable?
- When a disease has a long induction and latent period
- Rare/unknown diseases
Inefficient to follow participants over time
Cases in a Case Control Study
Cases = the numerators of the incidence/prevalence of disease in exposed / unexposed groups
Incidence of disease in exposed = a/?
Incidence of disease in unexposed = c/?
Controls in a Case Control Study
Controls = sample of base population that gave rise to the cases
*If it were a cohort study, we would have the total population or person years for both groups
If controls are selected correctly, a similar proportion of controls would have developed the disease if they had been exposed to the same exposure as the cases (counterfactual)
Case Control :: Cohort Study
A case control study is conceptually the same as a cohort study
Cases and controls are expected to a common base population in order for us to compare
Principles of Control Selection:
Base Population Principle
- Cases and controls should be representative of the SAME base population
- Method: choose a random sample of individuals from the same source of the cases, at specified time if applicable
- “Nested case control study” - each time a case is dx’d, controls are randomly selected from the “risk set” = subjects at risk at the time of dx
Sampling from the base population:
- Simple random sampling: controls are selected randomly from the base population. Each eligible individual has the same probability of selection
- Stratified sampling/freq. matching: base pop. is subdivided into strata (i.e. by age),
- More complex sampling schemes: two-stage + cluster sampling plans
Principles of Control Selection:
Deconfounding Principle
Confounding should not be allowed to distort the estimation effect.
Confounders that are measured can be controlled in analysis.
Unknown/unmeasured confounders should have as little variability as possible.
Ex: using siblings as matched controls = less variability for genetic confounders , hopefully less cofounding
Principles of Control Selection:
Comparable Accuracy Principle
- The degree of accuracy in measuring the exposure of interest for the cases should be equivalent to the degree of accuracy for the controls
- Doesn’t eliminate information bias
- Rationale is to reduce differential misclassification bias from differential accuracy in case info + control info
- Misclassification of exposure due to this principle causes non-differential error – typically biases toward lack of association/null
Selection of Control Groups
Population Controls
- Most often used when cases are selected from a defined geographic population
- random digit sampling, residence lists, drivers license records
Advantage:
- PI is assured that controls come from same pop. as the cases
-the distribution of exposures in the controls can be extrapolated to base pop.
Disadvantage:
- time consuming, expensive, hard to contact and recruit, may remember exposures differently than cases (recall bias)
-incomplete case ascertainment
-less motivation - may not cooperate
Selection of Control Groups
Hospital Controls
Used when cases are selected from a hospital population
Ex: Cases identified from admissions to hospital coronary care units, Controls from surgical, ortho, medical units of the same hospital–illness with NO relation to cases
Advantages:
- same selection factors that led cases to hospital led controls to hospital
-easily ID’d and accessible (less expensive)
-more willing to participant than pop. controls
Disadvantages:
-hospital controls are ill = may not accurately represent exposure history in the population that produced the cases
-Hospital catchment areas may be different for different diseases
Friend Controls
Controls selected from a list of friends/associates while case is being interviewed.
-reduces biases due to social class
Relative Controls
Used when genetic factors confound the effect of exposure
match on genetic background, ethnicity
Spouses could be good controls if seeking matching on environmental risk factors
Proxy Respondents and Deceased Controls
Used when subjects are deceased or too sick to answer questions, or for persons with cognitive disorders
Surrogates (spouses, children) generally provide accurate responses for broad categories of exposure info
detailed info less reliable
Strengths of Basic Case Control Studies
- Less expensive, time consuming than other basic designs
- Case-control sampling is efficient for studying rare diseases bc investigator fixes the ratio of cases to controls
- Efficient for studying diseases with long latent periods, since exposure history can be measured retrospectively
Population-Based Case Control Studies
case control studies with a defined cohort
- Special type of case control study that has certain features of a cohort design
- A base population is followed for a given period for detection of all incident cases
- Noncases are randomly sampled to compare the cases to
- In this case base population is sampled directly by the investigator to obtain controls
- Exposure info is obtained from all subjects after cases and controls are identified
Population-Based Case Control Studies
Case Group = consists of all/representative sample of incident cases that occur in the defined cohort over a specified follow-up period.
Control Group = selected either from individuals at risk at the time each case occurs, OR from base population
Nested Case Control Studies
Density Sampling of controls
- Controls are selected longitudinally throughout f/u period, typically matched to cases at the time of dx/identification
- Cases are compared with a subset of the “risk set” = subjects at risk at the time of dx
- Controls may become cases during the study period–in analysis they would be treated as both cases + controls
Density Sampling vs Cumulative Sampling
Density Sampling: the equivalent of matching cases and controls on duration of f/u; permits the use of straightforward statistical analysis
Cumulative Sampling: all controls are selected at the end of the observation period during which cases are id’d
- Density sampling preferred when observation period is long, esp. if the frequency of exposure changes over time.
- w/ cumulative sampling, if exposure frequency changes, cases + controls might differ in distribution even if exposure isn’t a real risk factor
Case-Cohort / Case-Base Sampling
- Controls are selected as a random sample of the base population at baseline
- Controls may subsequently become cases during the study period; these subject would then be treated as both cases and controls in analysis
- Prevalent cases should be excluded from the control group
- A sample of the base pop. at baseline can serve as control group for different sets of cases occurring in the same cohort
- If the baseline cohort sample is representative of the base pop., risk factor distributions and prevalence needed for population attributable risk estimate can be obtained.
Strengths of Population-Based Case-Control Studies
Relative to cohort studies:
- more efficient for studying rare diseases, since exposure data are collected on a fraction of noncases
- sample control group can be used to assess effects of different diseases
Relative to basic case-control studies:
- we can estimate the frequency of disease in the base pop. .
- we are more confident that controls are representative of the base population
- less susceptible to selection bias - controls are selected from the same pop. = exposure status less likely to affect it
Limitation of Case-Control Study within a defined cohort
Relative to cohort studies:
- Exposure information obtained after disease occurrence = more vulnerable to measurement error, esp. if collected retrospectively
- leads to bias in effect estimation
Relative to basic case-control:
- It tends to be more expensive or impractical because of the system needed to id all new cases in the reference pop.
- sometimes impossible
