Transcript Epidemiologic Measures

Case-Control Studies
Afshin Ostovar
Bushehr University of Medical Sciences
Bushehr, 2011
7/21/2015
1
Epidemiological
Studies
Analytical
Descriptive
7/21/2015
Case Report
Observational
Interventional
Case-series
CrossSectional
Clinical
trials
Ecologic
CaseControl
Field Trials
Crosssectional
Cohort
Community
Trials
2

The sophisticated use and understanding of
case-control studies is the most outstanding
methodologic development of modern
epidemiology. (Rothman 1986, p. 62)

A primary goal is to reach the same conclusions
in a retrospective study as would have been
obtained from a forward study, if one had been
done. (Mantel and Haenszel 1959, p. 722)
7/21/2015
3
Key advantages
7/21/2015

Statistically efficient for rare conditions

Logistically efficient for prolonged induction
or latency diseases

Can examine many exposures in one study

Ethical - cannot affect onset of disease
4
Case-Control Studies

Examines the association between disease and
potential risk factors by taking separate samples
of diseased cases and of controls at risk of
developing disease.

Information may be collected for both cases and
controls on genetic, social, behavioral,
environmental or other determinants of disease
risk.
7/21/2015
5
7/21/2015
6
Odds Ratio (cross-product ratio)
7/21/2015
7

The initial interpretation of the case-control study was the
comparison of exposure histories for a group of diseased cases
with those for non-diseased controls.

The critics argued that such comparisons provided no
information about the quantities of true epidemiologic interest,
namely the disease rates.

Cornfield (1951) corrected this misconception by
demonstrating that the exposure odds ratio for cases vs.
controls was equal to the disease odds ratio for exposed vs.
non-exposed.
7/21/2015
8

Selection of cases and controls

Matching

Case-control studies based on a defined cohort
7/21/2015
9
Selection of cases
7/21/2015

Sources for cases

Problems in case selection

Incident or Prevalent Cases
10
Sources for cases

Hospitals

Physicians— practices, or clinics

Registries
7/21/2015
11
Problems in case selection

If cases are selected from a single hospital, any risk factors
that are identified may be unique to that hospital as a result
of referral patterns or other factors, and the results may not
be generalizable to all patients with the disease.

Furthermore, if the hospital from which the cases are drawn
is a tertiary care facility, which selectively admits severely
ill patients, any risk factors identified in the study may be
risk factors only in persons with severe forms of the
disease.
7/21/2015
12
Incident or Prevalent Cases
7/21/2015

Number of cases available

Survival bias
13
Principles of Control Selection

The “study-base” Principle

The Deconfounding Principle

The Comparable Accuracy Principle
7/21/2015
14
The “study-base” Principle

Controls be randomly selected from disease-free
members of the underlying cohort, also known as
the source population or study-base at the times
that cases are being ascertained

When controls are in fact selected later, it
sometimes mandates the random selection of a
reference date for each control so that the
distributions of the case diagnosis dates and
control reference dates are comparable.
7/21/2015
15
The “study-base” Principle. Cont.
7/21/2015

Only exposures occurring prior to the
reference|diagnosis date would be taken into
account.

This principle also implies that whatever
exclusion criteria have been applied to the
cases must also be applied equally to the
controls.
16
The Deconfounding Principle

7/21/2015
This principle underlies the stratified
sampling of controls to render possible, or
improve the efficiency of, an adjusted
analysis designed to control confounding
17
The Comparable Accuracy Principle

7/21/2015
Controls be selected so that the errors of
measurement of their exposures and covariates
are comparable to the measurement errors of the
cases.
18
Selection of Controls

Sources of Controls
 Non-hospitalized
 Hospitalized
7/21/2015
Persons
Patients
19
Non-hospitalized Persons
 Probability
sample of the total population
 School rosters
 Selective service lists
 Insurance company lists
 Neighborhood controls
 Random-digit dialing
 Best friend control
 Spouse or sibling
7/21/2015
20
Hospitalized Patients as Controls

“captive population” = more economical

Ill-defined reference population

“All other patients admitted to the hospital” Vs
“a specific “other diagnosis”
7/21/2015
21
7/21/2015
22
How Many Controls per Case?

For a fixed number of study subjects,
statistical power for testing the null hypothesis
is optimized by having equal numbers of cases
and controls.

When the disease is extremely rare or
acquisition of cases particularly expensive,
however, it may be important and costeffective to increase the numbers of controls.
7/21/2015
23
How Many Controls per Case?

In order to have the same statistical power (to
reject the null hypothesis of no exposure effect
against local alternatives) as a design with equal
numbers of cases and controls, a design with M
controls per case would need only (M+1)|2M as
many cases.

For a fixed number of cases, the relative efficiency
of a design with M controls per case relative to one
that uses an unlimited number of controls is
therefore only M|(M + 1).
24
7/21/2015
How Many Control Groups?

multiple control groups were recommended by
Dorn (1959) to improve the case-control study so
that it would “provide a more valid basis for
generalization”.
If a whole series of control groups, e.g., of patients
with different diseases, gives much the same
answer and only the one affected group differs, the
evidence is clearly much stronger than if the
affected group differs from merely one other
group.”
7/21/2015

25
How Many Control Groups?

Failure to detect a difference among control groups may
give a false sense of security unless they were
deliberately selected to differ with respect to
unmeasured potential confounders.

Recent reviews of case-control methods have tended to
shy away from the use of multiple control groups. They
argue that there is usually a single “best” control group,
and that since the discovery of an adjusted exposure
difference with other control groups will force these to
be discarded, the effort involved will have been wasted.
7/21/2015
26
Case-Control Types

Primary-based Vs Secondary-based

Population-based Vs Hospital-based
7/21/2015
27
Matching

Group Matching (frequency matching)

Individual Matching (matched pairs)
7/21/2015
28
Problems with matching

Practical Problems

Conceptual Problems
7/21/2015
29
Problem of Recall

Limitations in Recall

Recall Bias
7/21/2015
30
Case-control studies based on
a defined cohort

Nested Case-Control Studies

Case-Cohort Studies
7/21/2015
31
7/21/2015
32
7/21/2015
33
7/21/2015
34
7/21/2015
35
advantages of conducting a casecontrol study in a defined cohort
1.
No recall
2.
Temporality is established
3.
Efficiency
4.
Compariblity
7/21/2015
36
Case-Crossover Design

Primarily used for studying the etiology of acute outcomes
such as myocardial infarctions or deaths from acute events
in situations where the suspected exposure is transient and
its effect occurs over a short time.

The question being asked is: Was there any difference in
exposure between the time period immediately preceding
the outcome and a time period in the more remote past
which was not immediately followed by any adverse
health effect?
7/21/2015
37