Title: Hypothesis Testing: The Design and Conduct of Case Control Studies
1Hypothesis Testing The Design and Conduct of
Case Control Studies
- PH 2610
- October 14, 2002
- Lowell E. Sever
2Assumptions
- Disease does NOT occur randomly
- Human disease has identifiable causal and
preventive factors
3Overview of Epidemiology
- Frequency, Distribution, Other Factors
- Hypotheses
- Determinants
4Development of a Conceptual Hypothesis
- Theory/ Conceptual
- Knowledge Hypothesis
- Formulate
- Refine
- Human Studies
- Animal Studies
- Logic
- Hunch
5Development of a Study Hypothesis
- Conceptual Study Hypothesis
- Hypothesis Who
- What
- Where
- When
-
- Natural History of Disease
- Population
- Exposure
6Hypothesis Development
- Objective the overall goal or purpose of the
study generally broad in scope - Study question what the study plans to address
regarding distribution or determinants of
disease also referred to as the research
question the specific aim(s) of the study
7Research Questions/Hypotheses
- Exposure (E) Disease (D) ??
- Is there an association between E D?
- Is the association real, i.e. causal?
8Hypothesis Development
- Formulation of a hypothesis
- derived from descriptive studies (i.e.,
studies that describe patterns of disease
occurrence in terms of frequency and
distribution) - tested in analytic studies
- observational case-control or cohort
- intervention clinical trial or field trial
9Hypothesis Development
- Hypothesis a study question that is specific to
risk or cause of disease - null hypothesis (H0) no relationship between
exposure and disease - alternative hypothesis (HA) assertion that
there is an association
10Hypothesis Testing
- A working hypothesis (research question) is
tested in two stages - 1) carrying out a well-designed study to assess
the risk factor(s) of interest along with
potential factors (confounders) that may also be
associated with development of disease
11Hypothesis Testing
- 2) conducting a statistical test (e.g.,
chi-square, t-test) to quantify significance of
difference - requires making an explicit statement about the
difference or relationship between the two groups
regarding exposure -
12Epidemiologic Study Designs
13Analytical Epidemiology
- Definition and ascertainment of cases and
non-cases - Definition and identification of exposure (risk
factors) - Determination of associations between risk
factors and outcome
14Chief Characteristic of a Case-control Study
- ..it begins with people with the disease (cases)
and compares them to people without the disease
(controls). - Source Gordis, page 142.
15Conducting the Study
- Define study population
- Define CASES CONTROLS
- Inclusion criteria
- Exclusion criteria
- Assess exposure status
16Distinguishing Characteristics
- Individuals with a particular disease or
condition (cases) are selected for comparison
with individuals in whom the disease or condition
is absent (controls). - Looks for an association by comparing history of
exposure between a group of diseased individuals
(cases) and a group of non-diseased individuals
(controls).
17Design of Case Control Study
Time
Direction of inquiry
Cases (diseased)
Exposed
Population
Not exposed
Exposed
Controls (disease free)
Not exposed
18Distinguishing Characteristics
- Tests a hypothesis about an association between
disease and exposure - Looks retrospectively to past exposure (or
current exposure as a surrogate for past
exposure) - Explores data to identify exposures for further
studies
19Case-Control Studies
- The first step in a case-control or retrospective
study is to detect a number of people with the
disease under study the cases. - We then select a number of people similar to the
cases who are free of the disease the controls. - The cases and controls are then investigated to
see if there are differences in the distributions
of risk factors of interest between them.
20Study Population
TOTAL POPULATION
Reference Population
CASES
CONTROLS
-Source Gordis, pg. 145
21Variables
- Exposure (independent variables)
- Data obtained through interviews or records of
past exposures - Data on more than one exposure can be gathered
during a case control study
22Variables
- Disease (dependent variable)
- The disease spectrum should be known.
- The CASE definition should take into account the
stage of disease and be established before data
collection. - Data are obtained from medical records,
interviews, physical examination, laboratory
examinations or testing.
23Study Population
- Selection of cases
- Case definition is very important
- All cases have an equal probability for
selection reduce selection bias - Selection of controls
- Identical to cases in every respect except
disease of interest
24Selection of Controls
- from the same population that gave rise to the
cases - independently of exposure status
- during the same time period this person would
have also been eligible to become a case - -Source Rothman Greenland, pg. 97
25Matching
- Matching is done so that the control has
identical (or, at least, very similar)values of
the confounding variable. - Common matching variable are age and sex.
26Matching Controls to CasesAdvantages
- There is direct control of potential confounders.
- It ensures that adjustment is possible.
- Under certain conditions, matching improves the
efficiency of he investigation.
27Matching Controls to CasesDisadvantages
- Data collection is more complex.
- Data analysis must take account of the matching.
- The effect (on disease) of the matching variable
cannot be estimated. - Adjustment cannot be removed
28Matching Controls to CasesDisadvantages
- There may be overmatching. This is where the
matching has been done incorrectly or
unnecessarily. That is, the matching variable
has some relationship with the risk factor or
disease, but it is either not a true confounder
or is so highly correlated with other matching
variables as to be superfluous.
29Analytical Epidemiology
- Definition and ascertainment of cases and
non-cases - Definition and identification of exposure (risk
factors) - Determination of associations between risk
factors and outcome
30Conducting the study
- Determining exposure status for CASES and
CONTROLS - Be aware of limitations in recall
- Both CASES and CONTROLS may not recall accurately
their past exposures - CASES may have a vested interest in recalling
more accurately (results in Recall bias)
31Research Questions
- Among the CASES, is an attribute more (or less)
prevalent than among the CONTROLS? - Is the mean level of factor Y greater among CASES
than among CONTROLS?
32Research Questions
- What characteristics are more common among those
with disease (CASES) than among those without
disease (CONTROLS)? - Are CASES more likely to have been exposed to X
than CONTROLS?
33Retrospective vs. Prospective
Retrospective (Case-control)
Disease
Present (Cases)
Absent (Controls)
Risk Factor
Present (Exposed)
b
a
Prospective (Cohort)
Absent (Not Exposed)
d
c
34Data Analysis
- Present data on reliability and validity of
measurements - Compare CASES and CONTROLS on factors that may be
related to disease and/or exposure status - Produce the appropriate measure of association
- THE ODDS RATIO
35Case Control Study
DISEASE (OUTCOME)
EXPOSURE
36The Odds Ratio
Disease
Y
N
a
b
ab
How much risk is too much risk?
Y
Exposure
cd
d
c
N
ac
N
bd
Odds of exposure if case a / (ac) / c /
(ac) a/c Odds of exposure if control b /
(bd) / d /(bd) b/d
37THE ODDS RATIO (O.R.)
- Definition
- The ratio of two odds
- In a case control study, it is the ratio of odds
of exposure among the CASES to the odds of
exposure among the CONTROLS. - (Source Last,J. Dictionary of Epidemiology.)
38Measure of Association in a Case-Control Study
Odds Ratio
- Odds of exposed vs non-exposed among cases and
controls
39Interpretation of OR
- Cases have OR times the odds of exposure than
Controls - Cases are OR times more (or less) likely to be
exposed to X than Controls
40Case-control Odds Ratio
OR (70/30) / (300/700) 5.4
OR (30/70) / (700/300) .18
41Data from a Study of Oral Clefts and Agricultural
Chemical Exposure in Iowa
Odds Ratio 18 x 266 2.85
48 x 35
42Neural Tube Defects and Water Disinfection
Byproducts
43Case-Control StudiesAdvantages
- Case-control studies are quicker and cheaper than
prospective cohort studies because there is no
waiting time involved. - Many risk factors can be studied simultaneously.
- Case-control studies are particularly well suited
to investigations of risk factors for rare
diseases.
44Case-Control StudiesAdvantages
- Case-control studies usually require much smaller
sample sizes than do equivalent cohort studies. - Case-control studies are generally able to
evaluate confounding and interaction rather more
precisely for the same overall sample size than
are cohort studies.
45Case-Control StudiesDisadvantages
- Case-control studies often do not involve a time
sequence, and so are limited in their ability to
demonstrate causality. - Being identified as a case might reflect survival
rather than morbidity. - Case-control studies can investigate only one
disease outcome.
46Case-Control StudiesDisadvantages
- Case-control studies cannot provide valid
estimates of risk, and can only provide
approximate estimates of relative risk. - Case-control studies potentially suffer from
exposure recall bias.
47The Case-Control Study
- Retrospective study design
- identifies cases
- selects controls
- determines history of exposure
- Measure of association
- odds ratio
48Hypothesis Testing
- Ho null hypothesis or no real relation between
exposure and disease - Ha alternative hypothesis that asserts exposure
affects disease
49Case-control Study
- Groups defined by outcome
- Search for exposure is retrospective
- Quick and inexpensive
- Good for rare outcomes
50Case-control Study
- Susceptible to bias (selection and information)
- May be confounded by unknown exposures
- Appropriate control group often difficult to find
51Case-control Study
- Should be the first approach to the testing of a
hypothesis - Useful for exploratory study of a variety of
variables (fishing expedition) - looking for clues
52Case-control Study
- May be definitive in describing relationship
between exposure and outcome if disease if very
rare and exposure is quite harmful - Adenocarcinoma of the vagina and DES
- given how rare adenocarcinoma of the vagina is -
numbers - Smoking and lung cancer
- given the latency in lung cancer development -
time
53Problems / Limitations
- Choice of appropriate controls
- Potential bias recall, response, selection,
ascertainment, interviewer - Information on exposure obtained after disease
(temporality?) - Cannot estimate disease frequency