The document discusses different study designs including observational and experimental studies. It focuses on two observational study designs: cross-sectional studies and case-control studies.
1) Cross-sectional studies examine the relationship between diseases and exposures as they exist in a population at a single point in time. They provide estimates of prevalence but not causation.
2) Case-control studies select participants based on disease status, with cases having the disease and controls not having it. Exposures are then assessed retrospectively. The odds ratio is calculated to measure association rather than relative risk. Case-control studies are well-suited for rare diseases.
6. Cross-sectional study
• “Examines the relationship between diseases and
other variables of interest as they exist in a defined
population at one particular time.”
• Selection is independent of exposure or disease
status.
• Done at a single point in time
• Current disease status is examined in relation to
current exposure status
• Carried out for public health planning and for
etiologic research
6
7. CROSS-SECTIONAL STUDIES
• exposure and disease are assessed simultaneous
in each INDIVIDUAL at a given point or SNAPSHOT
in time...
ONE SLICE IN TIME
7
8. CROSS-SECTIONAL STUDIES
• Because unit of analysis is INDIVIDUALS…
– Provides good estimate of Prevalence
– NOT useful for rare events because in a
one-time snapshot enough rare diseases
or rare exposures may not be captured (in
ecologic studies samples sizes are always
large!)
– May be subject to selection bias if certain
INDIVIDUALS refuse to participate in the
study for unknown reasons
8
9. Key features of cross-sectional studies
• Examine association at a single point in time, and
so measure exposure prevalence in relation to
disease prevalence
• Cannot infer temporal sequence between exposure
and disease if exposure is a characteristic
• Other limitations may include preponderance of
prevalent cases of long duration and “healthy
worker survivor effect”
• Advantages include generalizability and low cost 9
11. CASE CONTROL STUDIES
SOME KEY POINTS
• Most frequently used study design
• Participants selected on the basis of whether or
not they are DISEASED (remember in a cohort study
participants are selected based on exposure status)
• Those who are diseased are called CASES
• Those who are not diseased are called 11
CONTROLS
13. Study Population
DISEASED non-DISEASED
(Cases) (Controls)
exposed non-exposed exposed non-
exposed
13
14. Because participants are selected on the
basis of disease, exposures for ALL
PARTICIPANTS are obtained
RETROSPECTIVELY…
PAST PRESENT
Exposure recall Cases & Controls
Selected
Example: lung cancer cases and non- 14
cancerous controls recall past exposure to
15. SELECTION OF CASES
• FIRST decide on a specific case definition
based on a medically diagnosed condition
• Must consider what criteria will confirm the
case definition
– lung cancer confirmed by biopsy
– osteoporosis confirmed by bone density
measurements
– atherosclerosis confirmed by ultrasound of
carotid arteries
15
16. SELECTION OF CASES
• SECOND will you use INCIDENT or
PREVALENT cases?
• Incident…
– must wait for new cases to occur
– study can specifically measure exposure
relating to development of disease
• Prevalent...
– don’t have to wait while cases occur with
time - more practical!
– study will specifically measure exposure
relating to survival with disease 16
17. SELECTION OF CASES
• THIRD be aware of the unique qualities of
certain groups
– hospital admissions
– nursing homes
– screening participants
– day care facilities
• some groups may have better supporting
medical records
• some groups may be more homogenous and
present less confounding variables
17
18. SELECTION OF CONTROLS
• THE BIG PICTURE…
– Controls should be representative of the
referent population from which cases are
selected (I.e. comparable)
– They don’t have to be representative of the
source (I.e. total) population
– Controls should have the potential to
become cases (they have to be susceptible
for the disease of interest) 18
19. Total Population
Reference
Population
Cases Controls
Controls should be comparable to cases
19
20. Selection of controls
• Controls (referent group) are a sample of
the population that produced the cases
• Controls come from the same base
population as the cases
• Controls must be sampled independently of
exposure status
20
21. Sources of controls
1. Individuals from the general population
Advantage:
Controls would be comparable to the cases w.r.t.
demographic variables
Disadvantage:
1. Time consuming and expensive to identify
2. Interest in participation
3. Recall bias
21
22. 2. Individuals attending a hospital or clinic
• Illnesses of the controls should be unrelated to the exposure
under study
• Control’s illness should have the same referral pattern to the
health care facility as the case’s illness.
Advantages:
1. Less expensive
2. Easy to identify, good participation rates
3. Have comparable characteristics to cases
4. Recall of controls is similar to recall of cases
Disadvantage:
Difficulty in determining appropriate illnesses for inclusion
22
23. 3. Friends or relatives identified by the
cases
Advantage:
• Share the cases’ socioeconomic status, race ,
age, education etc..
Disadvantage:
1. Cases may be unwilling to nominate controls
2. Bias results if cases and controls share a
specific activity (exposure) 23
24. 4. Individuals who have died
• Deceased controls are used when some or all of cases are
dead.
• Identified by reviewing death records of individuals who
lived in the same geographic region and died during the
same time period as the cases.
• Used to ensure comparable data collection procedures
(proxy interviews)
Disadvantages:
• May not be representative of the source population that
produced the cases
• More likely to have used alcohol, drugs, smoking
24
25. Sources of Exposure Information
• Questionnaires
• In-person
• Telephone interviews
• Self-administered questionnaires
• Medical records
Accuracy of the source especially that exposure is
retrospective
25
26. MEASURING ASSOCIATION
• because study participants in Case Control
studies are selected based on disease
status...
– case control studies are ideal for the study
of rare diseases
– incidence can’t be calculated
26
27. MEASURING ASSOCIATION
• Because incidence can’t be calculated, a
relative risk can’t be calculated (RR is a ratio
of INCIDENCE in exposed and non-exposed)
• Instead of the RR, an ODDS RATIO is
calculated in case control studies
27
28. Analysis of case-Control studies
• The size of the population which produced
the cases is not known
• Can not calculate risk
• => calculate Odds
• The odds of an event is the probability that
it will occur divided by the probability that
it will not occur Odds among the exposed
• Disease odds ratio = Odds among the non-exposed
28
29. MEASURING ASSOCIATION
• Odds: NOT a proportion, but the ratio of the
# ways an event CAN occur relative to the #
of ways an event CAN NOT occur
Odds = P(event occurs) = p / ( 1 - p)
1 - P(event occurs)
• Odds Ratio: Odds of case being exposed
Odds of control being exposed
29
30. Cases Controls
Exposed a b
Unexposed c d
a/b ad
Odds ratio= =
bc
c/d
30
31. Is Use of Artificial Sweeteners associated
with Bladder Cancer?
Cases Controls
Ever Used 1,293 2,455
Never Used 1,707 3,321
Total 3,000 5,776
ODDS RATIO = 1,293 * 3,321 = 1.026
2,455 * 1,707
31
Hoover and Strasser (1980) Lancet 1: 837-840
32. Interpretation of the Odds Ratio…
If
O.R. = 1 then exposure NOT related to
disease
OR >1 then exposure POSITIVELY related
to disease
OR <1 then exposure NEGATIVELY related
to disease
Hoover and Strasser concluded what from their32
33. CASE CONTROL STUDY SUMMARY
• cases and controls are
representative of a referent
population
cases • controls have the potential
to become cases
Referent pop’n
• selection based on disease
and exposure assessed
Total population retrospectively
33
35. SELECTION OF CONTROLS
• The investigator can elect to use more than
one TYPE of control for each case… when
there is no ONE group similar enough to
cases
EXAMPLE: a particular leukemia case
may have both a neighborhood control
(similar to case in terms of environment)
and a sibling control (similar to case in
terms of genetic background)
35
36. SELECTION OF CONTROLS
• to avoid potential problems of confounding
some studies use MATCHING
– MATCHING: the process of selecting
controls so that they are similar to cases
on certain specific characteristics
36
37. Exposure Disease
Confounder
Confounders are third variables that are
associated with both the disease and the
exposure
37
38. SELECTION OF CONTROLS
• CHARACTERISTICS THAT ARE OFTEN
USED FOR MATCHING…
– age
– gender
– body mass index (weight / height2)
– smoking status
– marital status
38
39. BIAS IN CASE CONTROL STUDIES
BIAS: any systematic error (not random or by
chance) in a study which leads to an incorrect
estimate of the association between an
exposure and the disease of interest
• MAIN TYPES of bias in Case Control
Studies…
– selection bias
– recall bias
39
40. BIAS IN CASE CONTROL STUDIES
• SELECTION BIAS: systematic error due to
differences in characteristics between those
selected for a study and those not selected
EXAMPLE in CC Studies: When cases are
selected from a hospitalized population with
unique exposures, controls often are not
representative of the population that gave rise
to cases
40
41. BIAS IN CASE CONTROL STUDIES
• RECALL BIAS:systematic error due to
differences in accuracy or completeness of
recall to memory of past events or
experiences
EXAMPLE in CC Studies: Often cases faced
with a serious illness will more closely
scrutinize their past exposures and will be
more accurate and complete in their recall
than controls
41
42. What do you think will happen to our
estimate of the Odds ratio if cases recall
their exposure status better than controls?
D ND
Exposure a b
No exposure c d
Odds Ratio = ( a ) d / b c
42
43. Strengths of case-control studies
• Efficient for rare diseases
• Efficient for diseases with long induction
and latent periods
• Can evaluate multiple exposures in relation
to a disease
• can use smaller sample sizes
• cost/time effective when using previously
collected (RETROSPECTIVE) exposures
43
44. Weaknesses
• Inefficient for rare exposures
• May have poor information on exposures because
retrospective
• Vulnerable to bias because retrospective
• Difficult to infer temporal relationship between
exposure and disease
• can’t calculate incidence
• selecting appropriate controls can be challenging
44
45. When is it desirable to use case-
control design?
• Exposure data are difficult or expensive to
obtain
• The disease is rare
• The disease has a long latent period
• Little is known about the disease
• The underlying population is dynamic
45
Notes de l'éditeur
Sharon - the idea here is like the relation between oral contraceptives and thromboembolism. Because physicians were more likely to hospitalize women presenting with symptoms of thromboembolism if they currently used oral contraceptives, any case control investigation of the relationship between this exposure and disease that utilized only hospitalized cases could overestimate the true relationship since the proportion of exposed women would be artificially high among those hospitalized with thromboembolism compared with all cases of thromboembolism.
Sharon - the idea here is like the relation between oral contraceptives and thromboembolism. Because physicians were more likely to hospitalize women presenting with symptoms of thromboembolism if they currently used oral contraceptives, any case control investigation of the relationship between this exposure and disease that utilized only hospitalized cases could overestimate the true relationship since the proportion of exposed women would be artificially high among those hospitalized with thromboembolism compared with all cases of thromboembolism.
Sharon - the idea here is like the relation between oral contraceptives and thromboembolism. Because physicians were more likely to hospitalize women presenting with symptoms of thromboembolism if they currently used oral contraceptives, any case control investigation of the relationship between this exposure and disease that utilized only hospitalized cases could overestimate the true relationship since the proportion of exposed women would be artificially high among those hospitalized with thromboembolism compared with all cases of thromboembolism.
Sharon - the idea here is like the relation between oral contraceptives and thromboembolism. Because physicians were more likely to hospitalize women presenting with symptoms of thromboembolism if they currently used oral contraceptives, any case control investigation of the relationship between this exposure and disease that utilized only hospitalized cases could overestimate the true relationship since the proportion of exposed women would be artificially high among those hospitalized with thromboembolism compared with all cases of thromboembolism.
Sharon - the idea here is like the relation between oral contraceptives and thromboembolism. Because physicians were more likely to hospitalize women presenting with symptoms of thromboembolism if they currently used oral contraceptives, any case control investigation of the relationship between this exposure and disease that utilized only hospitalized cases could overestimate the true relationship since the proportion of exposed women would be artificially high among those hospitalized with thromboembolism compared with all cases of thromboembolism.
Sharon - the idea here is like the relation between oral contraceptives and thromboembolism. Because physicians were more likely to hospitalize women presenting with symptoms of thromboembolism if they currently used oral contraceptives, any case control investigation of the relationship between this exposure and disease that utilized only hospitalized cases could overestimate the true relationship since the proportion of exposed women would be artificially high among those hospitalized with thromboembolism compared with all cases of thromboembolism.
Sharon - the idea here is like the relation between oral contraceptives and thromboembolism. Because physicians were more likely to hospitalize women presenting with symptoms of thromboembolism if they currently used oral contraceptives, any case control investigation of the relationship between this exposure and disease that utilized only hospitalized cases could overestimate the true relationship since the proportion of exposed women would be artificially high among those hospitalized with thromboembolism compared with all cases of thromboembolism.