2. Introduction : need and the design of case control
study
Basic steps
Estimating the sample size
To describe measure of association in case control
study and its interpretation
To discuss potential biases in case control study
2
5. Definition:
“The observational epidemiological study of persons with the
disease (or other outcome variable) of interest and a suitable
control(comparison / reference) group of persons without the
disease”
(Dictionary of epidemiology; 4th edition, John M Last)
Subtype of analytical observational study
Also called as - Retrospective study design
- TROHOC study etc
5
7. Observational, non experimental
Explanatory (analytical)
Both exposure and outcome have occurred before the start of
the study
Proceeds backward from outcome(effect) to exposure (cause)
i.e Retrospective
It uses a control or a comparison group to support or refute an
inference
7
8. Disease Perspective
The disease is rare
The disease has a long
incubation and latent period
Data on exposure is
meager or expensive to
obtain
The study population is
dynamic
The risk factors for the
disease is not known
Researcher perspective
Limited resources in terms
of material and time.
Investigating multiple risk
factors
8
9. Research question- based on PECO
Methods – objectives and sample size
Selection of cases and controls
Matching
Ascertainment of exposure
Analysis and interpretation
9
11. Clear case definition for accurate classification of diseased
and non-diseased
Cases comprise all (or a representative sample of) members
of a defined population who develop a given health outcome
during a given period of time.
Efficient and accurate sources should be used to identify cases
Incident cases are preferable to prevalent cases
Sources of cases:
Hospital, disease registries, Special surveys etc .
11
12. Most important step
The guiding principle for valid selection of cases and
controls is that they represent the same base population
Controls must be sampled independently of exposure
status
Sources
Population controls
Hospital or clinic controls
Dead controls
Friend, spouse and relative controls (case nominate)
Neighborhood controls
12
13. At least 1 control for each case.
Can select 2 /3/ 4 … 100 controls per case: BUT
More Controls, More Resources
Do not gain much statistical power if more that 4
Controls per case are enrolled
Multiple controls a) Same type
b) Different type
13
15. Arises when cases and control may not be representative of
cases and control in the general population.
There may be systematic differences in the characteristic of
cases and control
Types:
• Berkson’s Paradox
• Neyman Fallacy
• Selective Referral
• Detection Bias
• Non-Response
• Length of Hospital Stay Bias
• Survival Bias
15
17. A type of selection bias which arises when the samples are not
taken from the general population but from sub population,
eg: hospital.
Hospital samples may exhibit spurious associations between
two variables, even though these variables are independently
distributed in the general population.
eg: medication-diseases associations Laxative use and
arthritic disease
OR (General Population): 1.48
OR (Hospital Sample): 5.00
17
20. A selection bias where the very sick or very well (or both) are
erroneously excluded from a study.
Excluding patients who have died will make conditions look
less severe.
Excluding patients who have recovered will make conditions
look more severe.
also called prevalence-incidence bias, selective survival bias
20
21. If cases within the population are differentially reported to
study hospital.
Tertiary Care Hospital: Complicated/Severe D Which may
differ etiologically from other cases.
21
22. A situation in which an association between an
exposure and an outcome is entirely or partially due
to another exposure which is called confounder
Must be associated with exposure of interest
Must be a risk factor for outcome of interest
Must not be on the causal pathway
Confounding masks the true effect of a risk factor on
a disease or outcome due to the presence of another
variable
22
24. During study design:
Randomisation
Restriction
Matching
During analysis:
Stratification
Standardization
24
25. Cases and controls might differ in the characteristic or exposure
other than the one that is targeted for the study
Purpose: To adjust - effects of relevant confounders
Involves only those variables which improve statistical efficiency or
eliminate bias from the effect of interest.
Matching is of two types:
(a) group matching
(b) individual matching
Unplanned matching
25
26. Matching of variables other than those which we are
sure are not risk factors for the disease of interest
either in a planned manner or inadvertent manner is
called overmatching.
26
27. Standardized methods should be used:
by Interview by review
Questionnaire records, biomarkers
Methods should be applied similarly in both case and control
groups
Selection on particular source depends on availability,
accuracy, logistics and cost
Accuracy is a particular concern in case-control study
27
28. Occurs when the means for obtaining information about the
subject in the study are inadequate so that as a result some of
the information gathered regarding exposure or disease
outcome is incorrect.
Types:
- Interviewers bias
- Surveillance bias
- Recall bias ( main problem in case control study)
- Reporting bias
28
29. When interviewer knows the hypothesis and also
knows who the cases are. Eg: OCP consumption
leads to breast carcinoma.
Check on this type of bias can be made by noting the
time of interview of cases and control.
Eliminated by double blinding.
29
30. A non random type of information bias
Refers to the idea : the more you look, the more you find.
eg: DVT reported after trauma.
REPORTING BIAS
Reporting bias is defined as "selective revealing or
suppression of information" by subjects (for example about
past medical history, smoking etc)
30
31. Also called as memory bias
Eg: A study to find out whether maternal infection
during pregnancy leads to congenital malformation in
the neonate.
Cases can better recall the past history as compared
to controls
Called as rumination bias by “Ernst Wynder”
Eliminated by : prospective studies
31
32. Compare rates of exposure across cases and
controls i.e Proportion exposed rates among cases
and controls
Measure of Associations
- Odds Ratio and confidence interval
32
33. a b
c d
33
Cases [CHD]
(with disease)
Controls[no CHD]
(without disease)
Exposed
[Smokers]
Not Exposed
[Non Smokers]
a + c b + dTotals
Proportions Exposed a / a+c b / b+d
34. Exposure Diseased Non diseased Total
Yes a b a + b
No c d c + d
Total a + c b + d a + b + c + d
34
Odds that cases were exposed = a/c
Odds that controls were exposed = b/d
35. = Odds that cases were exposed
Odds that controls were exposed
= a/c
b/d
= a * d
b * c
Odds ratio is a measure of strength of association.
35
36. OR = 1 Exposure is not related to the
disease
OR >1 Exposure is positively related to the
disease i.e causal to disease
OR <1 Exposure is negatively related to the
disease i.e protective to the disease
36
37. 95% CI value is taken.
An odds ratio of 5.2 with a confidence interval of 3.2 to 7.2
suggests that there is a 95% probability that the true odds
ratio would be likely to lie in the range 3.2-7.2 assuming
there is no bias or confounding.
Odds ratio should ideally be given with limits of confidence
interval.
37
38. Research Question:
Whether there is an association between coffee
drinking and occurence of pancreatic carcinoma??
38
exposure Cases
(pancreatic
carcinoma)
Controls
(healthy
individuals)
total
Coffee drinking
+
90 20 110
Coffee drinking
-
10 80 90
total 100 100 200
39. Research Question:
Whether there is any association between the
development of lung cancer and history of uranium
exposure through mining??
39
Exposure Cases
(lung cancer
pts.)
Controls
(Healthy
individuals)
Total
Uranium mining
+
23 0 23
Uranium mining
-
9 64 73
Total 32 64 96
40. ADVANTAGES
Can obtain findings quickly
Can often be undertaken
with minimal funding
Efficient for rare diseases
Can study multiple
exposures
Generally requires few
study subjects
DISADVANTAGES
Inefficient for rare exposure
Subject to bias
Difficult if record keeping
is either inadequate or
unreliable
Selection of controls can be
difficult
Difficult to ascertain
temporal relationship
40
41. Evaluating Vaccine Effectiveness
Evaluations of Treatment & Program
Efficacy
Evaluation of Screening
Outbreak Investigations
Indirect Estimation in Demography
Genetic Epidemiology
Occupational Health Research
Predictive Modeling
41
42. Study material from workshop on advanced
epidemiology – NIMHANS
Epidemiology – Leon Gordis, 5th edition
Parks textbook of Preventive and Social Medicine,
24th edition
42