Dr.

P K Bardhan

• Case-control studies are epidemiological investigations

designed to assess the association between disease
occurrence and an exposure suspected of causing (or
preventing) that disease.

• The key feature of a case-control study is selection of

subjects based upon their disease/outcome status.
The case-control study design has some distinct
properties:
• The study proceeds backwards from effect to cause.

• Both exposure (causal factor) and outcome (disease)

have occurred before the start of the study.

• A control or comparison group is employed to

support or refute an inference.
 Onset of Time
study
Exposed
Cases
(disease)
Not exposed

Exposed
Controls
(no disease)
Not exposed

Direction of Enquiry
https://untobaccocontrol.org/kh/smokeless-tobacco/wp-
content/uploads/sites/6/2017/09/Party-experience-Bangladesh.pdf
 Onset of Time
study
Chewers of
tobacco
Oral
cancer
Non-chewers

Chewers of
tobacco
Free of
Oral cancer
Non-chewers

Direction of Enquiry
 All
All Cases Controls

Cases Controls
(Sample) (Sample)
Selection of Cases.

The case identification should be complete, and the

source population (the population from which the cases
arise) should be well defined.

The cases may be sampled from:

• All patients with the disease of interest within a

particular geographical area.
 • From all cases that occur or are recorded in a health
organization. These cases may be identified by -
 reviewing hospital records, other medical records,
or death certificates,
 through institutional or population-based disease
registries,
 by means of surveillance systems or a survey
program.
Complete identification of cases in a well-defined source
population may be too time-consuming, not feasible, or not
possible.
A common alternative is to use a “convenience sample”.

Cases may be sampled from:

• All cases admitted to or discharged from particular
hospitals or establishments.
Disadvantage of Convenience sampling:

Although it is often easier to identify such cases, the

underlying source population may not be well defined,
making it difficult to generalize the results.

➢ Highly vulnerable to selection bias

➢ Generalisability is unclear

➢ High level of sampling error

Specify the Case Definition.

• The inclusion criteria should be both sensitive (minimize

the likelihood of missing true cases),
and
• specific (minimize the likelihood of including false cases).

 Usually a balance is made.

Selection of Controls.
Identify and select control group of subjects without
the diseases of interest.
In practice, the sources for selecting the controls may be:

Population-based study:
• Ideally, a probability sample from the source population,
if the cases are drawn from the source population,
contacted by telephone, by visiting residences, by
mailing letters, etc.
Hospital-based study:

• a sample of patients admitted to or attending the same

hospital or health facility as the cases;
• disease registries, representing individuals with similar
diseases;
• death certificates, representing persons who died of
other diseases;
• a sample of neighbors, relatives, friends, or associates.
The subjects’ exposure history may be determined by:

• The most objective means of characterizing exposure is

through the use of a biologic marker – in blood or other
specimens.
Difficulties inherent in the use of bio-markers:
 obtaining the specimens may involve an invasive
procedure, discouraging participation;
 many exposures do not have known bio-markers;
 the marker may be transient, and absent when the
measurement is taken.
• Interviews and questionnaires. This is subject to
reporting bias.

• From medical, occupational, or other records.

These methods of obtaining information are not
based upon self-reporting, and thus should avoid
reporting bias.
Selection Bias.

Bias is a systematic error in a study that distorts the results

and limits the validity of conclusions. This arises from the
way in which subjects are selected.

Selection bias can occur in several ways:

• Preferential diagnosis of exposed cases may lead to
selection bias.
• Low participation may lead to selection bias.
Problems of Recall
A major problem of case-control studies is that of recall.
Recall problems are of two types –
 Limitations in Recall: All human beings have limitations
of varying degrees in their ability to search memories
and recall information.
 Recall Bias: If a group of subjects in a case-control study has a higher
or a lower recall ability regarding exposure information in a systematic
way, then this will introduce a bias in the study, and may result in an
erroneous conclusion.

or “b”
Phocomelia & Thalidomide
Matching

Matching is a popular approach to control confounding

in case-control studies. Matching cases and controls
forces these groups to be similar with respect to
important risk factors and thereby makes case-control
comparisons less subject to confounding.
 Matching means that controls are selected which
have certain characteristics in common.
 The characteristics or variables are those that
would confound the effect of the putative risk
factor(s) of interest.
Analysis of Case-Control Studies
The analysis of case-control studies depends upon the type
of sampling – the unmatched or the matched approach.
Unmatched Design: The way the data is obtained and the
data thus obtained can be summarized as:

Cases Controls

Exposed A B

Not-exposed C D
 Design of Case-Control Studies

Odds of Exposure a/c b/d

Odds Ratio (OR) (a/c) / (b/c) = ad/bc
 Cases (Oral Cancer): 100, Controls: 100

Oral cancer
Cases Controls

Exposed 70 40
Tobacco
Chewing
Not-exposed 30 60

OR = (70 X 60) / (30 X 40) = 3.5

 Case-Control Study: Obese persons are at risk for Diabetes
Risk Factor: Obesity Disease: Diabetes

CASES CONTROLS
Diabetics Non-Diabetics
100 100

Obese Non-Obese Obese Non-Obese

35 65 15 85
 Diabetes No Diabetes
Total Diabetics = 100
Obese 35 15 Total Obese = 50

Non-Obese 65 85

Odds Ratio = (35 x 85) / (15 x 65) = 2975/975 = 3.05

A Hypothetical Example of a Case-Control Study of
CHD and Cigarette Smoking
*

Odds Ratio (OR) (112 x 224)/(176 x 88) = 1.62

*CHD: Coronary Heart Disease

Matched Case-Control Studies:
The analysis has to take account of the matched
sampling scheme in matched case-control studies. With
one matched control per case, the data can be
summarized as:
Controls
Exposed Not exposed
Cases
Exposed A B

Not exposed C D
The results have to be analyzed in pairs, since each
entry represents not one subject but two (a matched
case-control pair).
Each case-control pair can be categorized into one of
the 4 basic categories:
A: Both case and control exposed (concordant)
B: Case exposed, control not exposed (discordant)
C: Case not exposed, control exposed (discordant)
D: Both case and control not exposed (concordant)

OR = B / C
1 Case: Oral cancer patient
1 Control (matched for age & sex) } Pair
100 pairs recruited

Control
Zarda + Zarda -
Oral Cancer+
Zarda + 25 50

Zarda - 20 5

OR = 50 / 20 = 2.5
Diabetes & Obesity
Case-Control Study
 A Study on Risk Factors of Type-2 Diabetes Mellitus – A Case Control Study
Indian Journal of Public Health Research & Development. 2013, 4(1):236-240

Vidya.G.S.,
A H Suryakantha.
J J M Medical College.

Objectives: Bapuji Hosp CG Hosp

• To identify the risk factors of Type2 diabetes mellitus.

• To study the association of risk factors (host factors, socio-economic factors and
environmental factors) with the occurrence of Type2 DM.
• To find out the strength of the association between the risk factors and Type2 DM.

Design: Hospital based case control study carried out over a period of one year from
November 2010 to October 2011 in 2 hospitals in Davanagere, Karnataka.
SUBJECTS & METHOD
Cases: >15 yrs of age, both sexes, confirmed by Oral Glucose Tolerance Test,
admitted in the medical wards (n = 406).
Controls: Age- and sex matched non-diabetics taken from in-patients of the same
hospitals, from all wards except medical wards, admitted on the same day as
is the case (n = 406).
Exclusion: Type1DM, diabetics <15yrs, diabetics under treatment & gestational diabetics.

● A predesigned, pretested, semi structured interview schedule was used.

● Self reported data regarding age, sex, domicile, literacy level, occupation, monthly
income, family H/O type 2DM, diet, alcoholism, life style and exercise.

● Among female subjects information about intake of oral pills was collected.

● Anthropometric measurements.
Variables:
● Literacy level : Illiterate, School, & College.
● Occupation: Professional, Managerial, Clerical, Skilled, Unskilled & Unemployed.
● H/O type 2 DM among 1sto relatives : Absent, One parent & both parents.
● Type of diet: vegetarian or non-vegetarian.
● H/O alcohol intake: Absent, Occasionally & frequently.
● Life style: Sedentary, Moderate or Heavy work depending upon occupation.
● Exercise: Nil, Moderate or Heavy.
● Obesity(by BMI): underweight <18.5, normal 18.5-24.9, pre-obese 25-29.9, obese >30.
● Among female subjects H/O use of OCPs: Absent or present.

DATA ANALYSIS
Statistical tests: Proportion, Chisquare test (x2), Odds ratio (OR)
 (Sociodemographic profile)
Age
RESULTS
● Majority of cases belonged to >45yrs age group (77.6%).

● Males constituted 71% among cases.

● Most of cases were rural inhabitants (63.8%).

● About 18.7% were illiterates.

● Although there was no significant association between occupation of

cases & controls, majority of cases were unemployed(32.8%) as
compared to controls(28.6%).

● About 46% of cases belonged to lower (class IV & class V) SES.

However there was no significant association between SES &
diabetes.
 (Distribution among cases & controls)

**

**

**
 Odds Ratio
● Positive family history (OR 1.6),

● Alcohol intake (OR 1.5),

● Sedentary lifestyle (OR 2.8),

● Lack of physical exercise (OR 2.8)

● Obesity (OR 2.5)

- were significantly associated with diabetes.

 Oral contraceptive pills use among female subjects

(%)

OCP use among women were not found to be significantly associated with diabetes
 Conclusion

Although family history is a non-modifiable

risk factor, the disease can be prevented or
lessened by these modifiable factors:
- abstinence from alcohol,

- increasing physical activity,

- regular physical exercise, and -
- maintaining normal BMI.
● Logistic Regression: Bivariate, Multivariate, Conditional
● Multiple Linear Regression