Epidemiology Revision

Lecture 2

Department of Public Health Medicine

Penang Medical College

At the end of lecture, students should

be
Able to understand epidemiological studies
Know how to select an appropriate
epidemiological study
Know different types of error and bias
Able to classify and measure risk

Types of study design

Can be broadly categorized into descriptive and
analytical study.
Descriptive study: Describes health and disease in
a population, do not test hypothesis. E.g.
Describing disease prevalence, identify high risk
populations, describing the uses of health services
etc.
Examples of descriptive studies are case reports,
case series, ecological study and descriptive
cross-sectional study.

 Analytical study: test the hypothesis. E.g.

determine the association between risk
factors and disease.
Examples of analytical studies are
analytical cross-sectional study, casecontrol and cohort study.

Epidemiological studies in Medical

Research
Randomized Controlled Trials
Cohort
Case Control
Cross sectional
Ecological
Case series
Case reports

Increasing
Hierarchy of
Epidemiologic
Study Design

Selection of study designs

Did investigator
assign exposures?
YES

NO

Experimental studies

Observational studies

Random allocation?

Comparison group?

NO

YES
Randomized
controlled trial

YES

Non-randomized
controlled trial

NO

Analytical

Descriptive

Direction?

Exposure
Cohort
study

Outcome

Exposure

Outcome

Case-control
study

Exposure and outcome

occur at the same time
Cross-sectional
study

Cross-sectional study
Observation at one particular point in time/ over a period:
snapshot of the situation of population at a point of time
POPULATION

SAMPLE

Exposed

Disease

Not exposed

No Disease

Both the exposure and disease is determined

simultaneously.
Example of cross-sectional study:
Acute stress disorder after myocardial infarction: prevalence and associated
factors.

Advantages and disadvantages

Advantages
Very quick and relatively inexpensive
Starting point in prospective cohort and case control
studies
Useful in determining the disease trend

Disadvantages

Can not determine causal effect

Cannot establish temporality
Impractical for rare outcomes and remission
Prone to selection bias, information bias, and
confounding bias.

Case-control study

Onset of study
TIME

EXPOSED

CASES
NOT EXPOSED

EXPOSED

CONTROL
NOT EXPOSED

Direction of inquiry
(Backward directionality)
Compares a case group (with disease) with a control group (not diseased) with
reference to past exposure to possible risk factors.

Example

Onset of study
TIME

SMOKE

LUNG
CANCER

DO NOT SMOKE

SMOKE

NO LUNG
CANCER

DO NOT SMOKE

Direction of inquiry
(Backward directionality)

Advantages and disadvantages

Advantages
Suitable for a rare outcome or those with long lag between
exposure and outcome, and also common diseases
Relatively quick
Possible to evaluate many different exposures

Disadvantages

Increased bias in measurement of exposure

Problem in seeking control group and matching variables
Yields only odd ratio
Potential bias: recall and selection

Cohort study
Onset of study

TIME
DISEASE

EXPOSED

NO
DISEASE
DISEASE
NOT EXPOSED

NO
DISEASE
Direction of inquiry
(Forward directionality)

Subjects are defined based on the exposure status and followed over time to
determine disease development.

Advantages and disadvantages

Advantages
Powerful study design for estimation of incidence rate and risk
Able to establish causal effect
Maintain temporality between exposure and outcome
Assess various effects of a particular exposure
Reduces selection and information bias
Avoid bias in the exposure measurement
Disadvantages
Costly and inefficient for rare outcomes (unless attributable risk is
high)
Possible losses to follow-up cases (Non-response bias)
Requires long period of time to complete study
Requires large number of subjects
Difficult to measure confounding variables

Experimental study
Involves deliberate application of intervention or
treatment in the experimental group while no
expected changes in the control group and
comparing the outcomes for a duration of time.

Randomized Controlled Trial

The requirement for RCT:
Random allocation of subjects to various groups. Participants in both the
experimental and control group should be identical at the beginning of the
study, except for in the intervention that is being introduced.
Manipulation: by deliberate application or withdrawal of a tested factor
(could be treatment/any intervention).
Blinding (To reduce bias): No knowledge on type of intervention:
a) Single blinding: Participant is not aware on type of intervention received
b) Double blinding: Both participant and researcher (or doctor) are unaware
on type of intervention
c) Triple blinding: Participant, researcher and person who analyze the data
are unaware type of intervention

Randomized Controlled Trial Clinical

Trials

Process of development of new drug or vaccine involves:

Preclinical studies Phase 0 Phase I Phase II Phase III
Phase IV
Preclinical: In vitro (test tube) and in vivo (animal): establish
preliminary efficacy, toxicity & pharmacokinetic of drug
Phase 0: Test pharmacokinetic & pharmacodynamic property of
drug in human
Phase I: A small group of people (safety, feasibility & acceptability)
Phase II: A larger group of people (preliminary effect size, dosage,
efficacy, side effects)
Phase III: Large groups of people (efficacy of new treatment
compared to standard)
Phase IV: Post Marketing Surveillance Trial (application in clinical
setting & long-term safety)

Source of Error in Epidemiologic Studies

SOURCE OF
ERROR

RANDOM ERROR

SELECTION BIAS

SYSTEMATIC ERROR
OR BIAS

INFORMATION BIAS

CONFOUNDING

Error
Random error
Occurs by chance and usually cannot be predicted unless
from sampling error.
As random error increases, the precision of study
decreases.
Solution: Increasing sample size or by making the sample
representative of the population (study design).
Systematic error or bias
Occurs when there is a tendency to produce results that
are different in a systematic manner from the true value.
As systematic error increases, the validity of study
decreases.
Solution: Good study design

Bias
Selection Bias occur at the start of a
study due to method of selection is biased.
Berksons bias
Self selection bias
Lead time bias
Detection bias

Bias
Information Bias occurs because the data that
are collected or observed are incomplete or
incorrect.
Interviewer or observer bias
Loss to follow-up bias
Misclassification
Recall bias
Inter- interviewer bias
Hawthorne effect

Confounding
Definition distortion or the masking of an association
between exposure and a outcome because of third
extraneous factor.
Effects of confounding decrease, increase, or even
change the direction of the estimated association
between an exposure and outcome (conclusion is
biased)
Can be controlled by:
Randomization
Restriction
Matching
Stratification

Measurement of Risk

Determination of rates of disease by person, place and

time
Absolute risk (incidence, prevalence)

Identification of risks factors of a disease

Relative risk (or odds ratio)

Development approaches for disease prevention

Attributable risk/fraction

Relative Risk

Relative risk (RR) or risk ratio

= Incidence of disease among exposed
Incidence of disease among non-exposed

It shows how much one group is more likely to develop

disease than the other.

From a cohort/prospective study

If RR > 1, there is a increased risk for the exposed

group
If RR < 1, there is a decreased risk for the exposed
group
If RR = 1, there is an equal risk

Relative Risk

If 10 out of 1000 men who smoked over a 10 year period

develop lung cancer compared with 20 to 4000 men who
did not smoke who developed lung cancer, the relative risk
of developing cancer would be:
Lung
Cancer

No Lung
Cancer

Total

Incidenc
e

10 (a)

990 (b)

1000 (a + b)

1%

Non smoker 20 (c)

3980(d)

4000 (c + d)

0.5%

Total

4970 (b + d) 5000
(a+b+c+d)

Smoker

30 (a + c)

Relative risk = (a/ a+b) / (c/ c+d) = 1% / 0.5% = 2

Interpretation: There is two times greater chance

(probability) of developing lung cancer when a person is

a smoker compared to non-smoker

Odds Ratio

Odds ratio (OR) [Estimated RR)

= (cases with exposure) x (control without exposure)
(control with exposure) x (cases without exposure)

It is the ratio of the odds of exposure among diseased to

non-diseased.

From a case control/retrospective study

If OR > 1, there is a increased risk
If OR < 1, there is a decreased risk
If OR = 1, there is an equal risk

Odd ratio

If 100 out of 1000 men with lung cancer had smoked and
200 of the 4000 without lung cancer had smoked. The odds
of developing lung cancer in those who smoke is:
Lung
Cancer

No Lung
Cancer

Total

Smoked

100 (a)

200 (b)

300 (a + b)

Did not
smoke

900 (c)

3800 (d)

4700 (c + d)

Total

1000 (a +
c)

4000 (b + d) 5000
(a+b+c+d)

Odd ratio = (a/c) / (b/d)= ad/ bc = 100x3800 / 900x200 =

2
Interpretation: There is two fold greater odds of getting

lung cancer when a person smoked.

Attributable Risk

Attributable risk percent(AR%)

= Risk (exposed) Risk (unexposed) X 100
Risk (exposed)
= (RR-1)/ RR x (100)

Attributable risk - An amount of disease incidence that

can be attributed to a specific exposure
Excess Risk
= Risk (exposed) Risk (Unexposed)
= [a/(a+b)] [c/(c+d)]
100
80

Amount of risk that can be prevented

Population attributable risk (PAR)
= Risk (total) Risk (unexposed)

60
40
20
0

Risk Factor

Attributable risk

If 10 out of 1000 men who smoked over a 10 year period

develop lung cancer compared with 20 to 4000 men who
did not smoke who developed lung cancer, the attributable
risk for smoker is:
Lung
Cancer

No Lung
Cancer

Total

10 (a)

990 (b)

1000 (a + b)

Non smoker 20 (c)

3980(d)

4000 (c + d)

Total

4970 (b + d) 5000
(a+b+c+d)

Smoker

30 (a + c)

Attributable risk = (a/ a+b) - (c/ c+d) = 1 - 0.5 = 0.5

Interpretation: 0.5 or 50% of the lung cancer cases

among the exposed group is attributable to smoking (half

the cases of lung cancer would occur if all people quit
smoking).

Reading Materials
1. Abdul Rashid Khan and N.A Narayan. Lecture Notes

On Epidemiology. Edition 2; Chapter 7 10.

2. Timmreck, TC. An introduction to epidemiology. 3rd
edition, 2002.
3. Jekel JF, Katz DL and Elmore JG. Epidemiology,
biostatistics and preventive medicine. Second edition.

Thank you