Epidemiology

MODULE – 2

Epidemiological
Study Designs
MODULE

Epidemiological Study Designs

Module Description

This module has been designed to introduce the candidates about the various study
designs used in the epidemiological studies. This will help the candidate to develop
ability to know the conditions under which a particular study design is to be used.
The candidate will also be able to understand the role of error, bias and confounding
in the epidemiological research.

Unit 2.1

Descriptive Epidemiology

Unit 2.2

Basic Designs of Research Studies

Unit 2.3

Variation: Role of Error, Bias and Confounding

Unit Table of Contents
Unit 2.2 Basic designs of Research Studies
Topics Page No.
Learning Objectives
Learning Outcome
2.2.1 Descriptive and Analytical Study Designs
2.2.2 Experimental Study Design
2.2.3 Advantages and Disadvantages of Different Study
2.2.4 Calculations of Appropriate Parameters for all the Study
Designs
2.2.5 Conclusion
Summary
Activity
Activity – Key Answer
References
Multiple Choice Question
Multiple Choice Question – Key Answer
Introduction

Learning Objectives

• List the four main types of epidemiological studies

• Understand the difference between observational and experimental study
• Discuss the roles of cross sectional study in conducting population surveys
• Explain the steps involved in the conduct of randomised clinical trials

Learning Outcome

• Define cross sectional, case control, cohort and clinical trials

• Analyse the main deliverables of observational and experimental studies
• Recall the steps in conducting population census
• Differentiate the various types of randomised clinical trials

Epidemiology 1
Introduction

Epidemiological study designs are very important in establishing causal inferences.

However, no single epidemiological design confirms causality and none is incapable
of adding important evidence. In all studies there are limitations and pitfalls that
should be minimised as much as possible. There are various types of study designs
with differences in their types of pitfalls and their likelihood. The potential
contributions of various study designs to the epidemiological criteria for causality
are shown below. It should be noted that with the exception of consistency, to which
all designs contribute, and biological plausibility, to which no designs contribute
directly, all epidemiological studies contribute to some but not all criteria.

Table 2.2.1: Potential Contributions of Study Design to Causal Criteria

S/N Criteria Case-series Cross- Case- Cohort Trial

sectional control
1 Temporality Sometimes Sometimes Sometimes Often Usually
2 Strength or Sometimes Sometimes Often Always Always
dose response
3 Experimental Sometimes Sometimes in Seldom Sometimes Always
confirmation in case of case of following
natural repeated natural
experiment studies changes
following an
intervention
4 Specificity Sometimes Sometimes Yes for Yes for the Yes for the
diseases risk factors risk or
preventive
factors
5 Biological Not directly Not directly Not Not Not
plausibility directly directly directly
6 Consistency Yes Yes Yes Yes Yes

Epidemiology 2
Introduction

2.2.1 Descriptive and Analytical Study Designs

Classification of Study designs: There are five major categories of classifying study
designs

Fig 2.2.1: Major Categories of Classifying Study Designs

The term descriptive implies a study which provides information about the pattern
of disease or risk factors but not the underlying causes. The term analytic applies to

Epidemiology 3
Introduction

studies exploring hypotheses about causes of disease but, by inference, not primarily
concerned with patterns.

Retrospective studies are said to be concerned with data in the past and prospective
ones with data in the future. The terms retrospective and prospective have been used
synonymously with case-control and cohort studies, respectively. The distinction
between retrospective and prospective studies is also very blurred, for case-control
studies may enrol subjects prospectively and cohort studies may enrol subjects
retrospectively and both do, of course, collect data on risk factors in the past.

The observational study is one where the investigator observes the natural course of
events. The experimental study is one where the course of events is deliberately
altered. The investigation of a natural experiment is, strictly speaking, an
observational study. Most epidemiology is observational, for experiments are the
exception.

Another category of classifying study design lies in the presence or absence of

disease at the beginning of a study. Studies where the disease has already occurred
focus on the risk factors which lead to it or influence its course. Studies where the
risk factor is present but no disease has yet occurred focus on the occurrence of
disease and other outcomes.

Another division in study design is between studies which incorporate a specific

comparison group and those which do not. Those with comparison groups are
generally better for testing hypotheses about disease causation than those without,
and have usually been done with this as a primary goal.

Table 2.2.2: Study design and five categories of classifications

Epidemiology 4
 Introduction

Study Descriptiv Retrospective Observational/ Beginning with Specific

Design e/Analytic / Prospective Experimental diseases/causes comparison/no
Types of diseases such group
Case-series Descriptive Retrospective Observational Disease No
(clinical and
population)
Cross- Descriptive Retrospective Observational Both Usually not
sectional simultaneously
Case-control Analytic Retrospective Observational Disease Yes
Cohort Analytic Prospective Observational Usually causes Usually, yes
(Prospective and (though it may
and Retrospective be integral to the
Retrospectiv population
e)
Trial Analytic Prospective Experimental Usually disease, Yes, with
but sometimes exceptions
causes of disease

[Link] Cross-sectional study:

Fig 2.2.2: Cross-sectional Study

Epidemiology 5
Introduction

The word cross-section describes the shape that results from cutting an object
lengthwise. A cross-sectional study, therefore, exposes and studies disease and risk
factor patterns in a representative part of the population, in a narrowly defined time
period. Also known as prevalence study, to emphasise the key role of cross-sectional
studies in Epidemiology. The cross-sectional study seeks associations, generates and
tests hypotheses and, by repetition in different time periods, can be used to measure
change, and hence evaluate interventions. Its focus is simultaneously on disease and
population characteristics and risk factors. Comparisons between subgroups within
the sample are invariably made, but the study can also be deliberately designed with
comparison groups. The comparisons are usually based on differences in the
prevalence of risk factors and diseases, and the association between risk factors and
diseases.

An ideal cross-sectional study is of a geographically defined, representative sample

of the population studied within a slice of time and space. Rarely, cross-
sectional studies are of the whole population e.g. population census. Cross-sectional
surveys provide a ‘snapshot’ of health. The selection, compilation, and definition of

Epidemiology 6
Introduction

the population at risk and the listing of the sampling frame usually conform to the
snapshot analogy. But the measurement of risk factors and disease is usually made
over a period of time which varies from as little as a day to several years. In most
studies, the measurements are made over a relatively short period of time such as a
year or two.

The benefit of a study collecting data over a year is that any seasonal differences will
be evened out to give a more valid annual measure of prevalence. When the time
period of data collection is long, the degree of mismeasurement of the point
prevalence of disease depends on the natural history. If the disease is permanent
then the point prevalence may be overestimated because incident cases will occur
and be included during the duration of the fieldwork. Because most diseases are rare
so the effect will be small. If the disease incidence or base population is changing
over the duration of the study, this too will affect the disease prevalence. If there is a
dynamic but balanced state with new cases arising and old cases recovering in equal
numbers, then the mismeasure of the point prevalence is small. The
mismeasurement of point prevalence is likely to be important for diseases which
vary greatly by season or where the incidence of a disease is changing rapidly. The
cross-sectional study design is excellent for measuring the population burden of
disease using prevalence rates, which are the most reliable summary measures
obtained from such surveys.

In studies of the apparently well, cross-sectional studies may discover people with
previously unknown disease, that is, they uncover the iceberg of disease. The full
spectrum of disease can be described only by a combination of cross-sectional
surveys of the apparently well population and the diseased population, the latter is
more often obtained from clinical case-series than from cross-sectional studies. The
cross-sectional study can be of populations in different places, so comparisons can be
made. Studies can also compare people with different characteristics; for example,
there may be a sample of women and another of men, or one of people belonging to
a Chinese origin population and another of the Indian population. Such studies are
called comparative cross-sectional studies.

[Link] Case control study:

Fig 2.2.3: Case Control Study

Epidemiology 7
Introduction

The case-control study is a comparative study where people with the disease (or
problem) of interest are compared with people without that disease. The word case
is used to describe the characteristics and medical history of a patient. The
comparison, control or reference group supplies information about the expected risk
factor profile in the population from which the case group is drawn.
The cases can be obtained from a number of sources e.g. a clinical case-series, a
population register of cases, from the new cases identified in a cohort study, and
from those identified in a cross-sectional survey.

The ideal set of cases would be new (incident) and representative of all cases of the type of
interest to the study question in the population under study. The cases from population registers
and cohort studies usually meet this ideal the best. The cases identified in a clinical case-series
are usually highly selected; while those from a cross-sectional study are usually prevalent ones,

Epidemiology 8
Introduction

though there will be incident cases in a period prevalence study. The cases are compared with
controls, associations between the disease and potential risk factors are measured (usually by the
odds ratio), and through analysis of similarities and dissimilarities, hypotheses about disease
causes are generated or tested.

Information is obtained on the social and medical history of cases and controls and on potential
causal and confounding factors. As the causal factors have already had their effect in causing
disease in the case group, and the information required is recalled from the past, the case-control
study is sometimes referred to as a retrospective study.

In some studies, controls are recruited to match each case; for example, if a woman of 53 years
was recruited as a case, one could seek a control of similar age. This matching process is
reducing the risk of confounding, here by age and sex. If a mix of ages is likely to arise anyway,
the control group can be recruited without one-to-one matching.

Information is collected to confirm the presence of disease in cases and the absence of disease in
controls and on the past exposure to risk factors which may have caused the disease in both
groups. The concept is to find differences in exposure to the hypothesized causes in the past lives
of cases as compared with controls. These differences can be quantified and summarised either
as differences in prevalence of exposure, or more usually as the odds ratio which in defined
circumstances approximates to the relative risk. An exposure that may have caused disease will
be more common in cases than in controls giving an odds ratio greater than one, and one that
may protect against disease will be less common, giving an odds ratio less than one. Of all the
epidemiological designs, case control is most focused on establishing aetiology and least on
measuring burden of disease or risk factors, which is a by-product.

The estimate of risk in a case control study using the odds ratio, as a valid estimator of the
relative risk, is based on the following assumptions:

• The cases are incident cases drawn from a known and defined population;
• The controls are drawn from the same defined population and would have been in the
case group if they had developed the disease;
• Controls are selected in an unbiased way, e.g. independently of exposure status;
• Some types of study where the disease is rare.

[Link] Cohort study:

Fig 2.2.4: Cohort Study

Epidemiology 9
Introduction

The word cohort is derived from the Latin ‘cohors’ meaning an enclosure, company,
or crowd. In Roman times, a cohort was a body of 300–600 infantry. In
epidemiological terms, ‘cohort is a group of people with something in common, usually an
exposure or involvement in a defined population group’. The three synonyms for cohort
study design are

• Follow-up
• Longitudinal, and
• Prospective study

Cohort study involves tracking the study population over a period of time.

The cohort study population may be a general one, or one with characteristics of
particular interest, for example, people with a defined lifestyle or even a disease. The
hallmark of this design is that health outcome or health change data are obtained on

Epidemiology 10
Introduction

the same individuals in a population at more than one time, not just once as in the
cross-sectional study. The idea is to study part of the natural history of risk factors or
diseases in individuals, and to relate one or more characteristics, exercise for
example, to future outcomes such as coronary heart disease.

In causal research, cohort studies usually test the hypothesis that disease incidence
differs in people with different characteristics (exposures) at baseline; that is, there is
an association between exposure and outcome. The cohort study begins by
establishing baseline data, usually from a cross-sectional study, or less commonly by
the extraction of baseline data from sources such as the census or a routine
information system such as a birth register. The cohort can either be followed up
directly with repeated surveys of the same population or the baseline data can be
linked to health records, so providing information on outcomes of interest, usually
disease related but potentially also on risk factors. The new cases of disease
identified are incident cases and can be enrolled into a case-control study. Controls
can also be identified from within the cohort, and this is best done as each case
occurs. This is known as a nested case-control study.

Retrospective cohort study

Fig 2.2.5: Retrospective cohort study

Epidemiology 11
Introduction

Retrospective cohort study/ies’ are cohort studies done in the past without any
prospective work. Essentially, the cohort is identified from past records of exposure
status and this is the vital step. Usually, the outcome data is also obtained in the past
from records (but this information can be supplemented with direct questioning of
those subjects who are alive, and can be traced). Once identified the subjects can be
followed up over time (prospectively) so using both currently available and future
data on outcome. The difference between this design and the prospective cohort is
minimal; a retrospective cohort is assembled from historical records on exposure
status, the prospective cohort on exposure status in the present. Cohort studies are
often described as analytic but one of their main functions is to provide information
on the incidence and to describe the natural history of disease and not just to explore
or generate hypotheses, for which they are of course, extremely useful. If the cohort
study is based on a defined and characterised population, the incidence rates can
often be extrapolated beyond the study group to similar populations elsewhere. ‘The
most important information from a cohort study is on incidence rates’. The ratio of the
incidence rates in the exposed and non-exposed groups derived from the cohort
study is the relative risk; the primary basis for measuring the strength of an
association.

[Link] Trials:
Fig 2.2.6: Trials

Epidemiology 12
Introduction

Trials are studies where an intervention designed to improve health has been
applied to a population, and the outcome is assessed at follow-up. Such studies give
information about the causes of a disease, effectiveness of interventions to influence
the natural history of disease, and the costs and benefits of interventions.
Trials are experiments that are not done in the laboratory setting, and are on human
or whole animal studies. The term is also known by other terms such as ‘Intervention
studies, clinical trials, and community trials’.

The trial has, essentially, the same design as a cohort study with one vital difference,
that the exposure status of the study population has been deliberately changed by
the investigator to see how this alters the incidence of disease or other features of the
natural history. Clinical and public health trials usually have a practical question to
answer: whether a particular intervention is sufficiently effective to be introduced into
clinical or public health practice. Hence, trials must be based on a study population
with proper understanding of how it relates to the (target) population which will be
offered the intervention should it be shown to be successful. Otherwise, an
intervention which works in a selected population may not fulfil its goals when put
into public health or clinical practice in the general population.

Proof of concept trials: are trials designed solely to produce knowledge about cause
and effect; the intention being to test the efficacy of the intervention in actual practice

Epidemiology 13
Introduction

at a later date. For these trials, understanding the relationship of the study
population to the target population is not so essential (but still advised).

In preventive trials the intervention may be either an active intervention, say

enrolment into a diet and exercise programme, or the manipulation of a natural way
of life such as reducing the consumption of salt. Preventive trials are more difficult
to do than trials of treatment based on drugs (clinical trials). The first step in
conducting trials is to define a study population suitable for answering the question,
i.e. people either with disease (for clinical trials) or without (for prevention trials).
Ideally, this study population will be drawn from the target population. The study
population is divided then into two or more groups, the intervention group(s) and
the control group(s). While the intervention receives the intervention e.g. an exercise
programme, the control group may be offered the best known alternative (e.g.
meditation classes) or a placebo activity with no known effect on the outcome (e.g.
participation in a pottery class). The important thing is that the two groups gain an
equal amount of attention in the study. Otherwise, the changes seen might be
attributable to differences in the amount of attention each group receives, not the
intervention itself. The intervention group provides information on prognosis, the
control group on the natural history. In the ideal trial, the study and control
populations are at the same stage of the natural history of disease and are similar in
the characteristics that affect disease outcomes, differing only in exposure to the
intervention. To maximise the chances that the intervention and control groups are
the same at baseline, and hence avoid confounding, the trial design should ensure
that individuals in the study are assigned randomly to the groups. This process
solves the problem of finding the right control group. Where this is done, it is then
called a Randomised, Controlled Trial. Where there is no ‘best known alternative’ an
intervention which is ‘psychologically’ of similar impact to the study intervention,
but has no influence on the diseases process called placebo is used (a placebo, from
the Latin word to please). In this case, the study is known as a Placebo-Controlled
Randomised Trial.

2.2.2 Experimental Study Design

Epidemiology 14
Introduction

Formerly, "experimental epidemiology" meant the study of epidemics among colonies

of experimental animals such as rats and mice but in modern usage, Experimental
Epidemiology is often equated with Randomised Controlled Trials which can be
performed with both human and animal population. It is similar in approach to
cohort studies excepting that the conditions in which study is carried out are under
the direct control of the investigator who may carry out some intervention or
manipulation such as deliberate application or withdrawal of the suspected cause or
changing one variable in the causative chain in the experimental group while
making no change in the control group. The aims of experimental studies may be
stated as follows:

a) To provide "scientific proof" of aetiological {or risk) factors which may permit
the modification or control of those diseases.
b) To provide a method of measuring the effectiveness and efficiency of health
services for the prevention, control and treatment of disease and improve the
health of the community.

[Link] Randomised controlled Trials (RCT): The randomised trial is the gold
standard for evaluating the efficacy of therapeutic, preventive, and other measures
in both clinical medicine and public health study designs because randomisation, if
correctly conducted, prevents any biases on the part of the study investigators from
influencing the treatment assignment for each participant. If our study is large
enough, randomisation will also most likely lead to comparability between
treatment groups on factors that may be important for the outcome, such as age, sex,
race etc., as well as for factors one has not measured, or may not even be aware of as
important. The basic steps in conducting a RCT include the following:

Epidemiology 15
Introduction

Fig 2.2.7: Basic Steps in Conducting a RCT

1. The protocol

This is an essential feature of a randomised controlled trial. The protocol specifies the
aims and objectives of the study, questions to be answered, criteria for the selection of
study and control groups, size of the sample, the procedures for allocation of subjects
into study and control groups, treatments to be applied when and where and how to
what kind of patients, standardisation of working procedures and schedules as well as
responsibilities of the parties involved in the trial, up to the stage of evaluation of
outcome of the study. A protocol is essential especially when a number of centres are
participating in the trial. Once a protocol has been evolved, it should be strictly adhered
to throughout the study. The protocol aims at preventing bias and to reduce the sources
of error in the study. Preliminary test running of the protocol may be necessary to see
whether it contains any flaws. The final version of the protocol is then agreed upon by
all concerned before the trial begins.

2. Selecting reference and experimental populations

(a) Reference or target population: It is the population to which the findings of the trial,
if found successful, are expected to be applicable (e.g., a drug, vaccine or other
procedure). A reference population may be as broad as mankind or a whole city, it may
be geographically limited or limited to persons in specific age, sex, occupational or social
groups, e.g. population of school children, industrial workers, and obstetric population.

(b) Experimental or study population: The study population is derived from the
reference population. It is the actual population that participates in the experimental
study. Ideally, it should be randomly chosen from the reference population, so that it
has the same characteristics as the reference population. If the study population differs
from the reference population, it may not be possible to generalise the findings of the
study to the reference population.

It is important to choose a stable population whose cooperation is assured to avoid

losses to follow-up. The participants or volunteers must fulfil the following criteria:

Epidemiology 16
Introduction

i. They must give "informed consent", that is they must agree to participate in the
trial after having been fully informed about the purpose, procedures and possible
dangers of the trial
ii. They should be representative of the population to which they belong (i.e.,
reference population)
iii. They should be qualified or eligible for the trial. That is, let us suppose, we are
testing the effectiveness of a new drug for the treatment of anaemia. If the
volunteers are not anaemic, one will then say, they are not eligible or qualified
for the trial. In other words, the participants must be fully susceptible to the
disease under study.

3. Randomisation

This is a statistical procedure by which the participants are allocated into groups
usually called "study" and "control" groups, to receive or not to receive an
experimental preventive or therapeutic procedure, manoeuvre or intervention.
Randomisation is an attempt to eliminate "bias" and allow for comparability. Unlike
matching which can only match those factors which are known to be important to
assure comparability randomisation seeks to take care of other factors which are
important but whose effect is not recognised or cannot be determined. By a process
of randomization, hopefully, these factors will be distributed equally between the
two groups. In other words, by random allocation, every individual gets an equal
chance of being allocated into either group or any of the trial groups. Randomisation
is done only after the participant has entered the study that is after having been
qualified for the trial and has given his informed consent to participate in the study.
Randomisation is best done by using a table of random numbers.

4. Manipulation

In this stage, there is deliberate application or withdrawal or reduction of the

suspected causal factor (e.g., this may be a drug, vaccine, dietary component, a habit
etc.) as laid down in the protocol from the study (experimental) group. This
manipulation creates an independent variable (e.g., drug, vaccine, a new procedure)
whose effect is then determined by measurement of the final outcome, which
constitutes the dependent variable (e.g., incidence of disease, survival time, recovery
period).

5. Follow-up

Epidemiology 17
Introduction

This implies examination of the experimental and control group subjects at defined
intervals of time, in a standard manner, with equal intensity, under the same given
circumstances, in the same time frame till final assessment of outcome. The duration
of the trial is usually based on the expectation that a significant difference (e.g.
mortality) will be demonstrable at a given point in time after the start of the trial. It is
possible that some losses to follow-up are inevitable due to factors, such as death,
migration and loss of interest. This is known as attrition. If the attrition is substantial,
it may be difficult to generalise the results of the study to the reference population.

6. Assessment

This is the final step to assess the final outcome of the trial in terms of:

a) Positive results: That is, the benefits of the experimental measure such as reduced
incidence or severity of the disease, cost to the health service or other appropriate
outcome in the study and control groups.
b) Negative results: That is, the severity and frequency of side-effects and
complications, if any, including death. Adverse effects may be missed if they are
not sought.

The incidence of positive/negative results is rigorously compared in both the groups,

and the differences, if any, are tested for statistical significance.

Bias may arise from errors of assessment of the outcome due to human element.
These may be from three sources:

a) Subject Bias: There may be bias on the part of the participants, who may
subjectively feel better or report improvement if they knew they were receiving a
new form of treatment. This is known as "subject variation".
b) Observer Bias: That is the investigator measuring the outcome of a therapeutic
trial may be influenced if he knows beforehand the particular procedure or
therapy to which the patient has been subjected.
c) Evaluation Bias: There may be bias in evaluation, that is, the investigator may
subconsciously give a favourable report of the outcome of the trial.
Randomisation cannot guard against these sorts of bias, nor the size of the
sample.

In order to reduce these problems, a technique known as "Blinding" is adopted,

which will ensure that the outcome is assessed objectively. Blinding can be done in
three ways.

i. Single Blind Trial: The trial is so planned that the participant is not aware
whether he belongs to the study group or the control group.

Epidemiology 18
Introduction

ii. Double Blind Trial: The trial is so planned that neither the investigator nor
the participant is aware of the group allocation and the treatment received.
iii. Triple Blind Trial: This goes one step further. The participant, the
investigator and the person analysing the data are all "blind". Ideally, of
course, triple blinding should be used; but double blinding is the most
frequently used method when a blind trial is conducted. When an outcome
such as death is being measured, blinding is not so essential.

2.2.3 Advantages and Disadvantages of Different Study

Each study has advantages and disadvantages and no one study design is superior.
The ‘hierarchy of evidence’ whereby the trial is said to produce definitive evidence,
and other designs weaker evidence, is a narrow idea that only applies to evaluation,
particularly of drugs. Other designs are stronger for measuring the burden of disease
and in understanding causality. It should be noted that causal understanding comes
from all types of study. Some of the strengths and weaknesses of each study design
are given in the Table below.
Table 2.2.3: Advantage & Disadvantage of Different Study

S/N Themes Cross- Case-control Cohort Trial

sectional
1 Ease Difficulty Usually Difficulty Difficulty because
depends on difficult because of the of technical and
the study. because of need added ethical challenge of
Studies of for appropriate complexity of imposing an
natural control group follow-up intervention
living and problem of
populations recall bias
are hard
compared
with those at
schools or
other
institutions.
2 Timing Usually Usually Usually long Usually deliberately
finished finished within term(decades) designed to give
within months or a few though answers within a
months or a years except sometimes they few years or a

Epidemiology 19
Introduction

few years those on can be quick like decade(i.e. shorter

incident cases studies of birth than cohort studies)
of rare diseases. outcomes
3 Maintenance Study is Study is usually Long term Same as cohort but
and usually stopped continuity is when trials are in
continuity stopped essential and people with the
problematic as disease,
observations are commitment to the
on free living trial may be high
people
4 Costs Costs Costs usually High costs High cost for same
depend on comparable to because of the reasons as cohort
the study the cross large sample size and additionally
but lower sectional but as and the cost of due to high cost of
than cohort the study size is retaining staff intervention,
or trial of small, the and system to obtaining ethical
same size overall costs collect data over approval and trial
may be small the years management
5 Ethics Standard Standard ethical Standard ethical The ethics of trials
ethical issues as in issues are acute are complex and
issues and cross sectional especially as evolving and hinge
problem of and also of adverse on the concept of do
access to community outcomes may no harm and
sampling controls affect occupation informed consent
frame and insurance
premium.
Potential
intrusion of
repeated
contact and
measurements
6 Data Usually As analysis is Data tend to be Data concerning the
Utilisation under- straight- under-utilized central question are
utilized, as forward, data utilized.
more are usually
information fully analysed
is collected
than needed
7 Major Major Major Major Major contribution
contribution contribution contribution to contribution to to understanding
to burden of clinical both burden of the effectiveness of
diseases, knowledge, and disease and the intervention and
substantial sparking/testing casual analysis. indirectly to disease
contribution casual mechanism

Epidemiology 20
Introduction

to analysis hypotheses.
of Control group
association may supply
and may burden of need
confirm or data
spark
hypothesis
8 Observer Small Small studies Usually require Usually require
Bias studies may may be done by multiple multiple observers
be done by one observer observers but
one observer but large exceptionally
but for most studies usually studies may be
studies require a few small.
,inter-
observer
bias may be
a problem.
9 Selection Selection Studies of Selection bias Selection biases are
Bias bias arising prevalent cases due to non- particularly severe
from non- have selection response at because non-
response is bias, those of baseline is participation may
almost incident cases augmented by be high and
inevitable minimize this. loss to follow up intervention may
All studies have only be suitable for
recall bias some of the target
population.
10 Analytic Main output Proportions Incidence rate Incidence, survival
output is exposed and and the relative and numbers
prevalence, odds ratio incidence i.e. needed to treat or
though other relative risk prevent
measures
like the odds
ratio are
possible

2.2.4 Calculations of Appropriate Parameters for All the Study

Designs

Epidemiology 21
Introduction

Odds Ratio: Alternative approach to measure association in case-control studies, the

odds ratio is the ratio of the odds of the cases having been exposed to the odds of the
controls having been exposed. In Case-Control Studies, the odds ratios is the odds of
the cases having been exposed to the odds of the controls having been exposed or
the odds of exposed people developing the disease to the odds of non-exposed
people developing the disease. Uses of odds ratio include the following:

Fig 2.2.8: Uses of Odds Ratio

The odds ratio is a good measure of relative risk when the disease being studied is
rare.
Table 2.2.4: Odds Ratio

Suspected or Risk factors Cases(Disease present) Control(Diseae absent)

Present A b
Absent C d

Odds of exposure for case control studies:

Odds in cases = a/c

Epidemiology 22
 Introduction

Odds in contro l= b/d

Odds ratio (OR) = odds in cases/odds in control= ad/bc

• When exposure is associated with disease, Odds ratio > 1

• When exposure is associated with protection from disease, Odds ratio < 1

• When exposure is unassociated with disease, Odds ratio = 1

Relative risk is the rate of disease in exposed compared to rate of disease in

unexposed subjects. It is used to calculate measure of association in cohort studies.
Rate of disease in exposed subjects
Relative risk =
Rate of disease in unexposed subjects

In cohort study, the Relative risk is the rate of disease in exposed compared to rate of
disease in unexposed subjects.
Table 2.2.5: Cohort Study

Disease at entry Diagnosed as having disease during Total

point follow-up
Yes No
Yes A b a+b
No C d c+d

For a cohort study,

When the relative risk > 1, exposure is associated with disease

When the relative risk < 1, exposure is associated with protection from disease
When the relative risk = 1, exposure is unassociated with disease

Where the disease is rare

a a c c
 if a  b  if c  d
a+b b c+d d
a (a + b ) a b ad Odds ratio  RelativeRisk
Thus,  =
c (c + d ) c d bc

Epidemiology 23
Introduction

2.2.5 Conclusion

Study design can be likened to a system of interlinked and mutually supporting

methods. As the various designs having similar purposes are rooted in population
concepts of health and disease, and are conceptually overlapping, they are also
subject to similar errors, biases, and problems of sampling, and similar challenges in
data collection, analysis, and interpretation. In practice, a mix of designs is used to
solve a problem, and it may be difficult to name the design of the study. Most
studies can be distinguished by their focus on either disease or exposure, the
relationship of the observation to calendar time and the natural history of disease,
and whether there is an imposed intervention.

Epidemiology 24
Introduction

Summary

o Epidemiological studies can be broadly classified into observational or

experimental studies.

o The observational study is one where the investigator observes the natural
course of events.

o The experimental study is one where the course of events is deliberately

altered.

o Retrospective studies are said to be concerned with data in the past.

o Prospective studies are concerned with data in the future.

o A cross-sectional study exposes and studies disease and risk factor patterns in
a representative part of the population, in a narrowly defined time period. Its
focus is simultaneously on disease and population characteristics and risk
factors.

o The case-control study is a comparative study where people with the disease
(or problem) of interest are compared with people without that disease.

o Case control is most focused on establishing aetiology and least on

measuring burden of disease or risk factors, which is a by-product.

o Cohort study involves tracking the study population over a period of time.

Epidemiology 25
Introduction

Case Study-2
In a case study of prospective study design, comprising of 10000 subjects, 6000 subjects were
put on beta carotene and 4000 were not. Three out of the first 6000 developed lung cancer
and 2 out of the second 4000 developed lung cancer.

Question: How you will interpret this:

Answer: Relative Risk (Risk Ratio) is used to estimate risk for disease (calculated as
incidence of that disease) with exposure to a factor
Relative Risk= Incidence among exposed ÷Incidence among Non-exposed
In given case study, Exposure is beta carotene and disease is lung cancer
Incidence of lung cancer among those not exposed to beta carotene (Iexp) = 2/4000
Therefore, RR=1 (i.e, Iexp is same as Inonexp)
If RR =1, it implies ‘incidence among exposed’ is same as ‘incidence among non-exposed’.
Therefore, whether the person is exposed or not (to a factor),incidence of disease developing
later will remain the same. Thus, exposed (beta carotene) and disease the under study (lung
cancer) are not associated at all. Thus ‘beta carotene is neither causative nor protective for
lung cancer’

Epidemiology 26
Introduction

Bibliography

E-References

1. Basic Epidemiology - R Bontia ,R Beaglehole , T Kjellstrom

2. Epidemiology-An Introduction by Kenneth J. Rothman
3. Concepts of Epidemiology: An integrated Introduction to the ideas, theories, principles
and methods of Epidemiology by Raj Bhopal
4. Principles of Epidemiology - Self-Study U.S. Department of Health and Human Services
5 .A Short Introduction to Epidemiology -Neil Pearce

Epidemiology 27