Introduction to Study Designs in Epidemiology

1
Introduction to Study Designs in Epidemiology

• Lecturer Dr Ashraf
  El-Metwally
• Course, Module Bsc, Epidemiology
  Module
• Date
  09-October-2007

2
Aim of session

• To understand the concepts of different study
  designs
• To able to describe cross-sectional and cohort
  studies
• To learn about the advantages and disadvantages
  of these two study designs
• to be able to calculate three measurements of
  disease association Prevalence ratio, Risk ratio
  and Rate ratio (the last two known as relative
  risk).

3
Epidemiology

• The study of the occurrence, distribution
• and determinants of disease (and other
• health-related conditions) in populations

4
The Definition

• Study
• Disease or a health-related event
• Population

5
What questions?

• 1- What ( What is the burden or frequency of the
  disease)
• 2- Who ( Does the frequency vary according to
  Age, sex, race and socioeconomic conditions)
• 3- Where (Does the frequency vary according to
  geographic location)
• 4- When (Does the frequency vary calendar time)
• 5- Why (Why do some people get the disease and
  others do not get it)

6
Etiology is for Primary prevention

• Prevention of diseases is any activity which
  reduces the burden of the disease in the
  population.
• 1- Primary prevention
• Avoids the development of a disease
• 2- Secondy prevention
• Early disease detection, for early interventions
• Prevent progression of the disease.
• 3- Tertiary prevention
• Reduces the impact of an already established
  disease
• Restoring function
• Reducing disease-related complications.

7
Overview of study designs in Epidemiology

• Descriptive studies
• Focus on identifying the pattern and frequency
  of health events in a population.
• Estimate a specific measurement (mean, median,
  odds, prevalence, incidence).
• Analytic studies
• Focus on the search of determinants of health
  outcomes.
• Study a relationship between two events.

8
Study Designs in Epidemiology
STUDY DESIGNS
Descriptive
Analytical
Case report
Case series
Experimental
Observational
Cross-sectional

• Cross-sectional
• Cohort
• Case-control
• Migrant
• Ecological
• Nested case-control

Clinical trial
Field trial
9
Main Study Designs

• Cross-sectional survey
• Assess exposure and outcome simultaneously
• Case-control study
• Known outcome, assess prior exposure
• Cohort study
• Known exposure, assess (future) outcome

10

• Cross-sectional studies

11
Definition of a Cross-sectional study

• An analytical epidemiological study which
  involves the assessment of exposure and outcome
  simultaneously at a fixed point in time

12
Basic features

• Assesses both the exposure and outcome
  simultaneously, at a single point in time.
• Calculates prevalence, but not incidence.
• Analysed using prevalence ratio.
• The first step in testing associations.

13
Cross-Sectional Studies

• Advantages
• Quick, cheap
• Easy to obtain prevalence
• Outcome
• Exposure
• Easily adapted design
• Case-control study
• Prospective cohort study

• Disadvantages
• Cannot measure disease onset
• Problem of temporality (not a problem if exposure
  is constant)
• Not suitable for rare disease

14
Cross-sectional study design (when is it
appropriate?)

• Generate hypotheses
• Exposure is constant
• Outcome is not rare
• Outcome is not long-lasting (can be dealt with)

15
Cross-sectional study design (when is it
appropriate?)

• Exposure Smoking
• Outcome Lung cancer ?
• Exposure Iris colour
• Outcome Ocular melanoma ?
• Exposure Blood group (ABO)
• Outcome Hypertension ?
• Exposure Specific gene
• Outcome Urinary tract infection ?

16

• Cohort Studies

17
Cohort study

Developed the outcome
Exposed
Did not develop the outcome
Study population
Developed the outcome
Un-exposed
Did not develop the outcome
Direction of the research inquiry
18
Cohort Studies Advantages

• Can measure disease onset / incidence
• Can measure disease risk
• Best observational method to study aetiology
• Exposure measurement precedes disease
• Can examine gt1 outcome

19
Study of Pain in Schoolchildren

• EXPOSURES
• Anthropometics
• Lifestyle factors
• Mechanical factors
• Psychosocial / behavioural factors
• Other somatic symptoms
• Familial factors

• OUTCOMES
• Low back pain
• Chronic widespread pain
• Abdominal pain
• Headache
• Fatigue

20
Cohort Studies Disadvantages

• Resource intensive
• Time
• Money
• Follow-up period depends on disease
• Short Occupational heavy lifting - Low back
  pain
• Long Child abuse - psychological diseases in
  adults
• Not suitable for rare disease

21
Types of Cohort studies
FUTURE
PRESENT
PAST
FUTURE
PRESENT
PAST
Time
Prospective cohort
Known
Assess
exposure
outcome
Known
Retrospective cohort
Assess
exposure
outcome
Assess
outcome
22
Prospective Cohort (concurrent cohort)

• 1. Investigator ascertains exposure and no
  exposure at present time.
• 2. Cohort will then be followed for a given time
  period, and incidence will be measured during
  this follow-up period.
• Examples
• Framingham Heart Study, Nurses Cohort Study,
  Royal College of General Practitioners' Oral
  Contraception Study, British Doctors study.

23
British Doctors study

• Subjects identified from British medical register
  in 1951
• Questionnaire about smoking habits
• 34,349 men and 6,194 women
• Renewed 1957, 1966, 1972, 1978, 1990
• 2/3 decreased by 1990
• Causes of death

24
British Doctors (2)

• Non-smokers 1951
• - 89 remained non-smokers in 1990
• Former smokers 1951
• 5 current smokers in 1990
• Current smokers 1951
• 74 quitted by 1990 !

25
British Doctors (3)
26
Framingham Heart Study

• Started in 1948
• 5209 men and women, ages 35-54
• Examinations every two years
• Almost 50 years of follow-up
• Cardiovascular disease

27
Framingham Heart Study (2)
28
Retrospective Cohort (Historical cohort)

• Investigator ascertains exposure and no
  exposure from information recorded in the past.
• 2. Outcome of individuals in the cohort is
  ascertained at the present time.
• Examples Sellafield plant of nuclear fuels study

29
Sellafield plant of Nuclear Fuels study

• Effect of external radiation on overall mortality
  
• The study started in 1976
• Study population14, 282 workers employed at the
  plant from 1947 till 1975.
• Investigators reviewed records to retrieve
  information on the level of exposure.
• High level Vs. low level exposure groups were
  followed up till 1988
• Mortality experienced by both groups was then
  compared.

30
Selection of a cohort

• Cohort a group of people having a common
  characteristic (from the Latin cohors)
• A. Community-based studies (a sample of the
  population).
• B. Selection of special exposure groups (e.g.,
  occupational groups).
• C. Selection of groups with medical records
  readily available (armed forces).

31
Selection of a cohort (2)

• Selection of the particular group to serve
  as the study population depends on
• 1- The specific hypothesis under investigation
• 2- Logistic and Practical considerations

32
Hypothesis under investigation

• 1- Exposure to dyestaff and the risk of bladder
  cancer (rare exposure, can not select a
  population-based cohort)
• 2- Exposure to high dose of ionising radiation
  and the risk of blood cancer (conduct a cohort of
  highly exposed people)
• If the exposure is rare, a study of general
  population will have little ability to detect an
  effect. In that case, it is wise to choose a
  cohort which includes highly exposed individuals.
  

33
Logistic and Practical considerations

• Framingham Heart Study (stable community)
• British Doctors cohort
• 
  
• 
  easy to identify and follow
• Nurses Health study
• Researchers choosed a cohort of nurses and
  doctors not because the exposure was rare in the
  general population, but for practical reasons.

34
Comparison Group

• If the exposure is common in the general
  population, we can simply use an Internal
  comparison group
• Eg. Studying the relationship between smoking
  and bladder cancer
• If the exposure is rare in the general
  population, we usually study high risk groups and
  use an external comparison group.
• Eg. Studying the relationship between Exposure to
  dyestaff and the risk of bladder cancer

35
Follow-up

• Require
• Reasonable number of cases of disease onset (not
  suitable for rare diseases)
• Measure outcome
• Use same disease definition as at baseline!

36
Bias in Cohort Studies

• Can be many!
• Non-participation
• Of those invited to participate, responders are
  systematically different from non-responders with
  respect to the relationship under examination
• Loss to follow-up
• Of those who participate at baseline, follow-up
  responders are systematically different from
  non-responders with respect to the relationship
  under examination

37
Minimising non-participation or Loss to Follow-up

• Send reminders
• Can you assess bias from non-participation?
• - No, as we seldom can know the exposure and
  disease status of non respondents
• Can you assess bias from loss to follow-up?
• To some extent, by comparing their exposure
  status with those found at follow-up.

38
Analysis of cross-sectional and cohort studies

• Estimate of Disease Occurrance in Exposed group
• Estimate of Disease Occurrance in non-Exposed
  group

39
Analysis of cross-sectional and cohort studies
(2)

• In Cross-sectional study we calculate
• Prevalence in the exposed group, then
• Prevalence in the non-exposed group, then
• Prevalence ratio
• In a cohort study we calculate
• Incidence in the exposed group, then
• Incidence in the non-exposed group, then
• relative risk ( risk ratio or rate ratio)

40
2x2 Table
Outcome
-


Exposure
-
41
Analysis of cohort and case-control studies

• Cohort study
• (with outcome estimated as a cumulative risk)
• Cumulative risk of disease in exposed
• A / (AB)
• Cumulative risk of disease in non-exposed
• C / (CD)
• Risk Ratio
• A / (AB) / C / (CD)

• Cross-sectional study
• Prevalence of disease in exposed
• A / (AB)
• Prevalence of disease in non-exposed
• C / (CD)
• Prevalence ratio
• A / (AB) / C / (CD)

42
A Cross-sectional study example

• In a cross-sectional study investigating the
  relationship between watching TV and low back
  pain, a researcher interviewed 603 individuals.
  She knew the average number of hours spent
  watching TV during the past month and also
  collected information on low back pain symptoms
  during the past month. Results are presented in
  the following table..

43
A Cohort study example

• In a cohort study investigating the relationship
  between watching TV and low back pain, a
  researcher followed up 603 individuals. At
  baseline, she knew the average number of hours
  spent watching TV during the past month and also
  collected information on low back pain symptoms
  during the past month. At one-year follow-up,
  subjects who were free of LBP at baseline were
  again interviewed to know whether or not they
  developed LBP. Results are presented in the
  following table

44
Table showing results
LBP
LBP
Non-LBP
gt2hrs
Hours of TV per day
0-2hrs
45
Calculating Prevalence ratio and Risk Ratio

• 1- Prevalence/Risk of disease in exposed
• 119 / 298
• 2- Prevalence/Risk of disease in non-exposed
• 89 / 305
• 3- Prevalence Ratio/Risk ratio
• 119 / 298 / 89 / 305 1.4
• 4- Interpretation individuals who watch TV more
  than 2 hours per day had 40 higher risk of LBP
  (or 1.4 times higher risk of LBP) than those who
  watched TV less than 2 hours per day.

46
Example (Rate Ratio)

• A cohort of 15,000 subjects were followed up for
  10 years to investigate the relationship between
  smoking and lung cancer.
• 7000 smokers were followed up for 50,000 person
  years and 55 developed lung cancer.
• 8000 non-smokers were followed up for 70,000
  person years and 10 developed lung cancer.
• Calculate Rate Ratio

47
Answer for Rate Ratio example

• Incidence rate in smokers 55/50,000 0.0011
  11 per 10,000 person years
• Incidence rate in non-smokers 10/70,000
  0.00014 1.4 per 10,000 person years
• Rate ratio 11/1.4 7.9
• Interpretation Smokers have 8 times higher risk
  of lung cancer than do non-smokers

48
How to interpret Prevalence Ratio and Risk Ratio?

• If equals to 1 This means that there is no
  association between the exposure and the disease.
  
• If more than 1 (1 until infinity) This means
  that the exposure is a risk factors for the
  disease
• If less than 1 (from 0 to 1) This means that the
  exposure is a protective factor for the disease.

49
Cross-Sectional Studies (main points)

• Step number one in studying any potential
  association (e.g Depression pain)
• Simultaneously
• Temporality of the relationship can not be
  tested.
• Exposure constant, outcome not rare.
• Prevalences are compared (prevalence ratio), but
  incidence can not be estimated.

50
Cohort Studies (Main points)

• Resources
• Relative Risk
• Methodological ideal for the examination of a
  causal association
• Measures disease incidence in exposed and
  non-exposed
• Measuring exposure precedes the disease (no
  recall bias)
• Multiple exposures and outcomes
• Not suitable for rare diseases
• Non-participation or lost to follow-up (selection
  bias)