NAMA7903 Session 1

Study designs
21 Feb 2018 WJ Steinberg
 Quantitative Research

 Specific measurable
questions/characteristics
 Generates numbers
 Large representative samples
 Quantifies known
 Qualitative Research

 Broad questions, complex, in

depth, exploratory questions
 Generates words, concepts
 A few purposively selected
participants
 Later tested quantitatively
 Basic study designs
Quantitative (chapter 7 new book)

1) Descriptive (describes)
2) Analytic (explains)
a) case-control
b) cohort
c) cross sectional
3) Experimental Studies (evaluates
intervention)
a) randomized controlled (clinical) trial
b) community trial
 Choose the correct
study design for the job
Descriptive study design

Study Population

Study Sample

Measurement
Descriptive study design

Describes the distribution of disease

or risk factors or characteristics in a
certain population

Person, Place and Time

Audit - common
Incidence vs Prevalence

Incident cases:
Number of new cases during a specific
period (for example 1 year)

Prevalent cases:
Number of cases of a given disease at
a given time
 Exercise 1

 Formulate a good research

question that could be
answered in a descriptive study.
 Possible answers to
Exercise 1

 What is the prevalence of

smoking among registrars at
UFS?
 What is the age distribution of
diabetes at UAH?
 How many drugs per patient do
doctors at Pelonomi Hospital
prescribe?
Descriptive study design:
Issues to be considered

Study sample: Adequate

selection? Adequate size? Clear
inclusion and exclusion criteria?

Measurement: Reliable, well

defined, adequate response rate?
 Question 1

 In a community survey it was found

that 1000 of the 5000 children have
high lead levels in their blood

 800 of these 1000 children live on

plots where the soil is sandy, only
200 on plots where the soil is clay

 Does the sand cause the high lead

levels?
 Cause and effect

How do we determine risk factors for

disease?
How do we know that smoking causes
lung cancer?
How do we determine whether a
treatment is harmful to patients?
How do we know that thalidomide
early in pregnancy causes
congenital abnormalities in the
baby?
 Associations

 If we have an outcome (disease)

and we suspect a cause /
exposure / agent / ⇛our studies
need to show evidence of
association – the stronger the
association the more
suggestive the causation
 Basic study designs
Quantitative (chapter 7 new book)

1) Descriptive (describes)
2) Analytic (explains)
a) case-control
b) cohort
c) cross sectional
3) Experimental Studies (evaluates
intervention)
a) randomized controlled (clinical) trial
b) community trial
 Analytic study design:

 Seek to discover the cause of a

disease/condition; to answer
the question “Why?”
 NOTE: Analytic studies
demonstrate association, from
which we infer possiblecause
 Use the word “prove” sparingly
 Analytic study design:

Analytical study (to associate

risk factor/exposure to
disease/outcome)
 Comparison group
 Presence of confounding variables:
variables associated with risk factor
and outcome
 Consider the Sequence of events
 Exercise 2

 Formulate a study question

which could be answered by an
analytic study?
 Possible answers to
Exercise 2

 Does smoking cause lung

cancer?

More correct?
 Is smoking associated with lung
cancer?
 Types of analytic study
designs

 Case-control (retrospective –
starts at endpoint dx- works back
to exposure)
 Cohort (prospective – starts with
exposure – observes for outcome)
 Cross-sectional ( exposure and
outcome at one point in time)
 Case - control Study

Study Sample
All registrars that specialised at UAH in the last 30 yrs

Categorize by outcome Y (lung cancer)

Lung Cancer (cases) No Lung Cancer (controls)

Assess all for exposure X (smoking)

Is there a difference in exposure within the groups?
 Case - control Study :

Select a group of cases (persons with the

outcome) and a group of controls
without the outcome from the same
source population

Compare these groups regarding previous

exposure to various factors
 Study Sample

All registrars that specialised at UAH in the last 30 yrs

Categorize by outcome Y (lung cancer)

Lung Cancer (cases) No Lung Cancer (controls)

Assess all for exposure X (smoking)

Is there a difference in exposure within the groups?
 Case - control Study :

Choose control so as to eliminate

effect of possible confounders:
 Restrict study to a certain subgroup
 Individual matching: select a matching
control for each case
 Retrospective in time: preceding risk
factor information collected from
records or interviews
 Case - control Study :

Cases: Clear definition

Hospital records, doctors’ records, death
certificates

Controls:
Ideally from the same source population
as cases, so that if they had been
cases they would have been included
as cases
 Case - control Study :

Community control: neighborhood

control

Hospital controls: people who do not

have disease which is associated
with the disease or exposures being
investigated e.g BBA’s
 Case - control Study :

Possible problems in case-control studies:

 Choice of appropriate controls

 Recall bias: cases and controls unable

to recall past exposures; cases may
recall past exposures better; cases may
deny previous habits
 Comparability of measurements made
 Case - control Study :
Issues to be considered

 Study Population: clearly defined,

strata? How generalizable will results
be to other populations
 Sampling: Adequate selection? Size?
Controls matched? Inclusion criteria
sorted?
 No contamination between the 2 groups
(i.e. mutually exclusive)
 Case - control Study :
Issues to be considered

 Assess the Histories of all the

participants with respect to exposure
X - How many in each group had
exposure to smoking? Measure item
e.g. pack yrs
 Adequate response rate?
 Measurement criteria well defined,
repeatable, valid and No biases?
 Cohort study

Study Sample
All registrars at UAH

Categorize by exposure X (smoking)

Exposed Not Exposed

Assess all for outcome Y (lung cancer)

Is there a difference in outcome within the groups?
 Cohort study (follow –up)

Select a group without the outcome(s) of

interest but who has a certain exposure
and a group without the outcome(s) of
interest and without that exposure

Cohort: a group which shares a certain

characteristic e.g
Birth cohort: born at approximately the
same time
Work cohort: works at the same place
 Cohort study (follow –up)

 Prospective in time: first the exposure,

then the outcome

 Historical cohort analytical study

 Exposures are often various treatments

 Useful for rare risk factor, not rare

outcome
 Study Sample

All registrars at UAH

Categorize by exposure X (smoking)

Exposed Not Exposed

Assess all for outcome Y (lung cancer)

Is there a difference in outcome within the groups?
 Cohort study (follow –up)

Possible problems in follow-up studies

 Time needed to follow patients up

 Loss to follow-up

 Risk factor status can change over

time
 Comparability of the exposure groups

 Comparability of the measurements made

in the exposure groups
 Cohort Study :
Issues to be considered

 Study Population: clearly defined,

strata? How generalizable will results
be to other populations
 Sampling: Adequate selection? Size?
Controls matched? Inclusion criteria
sorted?
 No contamination between the 2
groups (i.e. mutually exclusive)
 Analytic studies
Exercise 3

 We could use two different

analytic studies to answer the
same question.
 What does each study require in
terms of cost, resources, time
etc.
 Which study would be more
feasible, and why?
 Answer

 A case-control study is usually

easier to execute. Start with
disease and go back into
patients’ history of exposure.
 Don’t need to wait for a long
time, cheaper,….
 Bias: “recall bias”
 Analytic studies
Exercise 4

 If you are trying to demonstrate

cause, not merely association,
which of these two designs
would be more suitable?
 Answer

 A cohort study would provide

you with firmer evidence,
although an intervention study
design is the strongest of all.
 Cross - sectional study

 A sample of the study

population is investigated and
information is collected on risk
factors (exposures) and disease
(outcome) at a point in time.
The information sought can be
current or past.
 Cross - sectional study

Select a sample of the group to be

investigated (irrespective of
exposure or outcome)
Collect information on risk factors
(for example smoking, alcohol use,
obesity) and diseases (for example
hypertension, diabetes) at a point
in time
 Cross - sectional study

 The study has a descriptive

component which enables
researchers to calculate the
prevalence of risk factors/ disease.
 The analytic part of the study
consists of comparing exposure
groups with respect to disease
presence.
 Cross - sectional study

Provides results regarding

 prevalence of risk factors and
outcomes
 associations between risk factors and
outcomes
Problems:
 Comparability of groups
 Causality cannot be concluded –
which came first, the risk factor or
outcome?
 Cross - sectional study

 Watch out for confounding

variables, recall bias
 Usually large sample sizes
required
 Cross - sectional study

 The total sample is divided into

four groups:
a) those exposed and diseased
b) those exposed and not diseased
c) those unexposed and diseased
d) those unexposed and not diseased
Table 7.1 p 81
 Break
 Basic study designs
Quantitative (chapter 7 new book)

1) Descriptive (describes)
2) Analytic (explains)
a) case-control
b) cohort
c) cross sectional
3) Experimental Studies (evaluates
intervention)
a) randomized controlled (clinical) trial
b) community trial
 Experimental study

 Intervention study, best way of

demonstrating causality, but
often not ethical to do so.
 Phrase a research question
which would need an
intervention study design.
 Answer

 Does smoking cause lung

cancer?

 Does drug A control

hypertension better than drug
B?
 Experimental study

Before-after study:
 One group
 Measurements made before and
after an intervention

 How sure can one be that change is

due to intervention?
 Experimental study

Community trial / Pragmatic trial:

Groups of individuals (eg schools,

clinics, GP’s practices, towns) are
randomised, not individual
participants, eg for health education,
monitoring of medication intake
 Experimental study

Study Population / Sample

Randomization

Rx: C
Rx: A Rx: B
Placebo

Assess all for outcome

Has sufficient time been given for outcome to develop?
 An early example of a trial

J Lind, 1753: treatments for scurvy

12 sailors, as similar as possible, on the
same diet
 2 received cider
 2 received vitriol
 2 received vinegar
 2 received sea water
 2 received oranges and lemons
 2 received nutmeg
 Experimental study

Experimental studies / Intervention

studies: evaluates the effect of an
intervention aimed at decreasing or
preventing morbidity or mortality,
increasing knowledge
 Randomised Controlled
Trials (RCTs)
RCT’s
 Compares different
medications/treatments, often a new
treatment and the standard treatment

 Compares medication/treatments to
placebo

Trial should not be conducted if it is already

known that one treatment is superior to
another
 Study Population / Sample

Randomization

Rx: C
Rx: A Rx: B
Placebo

Assess all for outcome

Has sufficient time been given for outcome to develop?
 Experimental study

Groups are followed for a set period

to determine
 Outcome (efficacy )
 Side-effects (safety)
During this time treatment must be
identical except for the interventions
being studied
 Experimental study

Contamination:
Participants in one group receive the
intervention of the other group

Co-intervention:
Participants receive some other
treatment in addition to the
intervention treatment
 Experimental study

Parallel design:
 different groups of participants receive the
different interventions

Cross-over trial:
 each participant receives all the interventions,
in a random order
 Washout period in between intervention, only
of use if participant’s condition after washout
returns to level before intervention (chronic
conditions)
 Experimental study

Ethics:
 Written informed consent
 Participant can withdraw
 Adverse events noted and reported to
for example ethics committee
 Under certain conditions participant
must be withdrawn, trial must stop
 Intervention study:
Issues to be considered

 Randomization, is it ethical, adequately

done
 Follow over time – sufficient time
allowed? (e.g. for side effects to
develop)
 Assess outcome, adequate response
rate? Measurement criteria well defined,
repeatable, valid?
 Are assessors “blind”?
 Randomisation

 Participants fulfilling inclusion criteria

and giving informed consent are
allocated at random according to a
random number table/computer
generated randomisation list
 Each participant thus has the same
chance of receiving any of the
treatments
 Ensures comparability of intervention
groups
 Randomisation

Block randomisation
 To ensure that at various stages of the
trial similar numbers are enrolled in
each group

Stratified randomisation
 To ensure that the groups are identical
regarding some prescribed variables
 Randomisation

 not alternate allocation

 not historical comparison
 not clinical judgment
 not patient’s choice
 not cool or nice or beautiful
 Blinding

 Participants and/or persons who

determine the outcome
(assessors) do not know in
which group the patient is, so
that this knowledge cannot
influence the evaluation/care
 Phases of Clinical Trials

Phase I trial
 Determining levels of drug in blood, absorption,
excretion, safety
 Usually done on healthy volunteers
Phase II trial
 First investigation in patients
 Usually a small group, often no comparison
group
Phase III trial
 Large randomised controlled clinical trials for
proof of efficacy and further investigation of
safety
 Strengths of study designs

Randomised controlled clinical trial

Not for a risk factor
Cohort study
Not for a rare disease
Not for long follow-up
Not if little is known about risk factors
Case-control study
Cross-sectional study - cannot deduce
causation
 Other types of evidence :

Systematic review
Meta-analyse
 Systematic reviews

 “A systematic review is a
literature review focused on a
single question that tries to
identify, appraise, select and
synthesize all high quality
research evidence relevant to
that question” (Wikipedia)
 Meta-Analysis

 “A meta-analysis combines the

results of several studies that
address a set of related
research hypotheses.”
(Wikipedia)
 Reviews

Types of questions:
 Frequency of a certain condition

 Diagnostic accuracy

 Risk factors/causation

 Prediction

 Intervention/treatment
 Reviews

Formulate the question of the

review in terms of:
 Population/participants

 Intervention or indicator

 Comparator (control) – optional

 Outcome

(PICO)
 PICO

 P Patient / Problem

 I Intervention

 C Control

 O Outcome
 Example

 Evidence of effectiveness (O) of

Herbal Medicinal Products (I) in
the treatment of Arthritis (P)
 Reviews

Identify all potentially eligible

studies:
 Data bases such as Medline, etc
publication bias?
 Grey literature: conference proceedings,
reports, theses, unpublished studies –
contact experts
 Registers (trial or disease)
 Reviews

Assess study quality:

 Methodology – sampling /
selection of
participants
measurement
confounding
 Reviews

Summarise and synthesise study

findings of selected studies:

Table: description of each study’s

sample size, setting, features,
results
Meta-analysis: statistical summary
of studies’ findings
 Reference

P Glasziou, L Irwig, C Bain, G

Colditz.
Systematic reviews in health
care. A practical guide.
Cambridge University Press
2001
 Reference

 Cochrane Collaboration –
reviews of effects of health care
interventions
www.cochrane.org

 Campbell Collaboration –
reviews of educational, social
and legal interventions
www.campbellcollaboration.org
 Systematic Review

study study study

criteria RCT’s

Summary of RCT’s
study---------outcome
study---------outcome
study---------outcome
 Meta-analysis

study study study

criteria RCT’s

Statistical pooling

Overall effect
 Levels of evidence AHCPR
Ia Evidence from meta-analysis of randomised
controlled trials

Ib Evidence from at least one randomised controlled

trial

IIa Evidence from at least one well-designed controlled

study without randomisation

IIb Evidence from at least one other type of

well-designed quasi-experimental study

III Evidence from well-designed non-experimental

descriptive studies, such as comparative studies,
correlation studies and case-control studies

IV Evidence from expert committee reports or opinions

and/or clinical experiences of respected authorities
 EBM: HAYNES 1996

CLINICAL STATE
CIRCUMSTANCES

CLINICAL EXPERTISE

PATIENTS’ RESEARCH
PREFERENCES EVIDENCE
AND ACTIONS
 Basic study designs
(chapter 7 new book)

1) Descriptive (describes)
2) Analytic (explains)
a) case-control
b) cohort
c) cross sectional
3) Experimental Studies (evaluates
intervention)
a) randomized controlled (clinical) trial
b) community trial
 Task for Session 1

Tasks - find 2 recent (2017/8)

full-length research articles:
 A randomised controlled trial

 An analytical study: cross-

sectional/case-control/cohort analytical
(NOT review articles or meta
analyses/protocols)
 Task for Session 1

Answer the questions of Task 1

Attach copies of the articles
(photo-copied hard copies or
electronically in pdf format) to
your answers
Mrs M Du Randt
7 March 2018