STUDY DESIGNS
Joaniter Nankabirwa-Wandera
MMED /MSc
Learning Objectives
โ€ข Classify study designs
โ€ข Define the unique features of the study
designs
โ€ข Discuss weaknesses and strengths
associated with each study design
โ€ข Discuss the most appropriate study design
to answer certain research questions
Determining a study design
โ€ข Identify topic of
interest
โ€ข State question of
interest
โ€ข State objectives of
the study
โ€ข Choose the best
design to answer
question
Topic
Research
question
Objectives
(Hypothesis)
Types of Study Designs
Observational
โ€ข -- Case report
โ€ข -- Case series
โ€ข -- Ecological Studies
โ€ข -- Cross-sectional
โ€ข -- Case-control
โ€ข -- Cohort
Experimental
โ€ข -- Randomized clinical trial
Overview of study designs
Observational studies
Case Report /Case Series
Case
reports
Case
series
Selective by
nature, providing
little information
May help identify
potential health
problems
Uses of case series
and reports
1 patient: Case Report
More patients: Case Series
Ecological Studies
Describe disease
occurrence on
population level
Evaluate an association
using the population -not
the individual -as the unit
of analysis.
The rates of disease are
examined in relation to
factors described on the
population level
Example: Association between
cancer and fat intake?
National Cancer Rate
High Low
High Low
National Diet Fat-Intake
Ecological Studies Key issues
Ecological fallacy: We do not know if the individuals who
have cancer were also the individuals with the high fat intake
โ€ข Explores correlations between group level exposure
and outcomes
โ€ข Unit of analysis: usually not individual, but clusters
(e.g. countries, counties, schools)
โ€ข Useful for generating hypothesis
โ€ข Cannot adjust well for confounding due to lack of
comparability (due to lack of data on all potential
covariates)
โ€ข Missing data is another concern
Example
Types of ecological studies
โ€ข Multiple group study
Compares disease rates among many regions during the
same period
โ€ข Time-trend studies
Comparison of disease rates over time in one population
โ€ข Mixed designs
๏ƒ˜Multiple groups + multiple time periods
Examples of time series
Uses of ecological studies
โ€ข Are appropriate designs when generating
โ€ข May be used for testing a new hypothesis
BUT have limited value due to in-built
design errors
โ€ข Only design possible when adequate
measurement of individual-level variables
is not possible/not ethical Example:
Holding off immunization
โ€ข When funds or time is limited to do
individual level studies
Advantages of ecological studies
โ€ข Are low cost and convenience
โ€ข Simple to analyze and present
โ€ข Often helpful for generating new hypotheses for
further research
โ€ข Aggregated data on exposure and health outcomes
often publicly available in state and national
databases
Disadvantages of ecological studies
โ€ข Cannot adjust well for confounding due to lack of
data on all potential covariates
โ€ข Measures of exposure are only a proxy based on the
average in the population
โ€ข Lack of available data on confounding factors.
โ€ข Missing data is another concern
โ€ข Prone to Ecological fallacy: We do not know if
grouped results apply on the individual level
Ecological fallacy
โ€ข Also called ecological inference fallacy or population
fallacy
โ€ข is an error where conclusions are inappropriately
inferred about individuals from the results of aggregate
data
โ€ข Example: If countries with more Protestants tend to
have higher suicide rates, then Protestants must be
more likely to commit suicide
โ€ข How can it be avoided: If individual data is not
available difficult to control for except through
modelling
Why ecological studies
โ€ข Low cost and convenience
โ€ข Some measurements cannot be made on individuals
โ€ข Ecologic effects are the main interest (at the
population level)
โ€ข Simplicity of analyses and presentation
โ€ข Often helpful for generating new hypotheses for
further research
Cross-sectional studies
โ€ข Exposure and disease are determined
simultaneously for each person ๏ƒ  a snapshot of
the population at one point in time
โ€ข No follow up in this design
โ€ข This can only indicate association, but cannot give
a time-sequence.
โ€ข Good for prevalence studies and testing
association
Strengths
โ€ข Fast
โ€ข Inexpensive
โ€ข May answer question of interest
โ€ข Used to determine both prevalence of outcomes and
exposures
โ€ข May be used to design future studies
โ€ข Start of cohort study
โ€ข May even be used for repeat cross-sectional analyses
Weaknesses
โ€ข Only prevalence cases, canโ€™t measure incidence
โ€ข Cannot determine whether exposure preceded
outcome (Outcome may precede exposure)
โ€ข Unable to establish causal relations โ€“ need for
longitudinal study designs
Case-Control study
Start: Cases and Control
Population
Cases
(with
disease)
Controls
(Without
disease)
Exposed
Non-
Exposed
Non-
Exposed
Exposed
Direction of inquiry
Present
Past
Case-Control study
โ€ข Compares cases to
controls in order find
out what factors are
associated and could
possibly have
contributed to the
problem
โ€ข Cases: Diseased
โ€ข Controls: Not diseased
โ€ข Association measure =
OR
Features of case control studies
โ€ข Less expensive than cohort studies and
experiments
โ€ข Small number of participants needed
โ€ข Bes t design for diseases with long latent period
โ€ข Very efficient for rare diseases
โ€ข Multiple etiologic factors can be studied at one time
โ€ข Cannot yield estimates of incidence or prevalence
โ€ข Riskier due to biases (especially recall bias)
Cohort studies
โ€ข A cohort is a group of subjects that are followed
over time starting at a specified point in time and
usually have something in common at the start
โ€ข Cohort study is where participants are selected
with respect to exposure
โ€ข Individuals exposed to a risk factor (study group)
are compared those not exposed (the control
group).
โ€ข Starts with measuring exposure, looks forward to
the outcome
Cohort contโ€ฆ.
Purpose
โ€ข To study incidence of
disease or other
conditions
โ€ข To study associations
between exposures
and outcomes (risk
factors,
determinants)
Measure association
= OR /RR
Nested Case-Control Study
Special case-control study and is within an ongoing
cohort study
advantage:
no recall bias, temporal relation between exposure and disease
possible, on selected group of persons more expensive test can be
performed
WHY OBSERVATIONAL STUDIES
โ€ข Usually the only option if the predictor is a
potentially harmful exposure or risk factor
โ€ข Even if the predictor is an intervention, an RCT may
not be feasible due to duration between exposure
and outcome
โ€ข Financial reasons: intervention studies can be
significantly more expensive than observational studie
โ€ข More intellectually interesting than RCTs?
.
END
OBSERVATIONAL STUDIES
Experimental
Designs
Clinical Trial: Definition
๏ƒ˜A clinical trial is a prospective study comparing the
effect of intervention(s) against a control
๏ƒ˜A research activity that involves the administration of
a test regimen
Note: A proper clinical trial
๏‚ง Always prospective
๏‚ง Must have 1 or more interventions
๏‚ง Must have a control group
Clinical Trial Objectives
โ€ข To asses the efficacy of a new intervention or drug
โ€ข To asses the effectiveness of a new intervention or
drug
โ€ข To asses the safety of a new intervention or drug
History of Clinical Trials
โ€ข Daniel 1:12-16. A diet experiment (Royal food and wine versus
vegetables and water) for 10 days
โ€ข 17th century: Lancaster (captain of a ship)
โ€“ experiment to examine the effect of lemon juice on scurvy for sailors.
โ€“ Ships with lemons were free of scurvy compared to ships without
lemons having scurvy.
โ€ข Lind (1753) โ€“ Study of 5 tx for scurvy in 10 pts (2/tx) plus 2 pts
without tx (control).
โ€“ First two pts given orange and lemons recovered quickly and was fit for
duty after 6 days, compared favorably with all other patients
โ€ข Smallpox experiment (1721) at the Newgate prison in Great
Britain. Voluntary inmates were inoculated and were free from
smallpox.
Key Elements of a Clinical Trial
โ€ข Selection of subjects
โ€ข Allocation of exposure
โ€ข Blinding
โ€ข Data collection
โ€ข Statistical issues
โ€ข Ethical considerations
Selection of Subjects
๏ฑPopulation at risk/Target population
๏‚ง ENTIRE group of individuals to generalize results
๏‚ง The intervention is intended to benefit this population
๏ฑAccessible population
๏‚ง Population to which the researcher has reasonable access
๏‚ง May be a subset of the target
๏ฑStudy population
๏‚ง Group of participants actually studied
๏‚ง Subset of population meeting eligibility criteria
๏‚ง Inclusion criteria โ€“ identifies the target population
๏‚ง Exclusion criteria โ€“ excludes people from the study, mainly for
safety reasons
Allocation of exposure
โ€ข Studies without controls โ€“
๏‚ง not a clinical trial although still an experiment
โ€ข Studies with controls: Types of controls/Comparison groups
๏‚ง Historical controls (compare with the past)
๏‚ง Simultaneous non-randomized controls (concurrent
group/quasi experiment)
๏‚ง Randomization - true controls
โ€ข Importance of controls
๏‚ง For comparisons
๏‚ง Allow to determine if outcome is caused by intervention vs
other factors
Randomized controlled Trials (RCT)
โ€ข Exposure assigned in a random way, participant have
equal chance of being in either control or intervention
group
โ€ข Balances the known and unknown risk factors for the
outcome under study
โ€“ eliminates confounding due to measured and unmeasured
factors
โ€ข Methods for randomization
๏‚ง Simple randomization (fish bowel method, tossing a coin)
๏‚ง Random number table
๏‚ง Computer-generated list of assignments
Blinding/ Masking
โ€ข The team assessing the outcome are not aware of
treatment assignment
โ€ข To prevent the biased assessment of the outcome
Single-blinded study
โ€ข Either pts or physicians are blinded to the tx
allocation
Double-blinded study
โ€ข Both pts and physicians are blinded to the tx
allocation
Analysis of trials
โ€ข Should always start with a baseline comparison of
groups
Note: Investigator cannot dictate what a participant does
in a clinical trial, may or may not comply with Rx
Intent-to-treat (ITT)
โ€“ Participants are analyzed according to the groups they were
randomized
โ€“ If 100 randomized to a treatment, all analysed whether took
it or not
Per-protocol analysis
โ€ข Analysis is only for participants who completed the treatment
Both analyses are recommended and compare
conclusions
Phases of Clinical Trials
โ€ข Phase I: clinical pharmacology and toxicity
โ€“ 1st experiment in human for new drug
โ€“ Primary concern: Safety
โ€“ Typically required 15-30 patients
โ€ข Phase II: Initial Assessment of Efficacy
โ€“ examine the efficacy and refine the safety
โ€“ Goal is to screen out ineffective drugs
โ€“ Has 30-100 participants
โ€ข Phase III: Full-scale Evaluation of Treatment Efficacy
โ€“ Compare new treatment with standard treatment
โ€“ Aim is to define the โ€˜bestโ€™ treatment
โ€“ Hundreds to thousands of participants
โ€ข Phase IV: Postmarking Surveillance
โ€“ monitoring the adverse effects
Ethical Principles in Trials
โ€ข Respect for Persons
โ€“ Voluntary, informed consent
โ€“ Protection of vulnerable populations
โ€“ Right to end participation in research at any time
โ€“ Right to safeguard integrity
โ€“ Protection of privacy and well-being
โ€ข Beneficence
โ€“ Non-malfeasance
โ€“ Benefits should outweigh cost/risks
โ€ข Justice
โ€“ Protection from physical, mental and emotional harm
โ€“ Fairness
ALL TRIALS MUST BE REGISTERED IN A CLINICAL TRIAL REGISTRY
Cross-over Design
Population
Sample
Intervention
Randomization
Placebo
Washout
Washout
Placebo
Intervention
Outcome Outcome
Factorial Design
Population
Sample
Int A and Int B
Int A and Pbo B
Pbo A and Int B
Pbo A and Pbo B
Outcome
Outcome
Outcome
Outcome
Experimental vs. Observational
Observational
โ€ข Predictor = exposure or
risk factor
โ€ข Usually biased
โ€ข Confounding usually an
issue
Experimental
โ€ข Predictor = intervention
(treatment or therapy)
โ€ขMinimize bias
โ€ขConfounding eliminated

STUDY DESIGNS.ppt

  • 1.
  • 2.
    Learning Objectives โ€ข Classifystudy designs โ€ข Define the unique features of the study designs โ€ข Discuss weaknesses and strengths associated with each study design โ€ข Discuss the most appropriate study design to answer certain research questions
  • 3.
    Determining a studydesign โ€ข Identify topic of interest โ€ข State question of interest โ€ข State objectives of the study โ€ข Choose the best design to answer question Topic Research question Objectives (Hypothesis)
  • 4.
    Types of StudyDesigns Observational โ€ข -- Case report โ€ข -- Case series โ€ข -- Ecological Studies โ€ข -- Cross-sectional โ€ข -- Case-control โ€ข -- Cohort Experimental โ€ข -- Randomized clinical trial
  • 6.
  • 7.
  • 8.
    Case Report /CaseSeries Case reports Case series Selective by nature, providing little information May help identify potential health problems Uses of case series and reports 1 patient: Case Report More patients: Case Series
  • 9.
    Ecological Studies Describe disease occurrenceon population level Evaluate an association using the population -not the individual -as the unit of analysis. The rates of disease are examined in relation to factors described on the population level Example: Association between cancer and fat intake? National Cancer Rate High Low High Low National Diet Fat-Intake
  • 10.
    Ecological Studies Keyissues Ecological fallacy: We do not know if the individuals who have cancer were also the individuals with the high fat intake โ€ข Explores correlations between group level exposure and outcomes โ€ข Unit of analysis: usually not individual, but clusters (e.g. countries, counties, schools) โ€ข Useful for generating hypothesis โ€ข Cannot adjust well for confounding due to lack of comparability (due to lack of data on all potential covariates) โ€ข Missing data is another concern
  • 11.
  • 12.
    Types of ecologicalstudies โ€ข Multiple group study Compares disease rates among many regions during the same period โ€ข Time-trend studies Comparison of disease rates over time in one population โ€ข Mixed designs ๏ƒ˜Multiple groups + multiple time periods
  • 14.
  • 15.
    Uses of ecologicalstudies โ€ข Are appropriate designs when generating โ€ข May be used for testing a new hypothesis BUT have limited value due to in-built design errors โ€ข Only design possible when adequate measurement of individual-level variables is not possible/not ethical Example: Holding off immunization โ€ข When funds or time is limited to do individual level studies
  • 16.
    Advantages of ecologicalstudies โ€ข Are low cost and convenience โ€ข Simple to analyze and present โ€ข Often helpful for generating new hypotheses for further research โ€ข Aggregated data on exposure and health outcomes often publicly available in state and national databases
  • 17.
    Disadvantages of ecologicalstudies โ€ข Cannot adjust well for confounding due to lack of data on all potential covariates โ€ข Measures of exposure are only a proxy based on the average in the population โ€ข Lack of available data on confounding factors. โ€ข Missing data is another concern โ€ข Prone to Ecological fallacy: We do not know if grouped results apply on the individual level
  • 18.
    Ecological fallacy โ€ข Alsocalled ecological inference fallacy or population fallacy โ€ข is an error where conclusions are inappropriately inferred about individuals from the results of aggregate data โ€ข Example: If countries with more Protestants tend to have higher suicide rates, then Protestants must be more likely to commit suicide โ€ข How can it be avoided: If individual data is not available difficult to control for except through modelling
  • 19.
    Why ecological studies โ€ขLow cost and convenience โ€ข Some measurements cannot be made on individuals โ€ข Ecologic effects are the main interest (at the population level) โ€ข Simplicity of analyses and presentation โ€ข Often helpful for generating new hypotheses for further research
  • 20.
    Cross-sectional studies โ€ข Exposureand disease are determined simultaneously for each person ๏ƒ  a snapshot of the population at one point in time โ€ข No follow up in this design โ€ข This can only indicate association, but cannot give a time-sequence. โ€ข Good for prevalence studies and testing association
  • 23.
    Strengths โ€ข Fast โ€ข Inexpensive โ€ขMay answer question of interest โ€ข Used to determine both prevalence of outcomes and exposures โ€ข May be used to design future studies โ€ข Start of cohort study โ€ข May even be used for repeat cross-sectional analyses
  • 24.
    Weaknesses โ€ข Only prevalencecases, canโ€™t measure incidence โ€ข Cannot determine whether exposure preceded outcome (Outcome may precede exposure) โ€ข Unable to establish causal relations โ€“ need for longitudinal study designs
  • 25.
    Case-Control study Start: Casesand Control Population Cases (with disease) Controls (Without disease) Exposed Non- Exposed Non- Exposed Exposed Direction of inquiry Present Past
  • 26.
    Case-Control study โ€ข Comparescases to controls in order find out what factors are associated and could possibly have contributed to the problem โ€ข Cases: Diseased โ€ข Controls: Not diseased โ€ข Association measure = OR
  • 27.
    Features of casecontrol studies โ€ข Less expensive than cohort studies and experiments โ€ข Small number of participants needed โ€ข Bes t design for diseases with long latent period โ€ข Very efficient for rare diseases โ€ข Multiple etiologic factors can be studied at one time โ€ข Cannot yield estimates of incidence or prevalence โ€ข Riskier due to biases (especially recall bias)
  • 28.
    Cohort studies โ€ข Acohort is a group of subjects that are followed over time starting at a specified point in time and usually have something in common at the start โ€ข Cohort study is where participants are selected with respect to exposure โ€ข Individuals exposed to a risk factor (study group) are compared those not exposed (the control group). โ€ข Starts with measuring exposure, looks forward to the outcome
  • 29.
    Cohort contโ€ฆ. Purpose โ€ข Tostudy incidence of disease or other conditions โ€ข To study associations between exposures and outcomes (risk factors, determinants) Measure association = OR /RR
  • 30.
    Nested Case-Control Study Specialcase-control study and is within an ongoing cohort study advantage: no recall bias, temporal relation between exposure and disease possible, on selected group of persons more expensive test can be performed
  • 31.
    WHY OBSERVATIONAL STUDIES โ€ขUsually the only option if the predictor is a potentially harmful exposure or risk factor โ€ข Even if the predictor is an intervention, an RCT may not be feasible due to duration between exposure and outcome โ€ข Financial reasons: intervention studies can be significantly more expensive than observational studie โ€ข More intellectually interesting than RCTs?
  • 32.
  • 33.
  • 34.
  • 35.
    Clinical Trial: Definition ๏ƒ˜Aclinical trial is a prospective study comparing the effect of intervention(s) against a control ๏ƒ˜A research activity that involves the administration of a test regimen Note: A proper clinical trial ๏‚ง Always prospective ๏‚ง Must have 1 or more interventions ๏‚ง Must have a control group
  • 37.
    Clinical Trial Objectives โ€ขTo asses the efficacy of a new intervention or drug โ€ข To asses the effectiveness of a new intervention or drug โ€ข To asses the safety of a new intervention or drug
  • 38.
    History of ClinicalTrials โ€ข Daniel 1:12-16. A diet experiment (Royal food and wine versus vegetables and water) for 10 days โ€ข 17th century: Lancaster (captain of a ship) โ€“ experiment to examine the effect of lemon juice on scurvy for sailors. โ€“ Ships with lemons were free of scurvy compared to ships without lemons having scurvy. โ€ข Lind (1753) โ€“ Study of 5 tx for scurvy in 10 pts (2/tx) plus 2 pts without tx (control). โ€“ First two pts given orange and lemons recovered quickly and was fit for duty after 6 days, compared favorably with all other patients โ€ข Smallpox experiment (1721) at the Newgate prison in Great Britain. Voluntary inmates were inoculated and were free from smallpox.
  • 39.
    Key Elements ofa Clinical Trial โ€ข Selection of subjects โ€ข Allocation of exposure โ€ข Blinding โ€ข Data collection โ€ข Statistical issues โ€ข Ethical considerations
  • 40.
    Selection of Subjects ๏ฑPopulationat risk/Target population ๏‚ง ENTIRE group of individuals to generalize results ๏‚ง The intervention is intended to benefit this population ๏ฑAccessible population ๏‚ง Population to which the researcher has reasonable access ๏‚ง May be a subset of the target ๏ฑStudy population ๏‚ง Group of participants actually studied ๏‚ง Subset of population meeting eligibility criteria ๏‚ง Inclusion criteria โ€“ identifies the target population ๏‚ง Exclusion criteria โ€“ excludes people from the study, mainly for safety reasons
  • 41.
    Allocation of exposure โ€ขStudies without controls โ€“ ๏‚ง not a clinical trial although still an experiment โ€ข Studies with controls: Types of controls/Comparison groups ๏‚ง Historical controls (compare with the past) ๏‚ง Simultaneous non-randomized controls (concurrent group/quasi experiment) ๏‚ง Randomization - true controls โ€ข Importance of controls ๏‚ง For comparisons ๏‚ง Allow to determine if outcome is caused by intervention vs other factors
  • 42.
    Randomized controlled Trials(RCT) โ€ข Exposure assigned in a random way, participant have equal chance of being in either control or intervention group โ€ข Balances the known and unknown risk factors for the outcome under study โ€“ eliminates confounding due to measured and unmeasured factors โ€ข Methods for randomization ๏‚ง Simple randomization (fish bowel method, tossing a coin) ๏‚ง Random number table ๏‚ง Computer-generated list of assignments
  • 43.
    Blinding/ Masking โ€ข Theteam assessing the outcome are not aware of treatment assignment โ€ข To prevent the biased assessment of the outcome Single-blinded study โ€ข Either pts or physicians are blinded to the tx allocation Double-blinded study โ€ข Both pts and physicians are blinded to the tx allocation
  • 44.
    Analysis of trials โ€ขShould always start with a baseline comparison of groups Note: Investigator cannot dictate what a participant does in a clinical trial, may or may not comply with Rx Intent-to-treat (ITT) โ€“ Participants are analyzed according to the groups they were randomized โ€“ If 100 randomized to a treatment, all analysed whether took it or not Per-protocol analysis โ€ข Analysis is only for participants who completed the treatment Both analyses are recommended and compare conclusions
  • 45.
    Phases of ClinicalTrials โ€ข Phase I: clinical pharmacology and toxicity โ€“ 1st experiment in human for new drug โ€“ Primary concern: Safety โ€“ Typically required 15-30 patients โ€ข Phase II: Initial Assessment of Efficacy โ€“ examine the efficacy and refine the safety โ€“ Goal is to screen out ineffective drugs โ€“ Has 30-100 participants โ€ข Phase III: Full-scale Evaluation of Treatment Efficacy โ€“ Compare new treatment with standard treatment โ€“ Aim is to define the โ€˜bestโ€™ treatment โ€“ Hundreds to thousands of participants โ€ข Phase IV: Postmarking Surveillance โ€“ monitoring the adverse effects
  • 46.
    Ethical Principles inTrials โ€ข Respect for Persons โ€“ Voluntary, informed consent โ€“ Protection of vulnerable populations โ€“ Right to end participation in research at any time โ€“ Right to safeguard integrity โ€“ Protection of privacy and well-being โ€ข Beneficence โ€“ Non-malfeasance โ€“ Benefits should outweigh cost/risks โ€ข Justice โ€“ Protection from physical, mental and emotional harm โ€“ Fairness ALL TRIALS MUST BE REGISTERED IN A CLINICAL TRIAL REGISTRY
  • 47.
  • 48.
    Factorial Design Population Sample Int Aand Int B Int A and Pbo B Pbo A and Int B Pbo A and Pbo B Outcome Outcome Outcome Outcome
  • 49.
    Experimental vs. Observational Observational โ€ขPredictor = exposure or risk factor โ€ข Usually biased โ€ข Confounding usually an issue Experimental โ€ข Predictor = intervention (treatment or therapy) โ€ขMinimize bias โ€ขConfounding eliminated