02: Study Design Flashcards
What is the importance of study design?
- data cannot directly answer questions of interest
- study design determines which data are collected/examined
- data require transformation via analysis and interpretation
- research questions we can ask depend crucially on the study design
What are the 3 overarching study design ‘families’?
- experimental studies
- observational studies
- meta-analysis studies
What are the 6 specific study designs?
- meta-analysis
- trials
- cohort
- case-control
- cross-sectional
- case report/case series
What is the unit of analysis?
What are the 2 types of meta-analyses?
- meta-analyses of experimental studies
- meta-analyses of observational studies
What do observational and experimental studies have in common?
can both make causal claims
What are observational studies?
interventions or exposures are not assigned or manipulated, just observed
- researchers observe what happens under different interventions/exposures
What are experimental studies?
interventions or exposures are assigned or manipulated
- generally considered higher quality evidence – less bias, higher confidence in causality claims
If there are ethical concerns, what study type is the only possible study type?
observational
- any exposure with evidence of harm
- smoking, asbestos, air pollution
- infectious disease status
If the exposure cannot be changed, what study type is the only possible study type?
observational
- race/ethnicity
- presence/absence of a genetic trait
- chronic disease state
What do all trials have in common?
interventions or exposures are assigned or manipulated by researchers
What are non-randomized trials?
treatment is assigned, but not at random
What are randomized trials?
treatment is assigned at random (most common)
- individuals meeting eligibility invited to enroll, randomized to a treatment or intervention, followed over time for outcome/endpoint occurrence
- expensive, time-consuming, not always possible
What is the idealized health research study design (gold standard)?
randomized trials
What do randomized trials require?
- pre-specified protocol – eligibility criteria, recruitment plan, randomization method, treatment/intervention and outcome or endpoint definitions, statistical analysis plan
- ‘equipoise’ – no evidence of harm or benefit for intervention
- blinding
What is blinding?
trial participants, treating providers, and statistical analysts do not know which patient received which treatment
What is a placebo controlled trial?
randomized trial where one group receives a treatment, while the other receives a placebo that looks like the treatment
What are the expected outcomes of randomized trials?
patients assigned to treatment A are expected to be comparable (exchangeable) with patients assigned to treatment B
- still susceptible to bias and flawed study design and implementation
What are observational studies with no controls?
- simply report outcomes of a single group
- ‘hypothesis-generating,’ anecdotal
- lowest quality evidence, cannot establish cause and effect between an exposure and outcome
- can draw attention to an important clinical question/concern
What are the types of observational studies with no controls?
- case series
- case report
What are case series?
> 1 individuals with common clinical feature
What are case reports
1 individual
What are the 2 types of observational studies with controls?
- followed over time
- no follow-up over time
What are the 2 types of observational studies with controls that are followed over time?
- cohort studies
- case-control studies
What are cohort studies?
select individuals from the source population into sample based on meeting cohort inclusion criteria (ie. not based on outcome status)
- followed over time for outcome status
What is assessed in cohort studies?
exposure status is assessed
- exposure status is not assigned/manipulated, but is based on real-world treatment
- to compare outcome frequency among exposed and unexposed, these groups must be exchangeable (comparable)
What are the advantages of cohort studies?
- does establish temporal sequence of events for causal inference
- can establish rates (events per person-time), incidence
- can evaluate multiple outcomes
What are the disadvantages of cohort studies?
- potential bias arising from differences between exposed and unexposed individuals (confounding)
- often requires a large sample, takes a long time to observe outcomes
- costly
- less efficient for rare outcome
What are case-control studies?
select individuals into sample based on outcome status
- outcome status is ‘starting point’ – determines selection into the sample
- look back in time for exposure frequency
What are cases in case-control studies?
individuals with the outcome of interest
What are controls in case-control studies?
individuals without the outcome of interest, comparable to cases
Is selecting cases in case-control studies hard?
no – usually straightforward
Is selecting controls in case-control studies hard?
yes
- selected from the same source population that gave rise to cases
- controls must be comparable to cases, similar with respect to potentially confounding factors
- selected independently of their exposure status – often selected via matching
What are the advantages of case-control studies?
- efficient for rare outcomes, outcomes that take a long time to develop
What are the disadvantages of case-control studies?
- potential bias arising from differences between exposed and unexposed individuals (confounding), control selection (difficult to select similar/comparable controls), exposure assessment occurring after the outcome (information bias)
- cannot examine multiple outcomes, cannot examine outcome rates or absolute risk/incidence
What is matching in case-control studies?
selection of controls to match specific characteristics of the cases
- can be 1:1 (cases:controls) or other ratio
What are the 2 types of matching in case-control studies?
- individual matching
- frequency matching
What is individual matching in case-control studies?
select a specific control per cases by matching variable (ie. date of birth)
What is frequency matching in case-control studies?
select controls to get the same distribution of variables as cases
What type of observational study with controls are not followed over time?
cross-sectional studies
What are cross-sectional studies?
exposure/intervention and outcomes are measured at the same time – ‘snap-shot’ view
What are the advantages of cross-sectional studies?
- simple to carry out, inexpensive
- can evaluate multiple outcomes
- can measure outcome prevalence (% with outcome at single time point)
What are the disadvantages of cross-sectional studies?
- does not establish the temporal sequence of events for causal inference
- cannot establish rates (events per person time) or incidence (proportion of population that develop an outcome)
Comparison of Time in Observational Studies
- cross-sectional study: present
- case-control study: present to past
- cohort study: present to future
What is the outdated use of retrospective vs. prospective?
- prospective: cohort and randomized trials
- retrospective: case-control
What is the key distinguishing factor of retrospective vs. prospective?
- prospective: exposure measurement cannot be influenced by the outcome
- retrospective: exposure measurement can be influenced by the outcome
What is the updated (better) definition of retrospective vs. prospective?
- prospective: exposure measured/collected prior to the outcome
- retrospective: exposure measured/collected at the same time/after as the outcome
Why do we care about different study designs?
- level of evidence depends on study design
- each study design is susceptible to different biases
List the evidence hierarchy based on quality of evidence, from highest to lowest.
increasing study quality, decreasing bias
- meta-analysis
- experimental studies
- cohort studies
- case-control studies
- cross-sectional studies
- case reports and case series
- expert opinion
What is the proposed new evidence-based medicine pyramid?
What is study registration?
involves submitting a study protocol to an online, public access, searchable registry
What type of studies is study registration available for?
both experimental and observational studies
What does study registration do?
- promotes transparency – prevents selective reporting, reduces publication bias
- prevents duplication – inefficient use of limited resources, patient time
- increases awareness for patients, healthcare providers
- condition for publication – in some cases required by law (ie. ClinicalTrials.gov)
What is a study protocol?
- complete description of PICO/PECO components
- complete description of the statistical analysis plan
What is a reporting guideline?
a simple, structured tool for health researchers to use while writing manuscripts
What do reporting guidelines do?
provides a minimum list of information needed to ensure a manuscript can be:
- understood by a reader
- replicated by a researcher
- used by a healthcare provider to make a clinical decision
- included in a systematic review
What are the reporting guidelines for randomized trials?
Consolidated Standards of Reporting Trials (CONSORT)
What are the reporting guidelines for the 3 types of observational studies?
Strengthening the Reporting of Observational Studies in Epidemiology (STROBE)
What are the reporting guidelines for meta-analyses?
Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA)