Causality and Study Designs

Question 1

Association

Answer 1

• Relationship between two or more variables

- Does not automatically mean that one variable is the cause of change in another

Question 2

Causation

Answer 2

-One event is the result of the occurrence of the other event

Question 3

Determining Causality

Answer 3

• Answer about the relationship between a possible cause/exposure (independent) and an outcome (dependent)
• Bradford-Hill’s criteria for causality are used to help decide if a variable contributed to an outcome
• Study designs differ to the degree that they meet Bradford-Hill criteria
• Higher level study designs provide stronger evidence for causality

Question 4

Temporality

Answer 4

-Cause must precede effect

Question 5

Reproducibility/Consistency

Answer 5

-Same results with different populations and study designs

Question 6

Strength

Answer 6

• Of association (or effect size)

- Bigger/stronger is better

Question 7

Specificity

Answer 7

• Nothing else causes this

- Nothing else is caused by this

Question 8

Dose-response/Biological Gradient

Answer 8

• Higher dose produces greater effect
• Lower dose produces lower effect
• Offending agent is discontinued and effect ceases

Question 9

Biological Plausibility

Answer 9

-Relationship makes biological sense

Question 10

Coherence

Answer 10

-Data doesn’t conflict with known facts

Question 11

Analogy

Answer 11

• Drug class effects

- Structural similarities

Question 12

Evidence Hierarchy for Study Designs

Answer 12

1. Meta Analyses
2. Systematic Reviews
3. Randomized Controlled Trials
4. Cohort studies
5. Case-control studies
6. Case reports/case series
7. Clinical textbooks/expert opinions
8. Animal Research

Question 13

Interpreting Evidence Hierarchy

Answer 13

• Observation vs experimental conditions
• Lower position: lower level evidence (expert opinion, case report, case series, cross-sectional)
• Moderate level: evidence associated with case-control and cohort studies
• Highest levels: done by RCT, systematic reviews, and meta-analyses

Question 14

Observational Studies

Answer 14

• Exposure not under investigator’s control; only observe
• Case reports/case series
• Cross-sectional
• Case-cohort
• Cohort

Question 15

Experimental Studies

Answer 15

• Investigators control the intervention

- Randomized controlled trial

Question 16

Case reports/Case Series

Answer 16

• Description of unique events in 1 patient (case report) or >= 2 patients (case series)
• Weakest forms of evidence: small patient numbers, no control group, many biases possible
• Good for: new conditions, novel treatments, adverse drug events, toxic exposures
• Should include literature review to provide context for the port and detailed information on the clinical course/follow-up of patients

Question 17

Cross-Sectional

Answer 17

• Descriptions of observations
• Larger sample sizes than case studies/series
• Can measure associations between variables (which came first/compounded?)
• Good for measuring prevalence in a population at a SINGLE time point
• Evidence quality dependent on data source

Question 18

Case-Control

Answer 18

• Start with patients who already have a condition/outcome and look back in time for etiologies/risk factors that could’ve causes that outcome
• Control group, contains similar people who didn’t experience that outcome
• Compare cases versus controls for differences in characteristics/exposures

Question 19

Case-Control Uses

Answer 19

• Strength of association is measured by Odds ratio
• CANNOT prove causality
• Good for investigating possible etiologies or risk factors for uncommon conditions

Question 20

Cohort

Answer 20

• Prospective/retrospective
• Already exposed to a suspected cause/risk factor and follow them forward to see outcome
• Large groups of individuals
• Can compare rates of an outcome for patients with and without the exposure (control)
• Relative risk reflects the strength of association
• Can demonstrate temporality
• Good for establishing risk factors for an outcome and calculating incidence

Question 21

Prospective

Answer 21

Start at the present time and follow into future

Question 22

Retrospective

Answer 22

Start at a past time and follow to the present

Question 23

Randomized Controlled Trial (RCT)

Answer 23

• Experimental design
• Investigators control the intervention and follow patients prospectively to measure outcome
• Decide who can/can’t participate
• Patients are randomized into experimental and controlled
• Minimize biases/confounders by study design
• Good for establishing a good standard’s efficacy
• Can demonstrate causality
• Has strength of association, temporality, and dose-response

Question 24

Evidence Hierarchy + Causality

Answer 24

• Ability to establish causation follows the evidence hierarchy
• Case-control shows association between suspected cause and outcome of interest
• Cohort shows association and establishes temporality
• RCT shows association, establishes temporality, AND demonstrates dose-response