analyzing rcts
TRANSCRIPT
Analyzing RCTs
Arin BasuPrepared for
HLTH 460-46227-04-2010
Objectives
• Principles of RCT• Describe RCT• How to analyze results• Advantages of RCT• Limitations of RCT• Interpret RCT
What is an RCT?SELECT POPULATION
RANDOM ASSIGNMENT
INTERVENTION GROUPCONTROL GROUP
FOLLOW UP
COMPARE OUTCOMES FOR BOTH GROUPS
Begin with homogenous group
Individual with disease
Use randomization to split
RANDOMIZATION CAN BE DONE USING RANDOM NUMBERS TABLESOFTWARE PROGRAMME
Follow up to observe outcomes
Individuals with desired outcomes
Individuals with the disease
Comparison of results
Intervention/Control Outcome status
PRESENT ABSENT Total
INTERVENTION A B A+B
CONTROL C D C+DTotal A+C B+D A+B+C+D
Experiment Event Rate (EER) = (A / (A+B)) * 100Control Event Rate (CER) = (C / (C+D)) * 100Absolute Risk Reduction = EER - CERRelative Risk = EER/CER
Types of RCT• Parallel Group RCT• Factorial RCT• Crossover RCT• Cluster RCT• Trial types
– Superiority– Non-inferiority– Equivalence
• Pragmatic RCT vs Traditional RCT
Parallel group RCT
Initial allotment
INTERVENTION CONTROL
Final Evaluation
Diseased individuals
Individuals who got better with treatment
Factorial RCTTREATMENT X
INTERVENTION CONTROL
TREATMENT Y
INTERVENTION
CONTROL
Crossover RCT
Intervention Control
Washout Period
Trial Types based on aims
• Superiority trial– Is treatment X superior to placebo or another
treatment Y?• Non-inferiority trial
– Is treatment X not inferior or as good as another treatment Y?
• Equivalence trials– Is treatment X as good as or certainly as harmful
as no more harmful than another treatment Y?
Pragmatic RCT
• RCT tight selection criteria• Internal validity high generalizability low• Two approaches – not do RCT at all, or
observational studies• Pragmatic RCT is a good middle way• Relaxed inclusion criteria• Concealment of allocation• Appropriate for behavioural intervention studies
Cluster RCT
Clusters of individuals and unit of analysis are clusters
Why are RCTs good?
• Randomization: balancing• Control of chance by appropriate sample size
selection• Control of bias by blinding
– A process where the investigator is unaware of the allocation status (single blinding)
– When the participant is unaware of allocation status (double blinding)
• Control of confounding by randomization
Limitation of RCT
• Expensive in terms of time and money• Not generalizable• Low external validity• Otherwise, RCTs are good study designs
Example of an RCT
Cohort Study
• Observational epidemiological study• Individuals are selected on the basis of their
exposure• Exposed and non-exposed individuals are
followed up in time• Followed up to observe emergence of the
outcomes• Rates of outcomes compared
Steps of Cohort studyStudy Population
Exposure staus
Exposed Non-exposed
Rate of disease occurrence in exposed
Rate of disease occurrence in non-exposed
Type of Cohort study
• Prospective cohort study– Exposure occurs first followed by outcome but
outcome status not known at the time of exposure assignment
• Retrospective cohort study– Exposure precedes outcome– However, outcome status is known at the time of
exposure cohort assembly
Analysis of results
Exposure Outcome Total
Present Absent
Exposed A B A+B
Non exposed C D C+D
Total A+C B+D A+B+C+D
Rate of the disease among exposed = A/(A+B)Rate of disease or outcome among non-exposed = C / (C+D)Relative risk = Rate among exposed / Rate among non-exposed
Why are cohort studies good?
• Since sampling is done on exposure therefore easier to control for sample size therefore chance
• Always followed up, therefore presence of exposure before outcome is guaranteed
• Good for studying many outcomes
Limitations of cohort studies
• Risk of biased observation– Bias during cohort formation– Bias during recording of outcomes
• Role of uncontrolled for confounding is a problem
Summary
• RCTs and cohort studies are two important study designs in epidemiology and health research
• RCTs can be extended to studies beyond health care
• Pragmatic RCTs enable complex interventions to be studied
• Cohort studies are best study designs among observational studies
Example of cohort study