Module 7 Flashcards
RCT
- individuals are located at random to receive one of a number of interventions
- experimental study
- often comparative study
Types of Designs
- Historical controls
- Non-randomised concurrent control
- Quasi-randomised design
Historic Controls
Compare results of new treatment on new patients to previous results of a historical group who received standard treatment
Non-randomised concurrent control
Two groups, each receiving different treatment roughly at the same time.
Quasi-randomised design
Allocation to intervention is not truly random. Eg. allocation by order of enrolment
Purpose of random allocation
‘gold standard’ evidence
eliminates bias in treatment assignment
covariates are equally distributed across groups at baseline
experimental and control groups treated the same
facilitates blinding of treatments from investigators, assessors, pts
Types of RCT
- Parallel trials
- Crossover RCT - must have washout period
- Factorial RCT
RCT sources of bias
SELECTION BIAS
- inadequate generation of randomisation sequence
- inadequate concealment of allocation
PERFORMANCE/DETECTION BIAS
- inadequate blinding
ATTRITION BIAS
- excluding participants or significant attrition
REPORTING BIAS
- analysing participants in the wrong group
- selective reporting of findings
Intention to treat analysis
Compares treatment groups as originally allocated regardless of whether participants received or adhered to treatment
- promotes external validity - aims to evaluate effectiveness of an intervention in routine practice
Per protocol analysis
Compares treatment groups as originally allocated but includes only those patients who completed the treatment protocol, compromises internal validity
Repeated Measures analysis of variance (RMANOVA)
- assume everyone is measured at the same time and equally spaced time intervals
- require restrictive assumptions about the correlation structure
- does not provide parameter estimates
- can not handle time dependent covariates (predictors measured over time)
Longitudinal Data Analysis
- assess changes in response over time
- measure temporal patterns of response to treatment
- identify factors that influence changes
- include time-varying predictors in the model
- investigate causality
- better handling of missing data
Types of Longitudinal Data Analysis
Mixed Effects model
- compare individual changes over time (trajectories)
- natural history
Marginal models (eg. GEE)
- compare populations over time
- evaluate interventions or inform public policies
Generalized Estimating Equations (GEE)
- extension of the general linear model of statistical regression for modelling clustered or correlated data
- offers robust estimates of standard errors to allow for clustering of observations
- produces consistent estimates of regression coefficients and their standard errors
- can deal with normal and non-normal outcome data
- useful when the aim is to investigate differences in population averaged responses
Correlation structures of GEEs
- takes into account within subject dependency of observations by specifying a priori ‘correlation structure’ for the repeated measures
- Correlation structure include: exchangeable, autoregressive, unstructured, independent
