Gordis: Chapter 7 and 8
Chapter 7:
Allocation of subjects to treatment groups
- alternatives to randomization
o studies without comparison
§ case study / case series
o studies with comparison
§ historical controls
· comparison group from the past
o quality of data collection
o can't be sure difference is due to treatment or other factors
· can be useful à disease is uniformly fatal
o decline in case fatality that parallels use of drug – strong support that drug has effect
§ simultaneous non randomized controls
§ randomization
· can't just match on important variables
o can only match on variables we know and can measure
o unknown variables
· stratified radomisation
o end up with 2 randomisaed groups, but having initially stratified the groups, we increase the likelihood the two groups will be comparable on stratified variables
- prognostic profile at entry
- masking / binding
- crossover
o each patient serves as his own control
§ repeated measures
- factorial design
- non compliance
o people who do not comply or who do not participate in studies differ from those who do comply and who do participate.
Chapter 8 : randomized Trials – Some Further Issues
- sample size
o difference in response rates to be detected
o estimate of the response rate in one of the groups
o level of statistical significance
o value of power
o one / two sided
Ways of Expressing Results of Randomised Trials
- Efficacy =
o ( rate for thos who received placebo – rate for those who received treatment ) / rate for those who received placebo
- ratio of the risks à relative risk
- number of patients who would need to be treated
o 1 / ( ratein untreated group – rate in treated group)
Generalisability of Results
- internal validity whether study was well done and whether the findings are valid
- external validity – generalisability
- problem of using intermediate measures as endpoints of effectiveness
o eg, because mortality takes years to manifest
o vs
o blood pressure / cholesterol levels
- publication bias
No comments:
Post a Comment