RCT 概述(overview).pdf

1  Randomized controlled trial (RCT)  Distinguish between randomisation and other ways of controlling for confounders  RCT types: basic parallel two arm RCT versus multiple intervention arms, factorial, crossover, as well the cluster randomised design, and n-of-1 trial  Superiority, equivalence and non-inferiority questions in trials  Pragmatic and explanatory attitudes to the design of RCTs 2  Confounding (randomization)  Bias (concealed allocation)  Placebo effect (placebo control group)  Chance (large sample) 3  Confounding variable is related to both the independent and the dependent variable (i.e. the exposure and the outcome) and is not considered an independent variable on the causal pathway.  If confounding variables are unequally distributed in the two exposure groups and are not taken into consideration in the design and/or analysis phase of research they can lead to spurious conclusions.  Randomization balances known and unknown confounders 4  A bias leads to systematic error  There are many types of bias in epidemiology  Selection bias: e.g. when the condition or risk factor under study itself makes people unavailable for the study. An often cited example of this is in studies that enroll participants from a workplace which can lead to a selection bias called the “healthy worker effect”  Concealed allocation (i.e. to intervention or control arm) can help prevent selection bias. 5  A concurrent control group is one chosen from the same population as the tes