RCTPres1.pptx

Randomization Randomization means: Everyone has the same known probability of receiving each of the possible treatments. A variety of methods may be used to achieve randomization, but one of the most common is the simple “coin toss.”  Advantages of Randomization Th...

Randomization Randomization means: Everyone has the same known probability of receiving each of the possible treatments. A variety of methods may be used to achieve randomization, but one of the most common is the simple “coin toss.”  Advantages of Randomization The assigned treatment group is unknown in advance to both subject and investigator (if allocation is concealed). Study groups tend to be comparable for all variables except the one being studied. Even unknown variables tend to balance out. If done well, randomization essentially eliminates selection bias. Lends confidence to the results of the trial.  The Perfect Study Identify a group of people Give them Wait 5 years, from population all Tx1 observe outcome at Time 0 Turn back time Start over with the same group Give them Wait 5 years, of people Tx2 observe outcome at Time 0 Compare Outcomes Estimate average Tx effect Tx1 vs. Tx2  As Good as it Gets: The RCT Identify a Randomly Give one group Tx1 group of people divide into and the other from population two groups group Tx2 at Time 0 Tx1 and Tx2 groups should be Wait 5 years, comparable, since randomization observe outcomes generally yields groups with in each group similar baseline characteristics. Compare Outcomes Estimate average Treatment effect Tx1 vs. Tx2  Applying The Intervention RCTs generally involve testing two or more different interventions. If there are systematic differences between the care provided to study groups beyond the assigned interventions, then Performance Bias may occur.  Assessment (1) Blinding (2) Reporting of Outcomes (3) Subject Accountability  Blinding Purpose: – Prevents Measurement and Reporting Bias. – Prevents Premature Drop-out. – Prevents unwanted co-interventions, i.e., helps reduce the possibility of performance bias.  Blinding Who is Blinded? – Study subject * – Person assessing the outcome * – Other study investigators – Statistician * Implied when using the phrase “Double Blinding”  Reporting of Outcomes All relevant outcomes for each group or intervention should be reported. Selective reporting of outcomes may lead to bias.  Subject Accountability All subjects who enter the study should be accounted for at study completion Selective loss of subjects and outcomes can bias the study results.  Analysis Intention to Treat Interpretation of the results Extrapolation of the results (external validity)  Intention to Treat Subjects are analyzed in the group to which they were randomized, regardless of the treatment they received.  Outcome Comparisons Main Analysis Intention-to-Treat – Preserves benefits of randomization – Generally, the preferred comparison – May underestimate the true effect of the intervention (i.e., biases towards the null) Treatment received – Often used for “efficacy” analysis – May bias the results  Study Power When a study reaches a negative or no difference conclusion, you should determine whether the study had enough POWER. Power is the ability for a study to detect a difference if, in fact, a true difference exists. Also termed the “Sensitivity” of a study  “True” State of Affairs Inference H0 False H0 True (study result) (there is a difference) (there is no difference) Type I Reject H0 OK POWER (There is a difference) (α) Accept H0 Type II (There is no difference) OK (β)  Extrapolation The study subjects should be similar enough to patients not in the study to allow you to generalize the study's results. The ability to extrapolate the study's results is often referred to as the Generalizability or External Validity of the study.  Extrapolation Do old men have big ears? Objective: To examine the association between age and ear size. Results: 206 male patients examined Mean age 53.8 yrs (30-93); Mean ear length 67.5 mm (52 - 84 mm)  Summary RCTs generally are considered the strongest type of epidemiologic study design. RCTs are the epidemiological equivalent of a basic science “experiment.” RCTs tend to be expensive and often are challenging to perform. Not all research questions are amenable to the RCT design.  Summary RCTs can provide the strongest and most direct evidence upon which to make a judgement about a cause-effect relationship. The generalizability of each study must be determined based on the inclusion and exclusion criteria Following the guidelines outlined in this lecture will help you to assess the validity of an RCT.