This preview shows page 1. Sign up to view the full content.
Unformatted text preview: methods _randomization_large_edge.jpg 15 Stratified Randomization Stratified Randomization
Use when it is important to be certain that Use some variables, such as age, race or gender, are equally distributed Stratified randomization:
First preFirst pre-divide study group by the variable of interest interest then then randomize by treatment This equalizes the distribution of various risk This factors (e.g., history of previous disease, smoking, age, race) in study groups 16 3. Apply Intervention 3. Apply Intervention
Intervention Intervention should be well-defined, reproducible and welljustifiable. IDEAL placeboIDEAL = placebo-controlled, double-blind trial doubleMasking: Neither subjects nor investigators know who received intervention or control treatment Participants and investigators cannot be totally blinded for some interventions.
Example: Example: surgical treatment Use defined protocol for both experimental and Use defined protocol for both experimental and placebo placebo groups Verify Verify interventions occurred for both groups.
17 Apply and Assess Intervention Apply and Assess Intervention
Schedule regular visits to clinic or study Schedule regular visits to clinic or study site site Make participation easy Make participation easy
Examples: Examples: deliver pills, provide classes at work call at home friendly staff work, call at home, friendly staff Assess Assess compliance
Examples: use diaries, pill counts, biomarkers (urinary riboflavin, urinary cotinine levels) cotinine levels)
18 Did the Intervention Happen? Did the Intervention Happen?
Low compliance may hide a real benefit of the Low intervention (How (How can you expect benefit if you don’t have an intervention?) BUT Poor Poor compliance in study suggests even poorer compliance in the “real world” Inclusion “nonInclusion of “non-compliers” gives a better picture of the value of the intervention in the “real world”
19 4. Assess Outcome. 4. Assess Outcome.
Outcomes Outcomes must be clearly defined and systematically determined in both the treatment and control groups to avoid surveillance bias. Oft...
View Full Document
- Spring '08