Study Design (Key, Odd)
- In experiments the investigator interferes with the explanatory variable
to help determine its effect. Observational (non-experimental) studies are
Randomized = the treatment is assigned based on chance mechanisms.
Double-blinded = observers (e.g., clinicians) and study subjects (e.g., patients) do not know which intervention subjects are receiving.
Controlled = there is a treatment and control group.
- Statistics is a way of learning from data.
- Self-controls (e.g., "before-after studies") are unable to determine if changes are related to the intervention or would have occurred
- In intention-to-treat analysis, outcomes are considered regardless of whether study subjects were compliant with their treatment. In
efficacy analysis, only outcomes in subjects who complete regimens are considered. The advantages of intention-to-treat analysis
include: (1) it reflects the way the treatment will perform in real-world settings; (2) it guards against investigator selectively excluding
outcomes non-favorable to their ideas. (Comment: The US Food & Drug Administration requires submission of intention-to-treat AND
effectiveness analysis results.)
- The 4 Basics according to Dallal are:(1) The research question must be fully focused and clearly stated.
(2) The study must be feasible (by which he means you can measure what you intend to measure).
(3) Every facet of the protocol must be carefully considered and documented.
(4) Research has consequences.Comment: These basics mean to me that a research must take their work seriously, NOT overstate conclusions, NOT falsify
exaggerate, and be guided by a deep conviction of the worth and dignity of the advancement of knowledge. This places special
responsibilities upon the researcher.
- Longitudinal measurements are able to discern the sequence of events in individuals over time (even if information on various
variables is collected simultaneously). Cross-sectional measurements cannot reliably discern the sequence of events in individuals over
- Randomization balances the group with respect to outcome determinants factors other than the "treatment" being studied. Thus,
contributors to the outcome other than the study factor will be prevented from confounding results.