Lecture 5 Flashcards
Systematic Review
- Overview of __ studies
- States 4 components
Overview of primary studies
States objectives, materials, methods, and results. Methods are explicit, transparent, and reproducible.
Also includes a protocol.
Why is a systematic review better than a journalistic review?
A journalistic review begins with the bias of the author- they will usually include their own work. They may also use selection (publication bias) to include studies that support their position.
Selection (publication) bias is a similar concept to
Confirmation bias.
Cochrane review
Type of systematic review.
Must be unbiased, reliable, up to date, and accessible.
Unlike journalistic reviews, ____ are required for systematic reviews.
References.
Evaluating systematic reviews: What checklists are used to assess methods?
CONSORT for RCT
PRISMA for systematic reviews and meta analysis
Chochrane Reviewers handbook
Evaluate if methods cause bias.
Sensitivity analysis
Exploration of possible results of the review if designed differently. Exclude/include things to see if outcomes change.
Negative study
Nothing statistically significant. Not published.
External validity assesses what?
If the results are generalizable to a particular population.
Compare overview, systematic review, and meta analysis.
Big circle: Overview
Systematic review inside with meta-analysis inside that.
Meta analysis
Will improve power.
Subcategory of systematic review.
Statistical/numerical only.
Results of several trials which all addressed the same question.
Why should you preform a meta-analysis prior to undertaking an interventional clinical trial?
Identifies if a question may already be answered.
What does each mean on a forest plot?
- Green box
- Horizontal line width
- Vertical line
- The size of squares are proportional to
- Diamond
- Width of diamond
- Mean
- CI
- Line of no effect
- To the weight each trial contributes to the analysis
- Pooled data for subgroup.
- CI
Forest plot: Overlap of horizontal line for any trial of the vertical line indicates
There is no real difference between groups
Forest plot: If no horizontal lines overlap the vertical line, what does that mean?
Real difference between groups is 95% likely. Always accept 5% error.
Homogeneity
Result of one trial compatible with the results of others
Can Infer on forest plot if many CI are overlapping
No direct test
Heterogeneity
Results of one trial are not comparable to the results of others. May be ale to infer on forest plot when there are few-no overlapping studies.
Chi square tests for heterogeneity
Chi square test
Stat of heterogeneity.
Detects whether or not there is greater variation between studies than expected by random probability.
Does NOT prove homogeneity if values are below.
DF
Trials - 1
What does it mean if Chi square value is above the DF
Trials were different (heterogeneity) but doesn’t tell you HOW they are different.
Inconsistency I^2
Assessment (not test) for heterogeneity. This converts chi square to a percentage of variability.
Relationship between heterogeneity and inconsistency
More hetero = more inconsistent
Do you want a low or high I^2 percentage?
Low.
0-40% is unimportant
30-60% is moderate
50-90 is substantial. Concerning.
75-100% is considerable. Shouldn’t publish. Too inconsistent.
Does a narrow CI indicate more or less power?
More power. Often results from small sample sizes.
