Renal MDM Flashcards
What does a high P-value (>0.1, maybe >0.08) tell you if you are conducting a statistical test for heterogeneity
High P-value means that the null hypothesis (that the estimate of effect is the same of all studies) cannot be rejected.
This means that there is low heterogeneity among the studies in your systematic review => they can be combined in a meta-analysis
Yay!
What are the advantages and disadvantages of meta-analyses?
- Pros
- More powerful than individual studies alone
- More precise estimate of effect size
- Can detect complications and differences among subgroups
- Can more accurately find the point in time when the intervention is effective or harmful
- Cons
- If dissimilar studies are combined, the findings are less helpful
- Does not give information on the biology of dilsease or clinical application of evidence
What are the two types of modeling we can use in systematic review/meta analysis?
Which one is used more frequently?
Which one results in wider confidence intervals?
Fixed Effect Model and Random Effect Model
- Fixed effect model
- Assumes that included studies ask the exact same question
- Only works if studies are truly similar
- Narrower confidence intervals
- Random effect model
- Assumes that included studies ask a slightly different but closely-related question
- Results in wider confidence intervals
- Used more frequently
What is the random effect model for a systematic reveiw?
A systematic review/meta-analysis that assumes that the included studies ask the a slightly different, but closely-related question. Results can be combined, but generally there is more heterogeneity than in the fixed-effect model
Used more frequently
How do we know if the results of multiple studies can be combined in a meta-analysis?
Two ways:
- Expert opinion
- Statistical test for heterogeneity (calculate an I2 statistic)
- Low I2 => low heterogeneity => can combine
- High I2 => high heterogeniety => should not combine
What is a funnel plot? What can we learn from it?
A funnel plot is a visual way to detect bias in the selection of studies for systematic reviews.
- No publication bias
- Large trials are included, regardless of findings
- Comparative effectiveness is closely grouped around effect size
- Publication bias possible
- Small studies are only published if they are “positive”
What is a systematic review?
A rigorous review of the evidene about a specific clinical question
- Follows a scientifically-based plan
- Decided on in advance
- Excplicit about every step
- Readers can check validity of the plan and methods
- Addresses a single, focused question
- Good when mulitple studies disagree
What kind of graph is this?
What can we learn from it?
Forrest plot
- Number of studies included
- Pattern of effect sizes
- Overall effect sizes
- Differential weighting of differently-sized studies
- Measures of heterogeneity
What is fixed effect model for a systematic review?
- A systematic review/meta-analysis that assumes that the included studies ask the exact same question so there is only 1 underlying result in all studies
- Only works if the studies are truly similar
What is the best way to avoid publication bias when conducting a systematic review?
Include unpublised studies in the systematic review
What is the purpose of a “full text review” when conducting a systematic review?
- See if the study ACTUALLY addresses the clinical question
- Extract all key data
- Assess the risk of bias in each individual paper
Papers are thrown away if they are review articles (not original research), or if they do not have the right kind of study, long enough follow up, different patient population, different intervention, etc.
What is publication bias?
What causes it?
How can it be fixed?
Publication bias is the tendency for published studies to be different from all of the completed studies of a question
- Cause: published studies are more likely to be “positive” or show an effect from the treatment or intervention
- Fix: Include unpublished studies in your systematic reveiw or meta-analysis
What “additional data” should be included in a meta-analysis?
Unpublished studies!
Shows that the researchers/you are actively trying to overcome publication bias
What are the general steps for conducting a systematic review?
- Clinical question
- Search
- Abstract Review
- Full text Review
- Data abstraction, quality assessment
What is a meta-analysis?
A type of systematic review that (mathematically) pools the results of several individual studies
- Must be adequately similar to combine
What kind of graph is this?
What can we learn from it?
Funnel plot
Can show us the risk of publication bias
- Low risk of bias
- Large studies centered around true effect
- Studies that show negative and positive effect are published
- Higher risk of bias
- Small studies are published only if they show a “positive” effect
