Intro to meta analysis Flashcards
What is meta analysis?
- Statistical technique used to combine and analyse the results of multiple individual studies on a specific research or topic
- produces a quantitative summary estimate of the effect size, along with measures of stat significance and precision
- relies on statistical methods to calculate summary effect sizes and assess heterogeneity
Why should we use meta-analaysis?
- Increases stat power
- precise estimation
- generalizabilty
- detection of heterogeneity
- Identification of subgroup effects
- synthesis of conflicting evidence
- exploration of publication bias
How are meta analyses conducted?
- Formulate the research question
- Create a protocol
- search for relevant studies
- study selection
- data extraction
- quality assesment
- calculate effect sizes
- assess heterogeneity
- select a meta-analysis model
- conduct the meta analysis
- assess publication bias
- interpret the results
- conduct sensitivity analyses
- present the findings
What steps are requited when formulating a problem?
P- Population or participant
I- Intervention or exposure
C- Comparator
O- Outcome of interest
S- study design
How do we define the population?
- define disease or condition of interest
- have explicit criteria
- Try not to be too exclusive
- Identify the important demograpic characteristics
- Identify the setting of the studies
- Handling of studies that involve a subset of population or participants
How are interventions defined?
- What is the intervention and control?
- Does the intervention have variations ? (dose, mode of delivery, frequency of delivery, who delivers the intervention)
- Which variations should be included?
- How will you handle the studies where only one intervention is included?
- How will trials that combine the intervention with another intervention of interest be handled?
How to define types of outcomes
?
- include outcomes that are meaningful to:
- clinicians
- patients
- society
- Define how outcomes will be measured
- Adress the outcomes that relate the beneficial effects as well as adverse effects
- Limit the outcomes to a max of 3 which have impact on patient care
How should we define study types?
- choose randomised controlled trials
- best way to eliminate bias and cofounding
- Better and more conservative eefect of estimates especially when compared to retrospective studies
- May eliminate “publication bias” to a reasonable extent
What should a good search be?
- Thorough
- Reproducible
- Objective
- Non restrictive
What type of sources should you research?
+ Bibliographic databases e.g., PUBMED, EMBASE, MEDLINE
Citation indexes e.g., SCOPUS, SciSearch
+ Dissertations and thesis on specific databases e.g., ProQuest
+ Grey literature e.g., conference abstracts
Hand searching (may not contain relevant search terms though)
+ Trial registers (national and international)
How should you plan the search?
-Develop a search strategy using a set of keywords
- document the search strategy (reproducible)
- Understand the use of filters in databases
- understand the use of boolean operators
What types of data should be collected?
- Study design data
- Participants
- Interventions data
- Outcomes measures data
- Results
What are some sources of data?
- Individual patient data
- Correspondence with investigators
- Original reports
What is imprecision?
random error that can overcome by larger samples
What are sources of bias?
- selection bias
- analytical bias
- exclusion bias
- data extraction bias
- reporting bias
What are the types of reporting bias?
- Language bias
- Citation bias
- Location bias
- Outcome reporting bias
- Publication bias
- Time lag bias
what are effect measures?
- Dichotomous data: risk ratio, risk difference, odds ratio
- Continuous data : mean difference
- Ordinal data : proportional odds ratio
- Time of event data : Harvard ratio
What is the difference between risk and odds?
-Risk : the probabilty of getting the event. Ranges between 0 and 1
Risk ratio : ratio of risks in the two groups
risk difference: difference between the risks in the two groups
Odds: Ratio of the probability that an event will occur to the probaility that the event will not occur
- odds ratio : ratio of odds of an event in the two grouos
- Odds = risk / 1- risk
- Odds> risk
What is heterogeneity?
- ensures that the observed differences in effect between two studies become more than what would be expected by chance alone
- Different types of variations:
- clinical diversity
- methodological diversity
- stats diversity
- clinical and methodological diversity gives rise to statistical diversity
What should be included when reporting findings?
- characteristics of included study
- can make the result generalisable
- requires a thorough methodology
