Meta-Analysis Flashcards
How are Meta-Analysis conducted?
-Combining the results of underpowered
studies would have the effect of increasing power and reducing the risk of a type II error
-using mathematical techniques: combine the results from different trials to reach a conclusion
-Trials are brought together for a collective weighing of their results
-Individual effect sizes for each study are combined to form a common estimate
What are the advantages of Meta-Analysis?
-Increase power by increasing the sample size (combining sample size)
-Improve estimate of effect size
-Resolve uncertainty when conflicting results exist among many studies
-Improve the generalizability (external validity) of findings (f.e. one study is done in older patients, one in mid-aged, one in young -> together they cover different ages of the population or in different regions)
What are the criteria used to exclude studies from the Meta-Analysis?
-Objective of study
-Description of subjects
-Type of dependent variable
-Study design
-Method of analysis
How are the results oftentimes plotted to depict the results of all trials?
Forest Plot
-the diamond represents all data points of the trials -> significant or NOT?
What does the diamond of the Forest plot of a Meta-analysis represent?
-The top and the bottom of the diamond is directed to the value of the OR on the X-axis
-the width of the diamond represents the 95% confidence interval
What does the Cochran Q test for?
-Test for Heterogeneity
-we want to minimize the heterogeneity, we want similar studies
is the variability between the studies in the Meta-Analysis acceptable
Another test for Heterogenity
Chi^2: Chi square test
What is the purpose of Sensitivity Analysis?
determines if the final result is influenced by the inclusion of one or more studies
-one big study may cause BIAS by influencing the result of the whole meta-study -> if the meta-analysis is significant after excluding the big study the significance of the Meta-Analysis will be underlined
What is the Funnel Plot?
Checks for symmetry and balance between the distribution of the results of the studies in a “funnel” (like a normally shaped curve)
-lack of funnel shape indicates PUBLICATION BIAS -> the meta-analysis favors studies with preferable results
What is a Systematic Review?
-Highest level of evidence
-complete summary of current literature
-an author assesses the evidence for benefit or harm
-often uses meta-analysis to combine results of studies
What is a Survival Analysis
-studying patterns of change in outcome variables over time = Survival Analysis
-evaluates the ultimate outcome (endpoint: often death) and the time it takes to get there
-often in cardiovascular or cancer trials
What are the analytical models used to depict Survival curves?
-Kaplan-Meier survival curve
-Cox proportional hazard model
Kaplan Meier Survival Analysis
-everyone starts enrolled alive (at 100%) -> treatment group and control group (standard or placebo)
-stepwise design: whenever a portion of the group dies the line (step) goes down
-observe how both groups are diverse as the number of patients goes down (death) -> is there a significant difference? -> Cox proportional hazard
Cox proportional hazard
-tells if there is a statistical difference
When looking for statistical significance we assess the impact of several factors (treatment, disease severity, concomitant condition, Gender)
-uses a technique of proportional hazard regression (Cox) - to assess the effects of 2 or more variables on the time to an event
-calculates a weighting coefficient for each independent variable in a multivariable equation
Hazard Ratio
-Ratio of hazards; Hazard ratio HR
-HR = hazard in the treatment arm/hazard in the control arm
Definition Hazard Ratio
-HR is an instantenous event rate
-probability of an event occurring over a small fixed period of time
-TIME is incorporated into their statistics!
What is the difference between HR and RR?
RR doesn’t take time into account
How do HR curves look like compared to Survival curves?
HR curves start at 0 and patients are followed over time to see if they develop the disease (increase in patients affected = incidence)
-COX proportional hazard is also used here: what percent of patients hit the event (disease) at what period of time
-TIME to EVENT plot
-after time the curves diverse and the COX proportional hazard tells if they are statistically different
What does an HR of 1.5 mean?
-HR of 1.5: 50% more patients in the treatment group (numerator) are having an event compared to the control group denominator
-HR of 0.5: 50% fewer patients in the treatment group are having an event compared to the control group
What is a pooled effect?
A Term used in Meta-Analysis
-pooling the results of multiple studies