Meta analysis Flashcards Preview

Epidemiology and biostatistics > Meta analysis > Flashcards

Flashcards in Meta analysis Deck (11)
Loading flashcards...
1
Q

Meta analysis what

A

statistical technique for combining findings from independent studies
Trials provide a precise estimate of treatment effect, giving due weight to the size of different studies included

2
Q

What does the validity of meta analysis depend on

A

Quality of systematic review on which it is based
Coverage of all relevant studies
Use of appropriate methods taking care of heterogeneity, bias, the robustness of main findings using sensitivity analysis

3
Q

What are the essential steps for a systematic review

A
PICOS
Population
Intervention
Comparator
Outcome
Study Design
4
Q

Heterogeneity what, what types are there

A

Meta analyses should assess heterogeneity ‘the presence of variation in true effect sizes underlying different studies’
Clinical-variation in true treatment, risk factor effects in magnitude or direction
Statistical-clinical differences between studies, methodological differences between studies, unknown study characteristics

5
Q

Equation for measuring heterogeneity

A

I^2=100*((chi^2-df)/chi^2)

df=degrees of freedom

6
Q

How to deal with heterogeneity

A

Fixed effect model: assumes common effect and within-study variation only. Open to bias: narrow confidence intervals and may become inappropriate in presence of heterogeneity
Random effects model: underlying effects are allowed to vary between studies.
In clinical contexts, fixed effect model first adopted

7
Q

What is fixed effect model

A

One true effect size which underlies all studies in analysis. Differences in observed effects are due to sampling error.
standard choice for public health because of likely differences.

8
Q

What is random effects model

A

true effect could vary from study to study
effect size might be higher/lower in studies where: participants are older,, more educated, healthier, more intensive variant of an intervention is used
Different effect sizes underlying different studies: mixes of participants and in the implementation of interventions.

9
Q

Sources of bias in systematic reviews and meta analysis

A

publication bias-type of bias related to what research is likely to be published

10
Q

What do forest plots and funnel plots show

A

forest-heterogeneity/inconsisteny

funnel–help detect effect of study size and publication bias

11
Q

What can detect sources of heterogeneity

A

a planned sensitivity/subgroup analyses