Meta analysis.

Question 1

What sort of review has increased objectivity?

Answer 1

Systematic.

Question 2

What sort of summary does a systematic review provide?

Answer 2

Qualitative only.

Question 3

What sort of bias can be introduced with a systematic review?

Answer 3

Publication bias.

Question 4

What does ‘meta analysis’ refer to?

Answer 4

The analysis of analysis.

Question 5

What does a meta analysis do increased compared to a systematic review?

Answer 5

Increase objectivity.

Question 6

What are the two main limitations of using a single study when looking at research evidence?

Answer 6

1. Poor study design and low numbers will result in false negative results.
2. Only a subset of the population will be studied.

Question 7

For studies to be combined what must the treatments used in each studies be?

Answer 7

Very similar.

Question 8

What three questions often drive a meta analysis?

Answer 8

1. What is the average effect of a treatment?
2. Who benefits most from the treatment?
3. In what situations will the treatment work?

Question 9

What is the purpose of producing a review protocol before conducting a meta analysis?

Answer 9

To indicate the methods used in advance that aim to minimize the systematic errors.

Question 10

The inclusion/ exclusion of studies in a meta analysis are defined by PICOS. What does PICOS stand for?

Answer 10

Patient/ population.
Intervention.
Comparison.
Outcome.
Study Design.

Question 11

What is PRISMA?

Answer 11

‘Transparent reporting of systematic reviews and Meta-Analyses’

Provides detailed guidelines regarding the preparation of Meta Analysis.

Question 12

Fill in the gaps regarding what is includes in a meta-analysis review protocol.

1. ______ selection.
2. ______ question.
3. Study_______ and ______ criteria (PICOS).
4. Identify ______ _______.
5. Select _______
6. ______ data.
7. ________ assessment
8. How will study results be _______?
9. How will results and _______ be presented?

Answer 12

1. Background.
2. Research.
3. Inclusion/ Exclusion.
4. Research evidence.
5. Studies.
6. Extract.
7. Quality.
8. Synthesised.
9. Conclusions.

Question 13

What is the individual unit used in the statistical analysis in a meta analysis?

Answer 13

A sample.

Question 14

What are the two types of goals in a meta-analysis

Answer 14

1. Synthetic goals.

2. Analytical goals.

Question 15

Describe what synthetic goals are in meta-analysis?

Answer 15

To estimate of an overall summary of the average effect across studies while increasing precision by obtaining data from multiple studies.

Question 16

Describe what analytical goals are in meta-analysis?

Answer 16

To identify and estimate differences between study specific effects (heterogeneity) between studies. To track down the source of this variation.

Question 17

If the endpoint of a meta-analysis is continuous what summary statistics should you calculate?

Answer 17

The differences in means and the standard error of this difference/ CI.

Question 18

If the endpoint of a meta-analysis is binary what summary statistics should you calculate?

Answer 18

The log (odds ratio) and its standard error.

Question 19

If the endpoint of a meta-analysis is a survival endpoint what summary statistics should you calculate?

Answer 19

The log (hazard ratio) and its standard error.

Question 20

What combined effect is shown in the diamond of forest plots?

Answer 20

The weighted average of the effects of each study.

Question 21

What is the weight assigned to each study proportional to?

Answer 21

The precision of the study.

Question 22

What is the equation to determine the weight given to a study in meta-analysis?

Answer 22

W=1/variance

Question 23

Why does the effect size vary between studies?

Answer 23

1. Random error.

2. True variation.

Question 24

What model determines the random error of a study?

Answer 24

Fixed effect model.

Question 25

What model determines how much true variation there is within a study?

Answer 25

Random effect model.

Question 26

What does the fixed effect model (used to determine random variation within a model) assume?

Answer 26

1. All studies in the meta anaylisis are drawn from the same population. 2. All factors that could influence the effect size are the same in all study populations. 3. Effect size is the same in all studies. Any variation in effect size between studies is hence due to random error.

Question 27

How does the random effect model differ from the fixed effect model?

Answer 27

It assumes that studies were drawn from different populations.

Question 28

What model determines the following? The combined effect is the estimate of one true effect size.

Answer 28

Fixed effect model.

Question 29

What model determines the following? The combined estimate of effect size is not an average of one value but an average o f a distribution of values.

Answer 29

Random effects model.

Question 30

If a fixed effect model gives rise to the best estimate of an assumed common treatment effect what does the random effects model do?

Answer 30

Provide a summary result of the average distribution of the of treatment effects across studies.

Question 31

When do you use the random effect model over the fixed effect model?

Answer 31

When there is evidence of heterogeneity across studies/ it is likely that heterogeneity is the case.

Question 32

You should decide if your analysis has heterogeneity between studies and use this decision to make a conclusive decision regarding whether you should use a fixed effect or random effect model. True or False?

Answer 32

False. It is good practice to present both and examine the difference between the two.

Question 33

How do you do a Cochran's Q test and why do you do one?

Answer 33

Square the deviations of each study estimate from the overall meta analysis estimate. This allows you to determine the weight of the contribution of each study .

Question 34

What distribution does Cochran's Q test follow?

Answer 34

Chi-squared distribution with df of n-1.

Question 35

What statistic test explains the % of variation across studies that can be explained by heterogeneity and not chance?

Answer 35

I² test.

Question 36

What are 7 sources of heterogeneity?

Answer 36

1. Differences in statistical analysis. 2. Differences in populations. 3. Differences in covariates. 4. Differences in design and implementation of combined studies. 5. Heterogeneity of a disease or its classification. 6. Variation of intervention/exposure measurement and/or definition across studies. 7. Bias.

Question 37

What are three methods to assess the source of heterogeneity?

Answer 37

1. Analyse different subgroups to see if the results differ. 2. Perform a sensitivity analysis. 3. Meta-regression methods.

Question 38

What does the following method describe? Exclude each study individually and examine the impact that removing each study has on the summary result.

Answer 38

Sensitivity analysis.

Question 39

Random effect models are always used when heterogeneity exists between samples in a meta analysis. True or false?

Answer 39

False. If too much exists you may not be able to pool the studies.

Question 40

Meta analysis increases the precision of estimates. What can they not eliminate in studies?

Answer 40

Bias.

Question 41

What type of scores can reduce bias in a meta analysis?

Answer 41

Quality scores.

Question 42

Name two ways in which you can improve study quality in a meta analysis.

Answer 42

1. Only include studies that meet a specified standard quality. Once this decision is made ignore the variation within studies. 2. Use objective quality-related study characteristics in regression or stratification.

Question 43

Using quality scores in a meta analysis are useful as they are easy to validate. True or false?

Answer 43

False.

Question 44

What tool can be used to assess risk of bias in clinical trials?

Answer 44

Cochrane Collaboration tool.

Question 45

What 6 forms of bias are mentioned in the Cochrane Collaboration tool?

Answer 45

1. Selection bias. 2. Performance bias. 3. Detection bias. 4. Attrition bias. 5. Reporter bias. 6. Other bias.

Question 46

What is attrition bias?

Answer 46

Bias caused by incomplete outcome data.

Question 47

What is detection bias?

Answer 47

Bias caused by insufficient blinding of participants and personnel.

Question 48

If only a subset of relevant data is available what type of bias is produced?

Answer 48

Publication bias.

Question 49

Access to only a limited amount of the available data can cause publication bias. What else can cause publication bias?

Answer 49

No reporting of null and non significant findings.

Question 50

What sort of plot can examine the likelihood of bias?

Answer 50

A funnel plot.

Question 51

What way round does the funnel have to be on the funnel plot for it to indicate no publication bias.

Answer 51

Upside down.

Question 52

What is plotted on the x and y axis on a funnel plot?

Answer 52

X- Odds ratio of treatment effect (log scale). | Y- Sample size.

Question 53

What doe the centre of the funnel represent on a funnel plot?

Answer 53

The mean effect. The funnel should be symmetric around this.

Question 54

What two things dosystematic literature reviews help you do?

Answer 54

1. Planning new research. | 2. Evidence based medicine to set policy..