Drug Lit Exam 2- Meta Analysis

Front 1

5 characteristics of narrative review (scope, search strat, criteria, quality, summary characteristics)

Back 1

•Broad in scope
•Search strategy and sources for included references are not usually specified
•Criteria for selection of articles referenced in the review are not provided
•Quality of appraisal of the referenced articles included in the review is variable
•Summary is often qualitative not quantitative

Front 2

narrative review- is it true primary lit?

Back 2

Not truly primary literature; similar to a textbook chapter

Front 3

narrative review - subjection to bias and why

Back 3

Subject to bias: selection of evidence and conclusions

Front 4

narrative review is useful for? (2)

Back 4

Useful for general review of a topic and for background information

May be useful in areas with limited or preliminary research

Front 5

narrative review reliability for clinical decisions

Back 5

Not generally reliable for clinical decisions

Front 6

Characteristics of Good Systematic Reviews (7) subject, review methodology, each study, abstraction, analysis, bias, subgroup/sensitivity

Back 6

•Subject is a specific well-formulated research question
•Review comprises explicit and reproducible steps
•Rigorous critical appraisal is performed on each study included in the analysis
•Unbiased abstraction of data from the studies
•Qualitative analysis of included studies
•Appropriate assessment of the risk of bias (including heterogeneity)
•Subgroup and sensitivity analysis prespecified

Front 7

necessary steps for a good systematic review (3)

Back 7

–Search strategy and databases searched specified in detail
–Selection of studies to include is based on explicit criteria
–Study exclusion/inclusion criteria are well defined

Front 8

•Quantitative summary of the data are carried out (Meta-analysis): when appropriate?

Back 8

–Calculation of the estimation of effect and confidence interval
–Note: Quantitative analysis is not always appropriate

Front 9

characteristics of review and title/abstract of a good systematic review

Back 9

•Review should be transparent to allow reader to assess strengths and weaknesses of the investigation
•Title and abstract should give a structured summary

Front 10

objectives of systematic review (4)

Back 10

–Summarize and integrate results from a number of individual studies
–Analyze the differences in the results among studies
–Evaluate effects in subsets of patients
–Generate hypothesis for further research

Front 11

systematic reviews can increase what 2 things?
can also use in lieu of RCT when?

Back 11

Avoid RCT when precluded by ethical issues
–Increase power by combining data from several studies
–Increase sample size when individual studies are too small to identify an effect

Front 12

meta-analysis definition (and relationship to systematic reviews)

Back 12

“The use of statistical techniques to integrate and summarize the results of included studies. Many systematic reviews contain meta-analysis, but not all.

Front 13

what actually is a meta analysis?

Back 13

•Are the statistics for a systematic review

Front 14

what does a meta analysis permit?

Back 14

Permit a quantitative approach for systematically assessing the results of previous research to detect a treatment effect previously not detected in individual studies

Front 15

What is an Appropriate Topic For Systematic Review and Meta-Analysis? (4)

Back 15

•Many similar studies exist but no consensus of opinion is possible from any single study.
•Data in subgroups may provide important data for individual decisions.
•Answers to key questions may be expected to change medical practice or help answer questions about treatment effects, adverse events, etc.
•A focused clinical question

Front 16

QUORUM Statement- what is it and why was it brought about, what was wrong with it

Back 16

Data suggest that key information is poorly reported in systematic reviews

•Checklist – preferred way to present the individual components of the meta-analysis

not enough emphasis on publication bias?? but still can use i think

Front 17

PRISMA- what is it

when released (year)

what does it consist of (3)

Back 17

•Replaced QUORUM statement
•Released July 21, 2009
•Checklist, flow diagram, elaboration document

Front 18

what does PICOS stand for

Back 18

•Patient Population/description of patient problem
•Potential Interventions
•Comparison to another intervention
•Relevant Outcome
•Study design chosen

Front 19

PICOS is useful for...

Back 19

Useful to structure the systematic review

Front 20

6 biases in systematic reviews

Back 20

–Publication Bias
–Selection Bias
–Statistical Heterogeneity
–Clinical Heterogeneity
–Sensitivity Analysis
–Subgroup vs. Sensitivity Analysis

Front 21

3 things to look for in a systematic review

Back 21

•Flow diagram, checklist (PRISMA)
•Structured abstract/title
•Risk of Bias

Front 22

publication bias

Back 22

Bias toward publication of positive results

Front 23

5 ways to minimize publication bias (like 5...sources)

Back 23

• Multiple databases should be searched
• Abstracts
• Bibliographies
• Unpublished data (contact FDA, pharmaceutical
companies)
• Contact experts in the field (dissertations, etc.)

Front 24

6 databases

Back 24

Medline,
EMBASE, CINAHL, Cochrane, AIDSLine, Index Medicus

Front 25

when listing sources of searches, what should you provide (2)

Back 25

give search dates and
specific search terms

Front 26

plot for checking publication bias

Back 26

funnel plot

Front 27

Funnel Plots-- what are they and what do they show

Back 27

• Scatter plots that graph treatment effects from individual studies on the horizontal axis against study sample size on the vertical axis.

Front 28

3 ways to present vertical axis of funnel plots

Back 28

sample size, log of sample size, or
the standard error of the effect estimate

Front 29

if there is little pub bias, what will a funnel plot look like

Back 29

looks like an inverted funnel (upside down funnel)

Front 30

statistical methods used to look for pub bias (2)

Back 30

Egger weighted regression method or Begg rank correlation method

Front 31

sensitivity of both stat tests for pub bias

which is better?

Back 31

–Sensitivity is low for both methods
–Egger may be slightly better

Front 32

% of systematic reviews published in Nov where pub bias was considered

Back 32

•36% systematic reviews published in November 2004 reported that publication bias was considered

Front 33

effect of therapy- what happens to it due to pub bias

Back 33

•Effect of therapy is often over estimated due to publication bias

Front 34

Review of antidepressant trials – estimates were how much larger if only published trials were analyzed?

Back 34

32% larger when only the published trials were analyzed – need to look hard for unpublished studies

Front 35

selection bias

Back 35

Studies selected for inclusion in the analysis influence results

Front 36

ways to avoid selection bias (3)

Back 36

Methodology for selection of studies should be determined prospectively
Criteria for inclusion/exclusion should be established prospectively
Flow diagram??

Front 37

how to avoid abstraction bias (how many reviewers, other shit (3 total))

Back 37

should use 2 independent reviewers; 3rd reviewer for consensus

–Data collection form (randomization, blinding, etc)
–Attempts to contact original study authors

Front 38

Multiple publication bias- what is it
how to avoid

Back 38

duplicate publications i think

–Description of how authors avoided it?

Front 39

English-language bias

Back 39

i assume this means only getting trials published in english

Front 40

“Baseline risk bias”

Back 40

---listen

Front 41

what is study heterogeneity

Back 41

Describe the degree of between-study variability in a group of studies

Front 42

how to avoid/detect study heterogeneity

Back 42

•There should be an assessment of consistency between studies

Front 43

issue with study heterogeneity, combining study results to draw conclusion

Back 43

•Not appropriate to combine the results from heterogeneous studies – creates bias

Front 44

meta-analysis process (2 steps)

Back 44

1) Calculate appropriate summary effect estimate and confidence interval for each study

2) Combine these statistics into a weighted average using an appropriate statistical model.

Front 45

4 typical summary stats used for meta analysis

Back 45

odds ratios, relative risk, difference in means, and effect size.

Front 46

weighted average

Back 46

Average where the results of some studies make a greater contribution to the total than others

Front 47

main determining factor for weight of each trial in calculating weighted average (2 possibilities)

Back 47

Sample size is the main determining factor in the weight for the trial, but event rate can be used as well

Front 48

in a weighted avg, variance is described by the...

Back 48

Variance is described by the confidence interval (the uncertainty of the estimate)

Front 49

Models used for calculation of summary estimates depend on

Back 49

the type of outcome (RR, OR)

Front 50

weighted average should have both...(2) wtf

Back 50

Should have both effect estimate and confidence interval

Front 51

Fixed-Effects Model

Back 51

Considers within-study variability - assumes that the true effect of treatment is the same in each study

Front 52

describe what the effect estimate and CI are telling us if we use fixed effects model

Back 52

Effect estimate resulting from this method of meta-analysis is the “fixed” treatment effect and the confidence interval describes how uncertain we are about that effect

Front 53

another name for fixed effects model

Back 53

Mantel-Haenszel Method

Front 54

Random-Effects Model

Back 54

Considers both between-study and within-study variability - assumes that the studies used are a random sample from the universe of all possible studies.

Front 55

random effects model- what are the effect estimates telling us

Back 55

–Meta-analysis estimates the mean and standard deviation of the different effects (there is no one “fixed” treatment effect)

Front 56

random effects model aka...

Back 56

DerSimion-Laird Method

Front 57

when would random effects/fixed effects models give similar results?

Back 57

if Homogeneity is present – both models give similar results

Front 58

random effects models vs. fixed effect models - differences (3)

Back 58

Random effects models are more conservative

they also generate a wider confidence interval

they are less likely to show a significant treatment effect compared to a fixed effects model

Front 59

which stat model to use? (2)

Back 59

•Most journals require the random-effects model to be used because it is more conservative
•Both models should be used; if differences in results arise, probably due to heterogeneity among the studies

Front 60

Statistical tests for heterogeneity (2)

Back 60

Q statistic
I2 statistic

Front 61

how do you interpret the Q statistic for heterogeneity (2)

Back 61

–A significant test (p < 0.1) indicates that there is heterogeneity

–A non-significant test does NOT mean that there is no heterogeneity (the tests have low power because number of studies is usually small, the non-significant test just means heterogeneity could not be found)

Front 62

Q statistic- power fluctuations

Back 62

Has low power when there are few studies and excessive power when there are many studies

Front 63

I2 statistic properties- compare to Q statistic

what does it quantify

interpretation

Back 63

•Considered better than Q
•Quantifies the amount of variation in results across studies beyond what is expected by chance
•Variable interpretations

Front 64

what if I2 = 0?

Back 64

I2 = 0 is assumed to mean that all variability in effect size estimate is due to sampling error within the studies

Front 65

I2 = 50
I2 = 25
I2 = 75

interpret these values in terms of heterogeneity

Back 65

• I2 = 50 means that total variability in effect
size is not caused by sampling error but by true heterogeneity between studies
(moderate heterogeneity)
• I2 = 25 lower heterogeneity
• I2 = 75 severe heterogeneity

Front 66

Clinical heterogeneity may result from differences in: (5)

Back 66

– patients
– treatments (drugs/doses used)
– duration of follow-up
– disease states treated
Also may be related to unknown or unmeasured
confounders

Front 67

things having to do with the patient that can affect clinical heterogeneity (4)

Back 67

• patient selection procedure
• baseline disease severity
• different age groups
• gender differences

Front 68

2 ways to deal with heterogeneity

Back 68

• Consider subgroup analyses
• Sensitivity analysis may be conducted if
questionable studies were included; results are strengthened if inclusion of these studies gives
similar results to exclusion of the studies

Front 69

4 types of subgroup analyses

Back 69

– age subgroups
– outcome measure subgroups
– dose subgroups
– disease stage subgroups

Front 70

when would a meta analysis not be appropriate ( in terms of heterogeneity)

Back 70

If heterogeneity is severe and subgroup analyses are not possible, a meta-analysis may not be
appropriate.

Front 71

meta analysis results- how are they displayed

Back 71

• Results of a meta-analysis typically are displayed in Forest plots

Front 72

what do forest plots display? (4)

Back 72

Forest plots display the summary statistic and confidence interval for each study, and the overall weighted average (pooled estimate) for all studies

• The plots sometimes display the relative weight
assigned to each study

Front 73

what to inspect/look at in forest plots

Back 73

First, visually inspect the plots for heterogeneity

Front 74

what should be displayed in the results of a meta analysis? (4)

Back 74

• Characteristics for which data were extracted should be presented:
• Study size
• Population, intervention, comparators,
outcomes, study design (PICOS)
• Funnel plots
• Flow diagrams

Front 75

4 things that systematic reviews can help do/provide

Back 75

•Can define the boundaries of what is known and help define a research hypothesis
•Can help resolve conflict
•Can create a more precise estimate of a treatment effect and effect size
•Can provide information regarding applicability of research results in specific subgroups of patients

Front 76

what do the results mean? like what...will they tell..you? (4)

Back 76

•Careful application - may be wrong
•Quality of the studies used (“Garbage in, Garbage out”)
•Should not change clinical practice – most of the time
•Limited application – only answer a specific question

listen to lecture

Front 77

idk if have to know paper examples

Back 77

---