Lecture 4 - Network Meta-analysis

Question 1

Relative treatment effect

Answer 1

Hazard ratio

Question 2

Clinical transitivity - 1.

Answer 2

Do treatments and populations tot up?
Could a mega-randomised trial with all included treatments be done?
Have first-line and second-line treatment trials in the same network been included?

Question 3

Clinical transitivity - 2.

Answer 3

Do the ‘nodes’ of the network hang together? (are placebos similar enough to be grouped together)
Are dosages of the same drug anticipated to have the same effect?

Question 4

Statistical transitivity

Answer 4

Is the network balance across planks on the effect modifiers, including risk of bias indicators?
Do some comparisons have a higher than average percentage of women?
Where the network has close loops, do intervention effects appear to be consistent?

Question 5

Consistency

Answer 5

Are direct evidence and indirect evidence in conflict i.e. inconsistent?

Question 6

Inconsistent

Answer 6

Estimate of effect from direct evidence is statistically different from indirect evidence

Question 7

Testing loop-specific inconsistency

Answer 7

Compare direct estimate with indirect estimate using a basic z-test

Question 8

Overall test

Answer 8

Design-by-treatment interaction model

Question 9

Markov chain

Answer 9

Allowed interventions to be probabilistically ranked

Rankings account both for effectiveness of an intervention and the uncertainty around it

Question 10

Ranking of treatments

Answer 10

Matrix of intervention by ranking, with a specific probability attached to each possibility

Question 11

SUCRA

Answer 11

Surface Under the Cumulative Ranking

Consolidates matrix of intervention into a unified score to rank interventions

Question 12

Network connections

Answer 12

Each drug in network needs to be linked by at least one comparison
Network appears transitive
Where do you think may there be inconsistency

Question 13

Things to look out for when reading NMAs (usual things)

Answer 13

Do pairwise meta-analyses make sense?
What is the heterogeneity?
What is the quality of the included studies?
Is the outcome clear and commensurate across studies?

Question 14

Things to look out for when reading NMAs (NMA-specific things)

Answer 14

Do studies look like they belong in same network (How do you know and how did authors check)
How did authors evaluate transitivity?
How did authors evaluate consistency?
What are the pairwise comparisons between interventions? (What kind of pattern do they reveal)