Only specialized statisticians discussed indirect comparisons in the past but over the years the topic developed into something, every statistician should know about.
In this episode, Benjamin and I talk about the important reasons for using indirect comparison (IC). We specifically address the following points:
- Reasons for IC
- H2H study design
- HTA assessment
- Regulatory discussions to inform the benefit-risk perspective
- Guideline development
- Clinical decision making
- Bucher,
- The classical Bucher approach vs matching adjusted indirect comparisons (MAIC)
- How to incorporated meta-analyses
- Different network-meta-analyses approaches (NMA): Bayes vs Frequentist
- systematic literature reviews (SLR)
- Data extraction sheet
- The iterative process of analyses
- Cochrane handbook
- Tools
- Visualizations
- Funnel plot – publication bias
- Forest plots – heterogeneity
- Inconsistency assessments – only if H2H also available
- Bias
- Different study designs
- Different populations
- Not exactly the same bridge comparator
- Differing assessments
- Different time points
- Multiple time points
- Pooling of doses
- Different analyses methods
- Precision vs bias
- Pre-specified vs post-hoc
- Secondary vs primary endpoints
- Power of IC
- Publish detailed analyses
Further references:
PRISMA http://prisma-statement.org/PRISMAStatement/
Earlier podcast episode:
Network meta-analyses: why, what, and how
Listen to this episode and know more about Indirect Comparison now!
Never miss an episode of The Effective Statistician
Join hundreds of your peers and subscribe to get our latest updates by email!
Get the
Hi Alex,
regarding the rule of thumb of needing 4 times as large a sample size for an indirect comparison to obtain the same precision as a head-to-head comparison, you might have read that here (towards the end of section 2.1): https://journals.sagepub.com/doi/10.1177/0962280207080643
Kind regards
Markus
Thanks Markus!!
Hi Alexander,
Great episode!
The rule of thumb can be “felt” considering that the variance of the indirect comparison is the sum of the variance of each individual trial. For example:
• Let’s consider 2 trials, trial 1: A vs Placebo and trial 2: B vs placebo with a variance V and N patients enrolled in each (to simplify let’s assume that the 2 trials have the same sample size and variance)
• The variance of the indirect comparison is 2*V (Bucher formula)
• To divide by 2 the variance of the indirect comparison, you need to multiply by 2 the sample size of each trial, so that the variance in each trial would become V/2
• That’s why to compare A vs B indirectly, you need 2*N patients in both trials; i.e 4*N patients in total; compared to N patients that you would have needed if you had compared A vs B directly
Kind regards,
Vanessa
Awesome easy explanation!