8

Background: I am working on a systematic review including several imaging modalities for coronary artery disease, but the evidence network is quite large, including different modalities, often compared one to the other in an ample network.

Network meta-analysis is an established approach for randomized controlled trials, with several potential approaches available in WinBUGS, Stata, R, and SAS.

However, I am not aware of the possibility to conduct network meta-analysis of diagnostic test accuracy studies.

Question: Is there any meaningfully robust approach to conduct a network meta-analysis of diagnostic test accuracy studies?

Attempt: In my opinion we could possibly use the diagnostic odds ratio (DOR) as effect estimate, and then pool it with standard techniques within an evidence network framework, for instance using the netmeta R package or similar approaches. (See: Which is the best method for network meta-analysis?).

UPDATE: Upon feedback from GGA and extensive search, we can mention as potentially suitable approaches: a Bayesian method proposed by Menten and Lesaffre to conduct a Bayesian network meta-analysis of diagnostic test accuracy studies (Menten and Lesaffre, BMC Med Res Methodol 2015), and two different Bayesian methods proposed by Nyaga et al (Nyaga et al, Stat Methods Med Res 2016; Nyaga et al, Stat Methods Med Res 2016).

Community
  • 1
Giuseppe Biondi-Zoccai
  • 2,244
  • 3
  • 19
  • 48
  • 1
    Is there an established gold standard for these studies? – AdamO Dec 06 '16 at 14:57
  • @AdamO: You have several good resources for network meta-analysis of randomized clinical trials (eg https://www.amazon.com/Network-Meta-Analysis-Comparison-Procedures-Technology/dp/1633210014) and several good resources for meta-analysis of diagnostic test accuracy studies (eg https://www.amazon.com/Statistical-Diagnostic-Medicine-Probability-Statistics-ebook/dp/B00MYZISXA). However, as far as I am aware there is no example of network meta-analysis of diagnostic test accuracy studies. – Giuseppe Biondi-Zoccai Dec 06 '16 at 15:40
  • 1
    I don't know the robustness of these methods since I'm not an expert statisticians, but here there are some references, https://www.ncbi.nlm.nih.gov/pubmed/27655805, http://journals.sagepub.com/doi/10.1177/0962280216682532, in alternative there's a manual on the cochrane dta methods groups where it explains how to do meta-analysis comparing multiple diagnostic tests using a glm approach – GGA Feb 01 '17 at 09:08
  • 1
    Why not remove your edit and make it into an answer? You are allowed to answer your own question. You might have to add another couple of sentences as link-only answers are discouraged but if people see the question has been answered they are more likely to visit your question. – mdewey Feb 18 '17 at 12:06
  • @mdewey Will definitely do it shortly. I just need to study is somewhat more detail this thesis: https://repository.library.brown.edu/studio/item/bdr:674079/PDF/ – Giuseppe Biondi-Zoccai Feb 18 '17 at 13:09

1 Answers1

1

So this does not go unanswered here, based on comments from @GGA and from the OP's further research, are some references which the OP edited into the question as an update.

Upon feedback from GGA and extensive search, we can mention as potentially suitable approaches: a Bayesian method proposed by Menten and Lesaffre to conduct a Bayesian network meta-analysis of diagnostic test accuracy studies (Menten and Lesaffre, BMC Med Res Methodol 2015), and two different Bayesian methods proposed by Nyaga et al (Nyaga et al, Stat Methods Med Res 2016; Nyaga et al, Stat Methods Med Res 2016).

mdewey
  • 16,541
  • 22
  • 30
  • 57