Investigating Incoherence Gives Insight: the case of anti-platelet therapy for stroke prevention

Authors: Dewilde S, Hawkins N
Published in: Journal of Clinical Epidemiology. 2012 Aug;65(8):835-45

Abstract

Objective: To identify confounding factors that may explain the incoherence between direct and indirect evidence in a published analysis comparing extended-release dipyridamole (ERDP) plus aspirin to clopidogrel for the reduction of stroke.

Study design and setting: An existing analysis was updated with new studies from a systematic literature review. Clinicians reviewed the studies for potential confounders. Network meta-analyses were conducted including or excluding potential confounders, and were estimated based on direct, indirect, or a combination of direct and indirect evidence. Model fit was compared using the residual deviance and the deviance information criterion (DIC); node splitting was used to test for incoherence between the networks.

Results: Six trials and one meta-analysis were identified; aspirin dosage was identified as a potential confounder. The odds ratio (OR) for stroke of aspirin plus ERDP vs. clopidogrel based on indirect evidence without aspirin dosage adjustment is 0.85 (0.68–1.05); when accounting for the aspirin dose–response relationship it is 0.96 (0.73–1.25); and the direct evidence based on PRoFESS resulted in an OR of 1.02 (0.93–1.12).

Conclusion: When analyzing networks of evidence, attention should be paid to identifying and adjusting for potentially confounding factors. Investigating rather than ignoring inconsistency in the data set leads to clearer insight into relative efficacy.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: