[Skip to Content]
Access to paid content on this site is currently suspended due to excessive activity being detected from your IP address Please contact the publisher to request reinstatement.
[Skip to Content Landing]
September 20, 1995

Bayesian Analysis and the GUSTO Trial

Author Affiliations

St Michael's Hospital Toronto, Ontario

JAMA. 1995;274(11):873-874. doi:10.1001/jama.1995.03530110035029

To the Editor.  —Drs Brophy and Joseph1 suggest that GUSTO2 does not directly address which thrombolytic strategy is superior and that the large sample size used enabled a statistically significant but not clinically meaningful difference to be detected. Using a Bayesian analysis, they conclude that the clinical superiority of t-PA over SK remains uncertain. The rationale and interpretation of their Bayesian approach are greatly flawed for the following reasons:

  1. Thrombolytic treatment (vs placebo) has resulted in a 2.6% absolute reduction in mortality; an additional 1.0% achieved with a better thrombolytic strategy in GUSTO is impressive. This a priori definition of a clinically meaningful difference, which Brophy and Joseph label as "arbitrary," was also consistent with the design goals and sample sizes of the Gruppo Italiano per lo Studio della Streptochinasi nell'Infarto Miocardico (GISSI-2) trial3 and the Third International Study of Infarct Survival (ISIS-3) trial.4

  2. The three trials

First Page Preview View Large
First page PDF preview
First page PDF preview