We reported a few weeks back on two studies in the New England Journal of Medicine claiming to offer definitive proof that trasylol, the once-popular antifibrinolytic drug used in tons of cardiac procedures, significantly increased the risk of post-operative MI and renal failure. 

An even newer report by Pagano and colleagues in the Journal of Thoracic and Cardiovascular Surgery argues that the effect observed in the NEJM papers was pure artifact, and that the original authors did not sufficiently control for the fact that patients receiving Trasylol (aprotinin) were just more sick to begin with.

 Or, as they put it:

[Our] findings were at odds with the study by Mangano and colleagues. In that study aprotinin was used in patients with a more adverse risk profile, among whom there was a 21% greater incidence of congestive heart failure than in the control group, 37% more pulmonary disease, 32% more renal disease, 50% more carotid disease, and 2-fold more concomitant valvular disease, all factors well known to influence survival.

You can almost hear the sniping in their voices when they mention that study. Now, Mangano and colleagues did take some measures to control for these significant differences in patient groups, which they described in their paper as follows.

Selection bias not controlled by multivariable methods was assessed with a propensity-adjustment method. Using nonparsimonious logistic-regression modeling, we developed propensity scores for the use of any antifibrinolytic treatment (vs. no treatment), including 45 treatment-selection covariates, and propensity scores for specific treatments. Covariate interactions proved unnecessary for the balance of baseline characteristics. The discriminate power of the propensity scores was quantified by measurement of the receiver-operating-characteristic area (the C-index). Covariate adjustment was performed with the derived propensity scores and drug-indicator variables.

Unfortunately, we don’t speak Martian, and so we can’t really evaluate the sufficiency of these measures. But, needless to say, Pagano and colleagues were not impressed.

Two things do seem clear. One, if your last name ends in -ano, it is your destiny to write about aprotinin (anyone else notice that?). Two, there’s enough evidence on both sides of the argument to suggest that something is fishy about this drug, so even despite this compelling new research, it’s unlikely aprotinin will ever shed its bad reputation and come back to market.