Biotech Values

[jq1234]

NEJM Aug 2011 editorial pointed out some of these issues:

http://www.nejm.org/doi/full/10.1056/NEJMe1107516

One interesting point is how the interpretation of ROCKET AF depends on the results from the three different protocol-specified analyses of the primary outcome of stroke or systemic embolism. The primary analysis included only patients who were treated according to protocol and were followed for outcome events only for the period during which they were actually receiving the assigned treatment (or within 2 days after the last dose). This analysis, designated the “per-protocol, as-treated” analysis, resulted in a conservative test for noninferiority5 and showed that rivaroxaban was significantly noninferior to warfarin. The authors also tested for superiority using an intention-to-treat analysis, which did not show superiority for rivaroxaban over warfarin.

Between these two analyses, the authors conducted another analysis in the “as-treated safety population.” Here, they included all patients who received at least one dose of a study drug and were followed for events while receiving the drug (or within 2 days after the last dose), regardless of adherence to protocol. It is not surprising that the annual event rates in the two study groups in this analysis were much closer to those of the per-protocol analysis than to those in the intention-to-treat analysis. As a consequence, in the safety analysis, the between-group P value was significant, even though the results do not show superiority for rivaroxaban over warfarin, since the intention-to-treat analysis was negative. Thus, the multiple analyses have muddied the waters regarding rivaroxaban's efficacy and effectiveness over warfarin.

Trials comparing new oral antithrombotic agents with warfarin are dependent on the quality of the management of the warfarin cohort. Overall compliance varies. Trials like ROCKET AF and RE-LY use algorithms for imputing the dummy INR of the warfarin placebo in patients who are not receiving warfarin in order to maintain blinding. They also use the concept of “time in therapeutic range” to assess the quality of warfarin management. Such trials typically use a method described by Rosendaal et al.,6 in which the measured INR values and the interval between INR tests are both taken into account. On the basis of this approach, INR values were within the therapeutic range a mean of 55% of the time in ROCKET AF and 64% in the RE-LY trial. So the interpretation of noninferiority in a given trial may also depend both on the homogeneity and treatment accuracy of the warfarin cohort and on the dummy INR algorithm that is used.