4.2.1. Efficacy

In ENGAGE AF-TIMI 48, edoxaban 60 mg was noninferior to well-controlled warfarin therapy. However, the trial failed to show superiority. When breaking down the NI primary outcome, we notice that edoxaban 60 mg has slightly fewer patients with ischemic stroke events than warfarin, but the largest driver behind the overall difference was the lower number of hemorrhagic stroke events in the edoxaban 60 mg group. The NI margin in ENGAGE AF-TIMI 48 was slightly more restrictive to those seen with dabigatran and rivaroxaban in the ROCKET-AF and RE-LY trials (NI margin of 1.46 in both); the margin in ENGAGE AF-TIMI 48 was equal to that in the ARISTOTLE trial that assessed apixaban (NI margin of 1.38). As such, the NI margin appears consistent with previous DOAC trials.

Although NI testing over the primary outcome of stroke and SEE produced an HR with a 97.5% CI that did not cross the 1, subsequent superiority testing did not produce a statistically significant finding. time frame for the additionally included from study drug to warfarin or another on-market new DOACs. Another difference is the population analysis used. In the NI primary outcome, a mITT population analysis set was used; in the primary superiority testing, an ITT population analysis set was used. The differences in these calculation sets are that ITT would include any patient randomized, regardless of whether the patient received any dose of the intervention therapy, and the overall study period would include the additional time during which patients were transitioned from the edoxaban 60 mg intervention to open-label anticoagulation therapy. When looking at the breakdown of the individual outcomes with superiority testing, we find that the rate of ischemic stroke is similar in the edoxaban 60 mg arm and in the warfarin arm, as opposed to being lower with NI testing. The increase was largely driven by ischemic stroke events disproportionally affecting the edoxaban 60 mg arm, leading to an almost equal number of ischemic strokes in both the edoxaban 60 mg and warfarin arms. Since around two-thirds of the patients in the ENGAGE AF-TIMI 48 moved to open-label warfarin, this would indicate a possible increased risk of ischemic stroke in patients transitioning from edoxaban to warfarin until such time that their INR became therapeutic.

Subgroup analyses were performed for the primary outcome, and the results were similar to the base case. Trends from the subgroup analyses suggest that edoxaban treatment may be less efficacious than warfarin in patients with high CrCL. However, since the trial was neither powered for nor meant to test these subgroups, the results from the subgroup analyses are to be considered exploratory in nature and should only be used for hypothesis generation that would require further hypothesis testing.

Secondary outcome analysis showed fewer events in the edoxaban 60 mg arm than in the warfarin arm. Specifically, edoxaban 60 mg was associated with fewer instances of myocardial infarction, CV mortality, and overall mortality. These outcomes, however, should not be statistically analyzed because of the failure of establishing superiority in the second step of hierarchal testing.

The eight IDCs reviewed (see 0) showed similar results, mostly supporting the notion that edoxaban 60 mg has similar efficacy to VKA (warfarin) and other DOACs. However, all comparisons between the different DOACs can only be informed indirectly through a comparison of single trials that independently compared a particular DOAC with a VKA. Such intertrial comparisons tend to be problematic, as adjustment of cross-trial differences can never be adequately achieved and only trials directly evaluating one DOAC against another can provide any confidence with regard to the respective merits of these drugs. Such an informative challenge precludes analysis using a random-effects model. Thus, a fixed-effects model was used with all the reviewed IDCs. The fixed-effects model makes unrealistic assumptions about the true treatment effect, specifically that all trials share the same common effect and that any differences between trials are due to sampling error. In other words, the fixed-effects model assumes that all the differences in the study methodology and patient characteristics between trials have no influence on the true treatment effect. Such assumptions are inappropriate given the clinical heterogeneity in the evidence network, consisting of the pooled populations of the four phase III studies of the DOACs presently on the market. However, considering the nature of the available evidence, no better IDC of DOACs could have been produced.

As with other DOAC trials, the composite outcome in ENGAGE AF-TIMI 48 is hard to translate into useful clinical insights. Breaking down the primary outcome into its individual components would lead to statistical uncertainty in the reliability of the results, as the trial is not powered to detect differences in the individual components, and adjusting for multiple outcomes and comparisons becomes impracticable. It is thus hard to say anything beyond that edoxaban 60 mg is noninferior in efficacy to warfarin and is similar in this regard to the other DOACs. Considering that warfarin treatment requires strict adherence, frequent testing and dose adjustments, and specific dietary restrictions, edoxaban can be viewed as a more convenient alternative to warfarin.

4.2.2. Harms

Major bleeding events were the primary safety outcome of ENGAGE AF-TIMI 48 and represent the biggest safety concern associated with antithrombotic drugs. In ENGAGE AF-TIMI 48, edoxaban 60 mg resulted in a statistically significantly lower rate of major bleeding events than did warfarin. The lower incidence of major bleeding was also mirrored in minor bleeding, clinically relevant non-major bleeding, and life-threatening bleeding. Only gastrointestinal bleeding showed a numerically higher incidence in the edoxaban 60 mg arm than in the warfarin arm. The rate of nonbleeding adverse events and nonbleeding SAEs were similar for edoxaban 60 mg and for warfarin.

Currently, there is no direct comparison relating to safety between edoxaban and any other DOACs. The available IDCs suggest that edoxaban does result in fewer major bleeding events than does warfarin. However, the results of the IDCs of edoxaban with other DOACs are mixed, with some showing statistically significant results in favour of edoxaban, and others not showing statistical significance. It is important to consider the results of the IDCs with considerable caution. The reliability and accuracy of these analyses are limited given that there are no studies directly comparing DOACs head to head, that the four phase III trials comparing the individual DOACs to warfarin show substantive heterogeneity in their study methodology and clinical populations, and that the fixed-effects model was used in all IDCs reviewed. As such, all the results of the DOAC IDCs should be considered as exploratory in nature and in need of further hypothesis testing. However, given the data currently available, no IDC of better quality could have been produced.