Andexanet alfa: expensive and harmful?

We have previously discussed the many pharmaceutical advertisements published by the New England Journal of Medicine, thinly disguised as science, such as the original open label uncontrolled look at andexanet alfa. (Connolly 2019) At that point, I concluded that andexanet alfa should clearly not be used (although our pharmacies didn’t listen and wasted a lot of money stocking this experimental chemical), and that we should wait for proper RCTs looking at clinical outcomes. We now have the first of the RCTs: the ANNEXA-1 trial.

The paper

Connolly SJ, Sharma M, Cohen AT, et al. Andexanet for Factor Xa Inhibitor-Associated Acute Intracerebral Hemorrhage. N Engl J Med. 2024 May 16;390(19):1745-1755. doi: 10.1056/NEJMoa2313040. PMID: 38749032 NCT03661528

The Methods

This is an industry-funded, open-label, multicenter RCT.

Patients

They initially included all patients with an intracranial hemorrhage who were taking a factor Xa inhibitor and had taken the most recent dose within 15 hours.

However, they made multiple amendments to the study over time, with no real explanations in the manuscript, which is really concerning in a trial completely run by the company trying to sell us this drug. They added exclusions including a lack of a primary subdural or subarachnoid hemorrhage, an estimated hematoma volume between 0.5 and 60 mL, and a maximum NIHSS score of 35.

They excluded patients with a GCS less than 7, an NIHSS more than 35, with a thrombotic event within 2 weeks, or if surgery was planned within 12 hours. 

Intervention

Andexanet, either in a high or low dose (although strangely that dose is not specified in the manuscript, and apparently only available in a supplement.) 

Comparison

Usual care, excluding andexanet but potentially including prothrombin complex concentrate. 

Outcome

The primary outcome was hemostatic efficacy at 12 hours. 

They use a different definition of hemostatic efficacy than the ANNEXA-4 trial, but unfortunately it is completely unclear when the definition was instituted, as it is not listed in the original clinicaltrials.org listing, and only added after the majority of patients had been recruited. 

The definition they use is “a change in the hematoma volume of 20% or less (excellent hemostatic efficacy) or 35% or less (good hemostatic efficacy) within 12 hours after baseline, an increase in the NIHSS score of less than 7 points at 12 hours, and receipt of no rescue therapies such as andexanet, prothrombin complex concentrate, or surgery to decompress the hematoma within 3 to 12 hours after randomization.”

The Results

They planned to enroll 900 patients, but stopped the trial early with 550 patients enrolled, 452 of which were included in the primary efficacy analysis. 

The median age was about 78 years, and the baseline NIHSS was 9 in both groups. Patients were enrolled in a median of about 2 hours from symptom onset. The median baseline hematoma was about 10 mL. Surprisingly (to me at least), only about 10% of these bleeds were traumatic. (The vast majority of the bleeds that I see on factor Xa inhibitors are associated with a fall.) 80% of the intervention group received ‘low dose’ andexanet. 86% of the usual care group received PCCs, almost all of which was 4 factor. 

The surrogate composite primary outcome of hemostatic efficacy was improved with andexanet (67% vs 53%, ARR 13.4%, 95% CI 4.6-22.2%, p=0.003). The only component of the composite outcome that was statistically positive was hematoma volume (the least clinically important outcome.)

However, all clinical outcomes (what they call the ‘safety outcomes’) look worse with andexanet. There was a statistical increase in thrombotic events (10.3% vs 5.6%, p=0.048). Ischemic stroke was statistically higher (6.5% vs 1.5%). MI and death were also both higher with andexanet, although not statistically so. In their post-hoc analysis using 30 day modified Rankin scores to assess functional outcomes, there was clearly no benefit, with 31% of the control group having a favourable outcome as compared to 28% of the andexanet group. 

My thoughts

Despite being reported as positive, this is clearly a negative trial, with all clinical outcomes being either neutral or worse with treatment with andexanet. Despite being a multicenter RCT, this is a very low quality trial with very high risk of bias.

This trial was not blinded, which is especially problematic in a trial with heavy financial conflicts of interest. It would probably have been less problematic with an objective outcome, but their final (if not original) primary outcome includes very subjective functional outcomes, as well as subjective decisions about clinical management. It is a bad outcome for an unblinded trial. They do not provide any rationale for leaving this trial unblinded, and it would have been a very easy trial to blind, which is a huge knock against the quality of the trial. 

As mentioned above, they made multiple amendments to the protocol without good (or any) explanations, which increases the risk of bias and p-hacking, especially in a conflicted study. The changes in population also make the results more difficult to apply clinically, although ultimately that doesn’t seem to matter, as the results suggest we should not be using andexanet.

The exclusion of patients with planned surgery within 12 hours makes no sense. These are the patients who will likely have the best outcomes, and in whom coagulopathy reversal is probably the most important. This exclusion only makes sense if your primary outcome is based around hematoma size. They claim their primary outcome includes functional outcomes, in which case this exclusion makes no sense. However, that primary outcome wasn’t registered until 4 years after this initial registration. This exclusion seems to hint that this was not their originally planned primary outcome. 

Although transparency is appreciated (especially compared to the average publication, where conflicts are only evident after searching independent sites), the first section of this paper is rife with conflict of interest statements that I haven’t encountered before. It is not surprising, but this trial was designed by the drug company, written by a medical writer hired by the drug company, and the authors all have confidentiality agreements with the drug company. There is incontrovertible evidence that such conflicted trials are biased in favour of finding positive results for the company sponsoring the trial. When reading a trial with conflict of interest, you must downgrade the results to account for that known bias. A trial with conflict of interest (especially an unblinded trial) that demonstrates net harm is particularly damning, and makes it very clear that no one should be using this drug at this time. 

Like many trials, this trial was stopped early. In their methods, they state that they changed their stopping criteria from a p value of 0.001 to a p-value or 0.031. In other words, they decided to use a much more relaxed statistical measure, and do not explain this choice. Stopping trials early is problematic for many reasons. In this trial, the safety (aka real clinical) outcomes, including death, all look worse with andexanet, and running the trial to completion might have provided more definitive evidence of harm, including the possibility that this drug is increasing mortality. 

Bottom line

At this point, it is very clear that andexanet should not be used clinically. There is still a lot of uncertainty in the data, and further research is warranted, but for now prothrombin complex concentrates are the standard for factor Xa reversal. Hospitals should probably stop wasting money on andexanet unless better independent data showing improved clinical outcomes is published. 

Other FOAMed

Andexanet Alfa: More garbage science in the New England Journal of Medicine

ANNEXA-1: Andexanet Alfa Associated with Harm in DOAC Reversal – REBEL EM

Evidence based medicine is easy

The EBM bibliography

Evidence based medicine resources

EBM deep dives

References

Connolly SJ, Crowther M, Eikelboom JW, et al. Full Study Report of Andexanet Alfa for Bleeding Associated with Factor Xa Inhibitors. N Engl J Med. 2019

Connolly SJ, Sharma M, Cohen AT, Demchuk AM, Członkowska A, Lindgren AG, Molina CA, Bereczki D, Toni D, Seiffge DJ, Tanne D, Sandset EC, Tsivgoulis G, Christensen H, Beyer-Westendorf J, Coutinho JM, Crowther M, Verhamme P, Amarenco P, Roine RO, Mikulik R, Lemmens R, Veltkamp R, Middeldorp S, Robinson TG, Milling TJ Jr, Tedim-Cruz V, Lang W, Himmelmann A, Ladenvall P, Knutsson M, Ekholm E, Law A, Taylor A, Karyakina T, Xu L, Tsiplova K, Poli S, Kallmünzer B, Gumbinger C, Shoamanesh A; ANNEXA-I Investigators. Andexanet for Factor Xa Inhibitor-Associated Acute Intracerebral Hemorrhage. N Engl J Med. 2024 May 16;390(19):1745-1755. doi: 10.1056/NEJMoa2313040. PMID: 38749032