AMAG-423 in severe preeclampsia: a digoxin immune Fab Phase 2b/3a stopped by the DSMB for futility

What was tried

NCT03008616 was a Phase 2b/3a randomized, double-blind, placebo-controlled study of AMAG-423 in antepartum patients with severe preeclampsia between 23 0/7 and 31 6/7 weeks of gestation. AMAG-423 is an ovine digoxin-specific Fab fragment, the same molecular entity sold as Digibind and DigiFab for reversal of acute digoxin toxicity. AMAG Pharmaceuticals acquired the asset and repositioned it for preeclampsia.

Patients received 3.2 mg/kg IV every 6 hours for 4 days on top of standard expectant management. Masking was quadruple. The primary endpoint was a neonatal composite of severe intraventricular hemorrhage (grade 3 or higher), necrotizing enterocolitis, or death by 36 weeks corrected gestational age. The trial enrolled 58 patients before the DSMB recommended stopping for futility on 13 August 2020. No safety concern was raised.

The biological hypothesis

The rationale was a single causal chain. Endogenous cardiotonic steroids (ECS), including marinobufagenin and endogenous ouabain, circulate at elevated concentrations in patients with severe preeclampsia. ECS bind and inhibit the alpha subunit of Na+/K+-ATPase (ATP1A1) in vascular smooth muscle and renal tubules. Sustained inhibition raises intracellular sodium and calcium and increases vascular tone. A digoxin-specific Fab that cross-reacts with marinobufagenin and other bufadienolides could in principle scavenge circulating ECS and restore pump activity.

Preclinical support was real but narrow. Vu et al. showed that DigiFab reverses marinobufagenin-induced Na+/K+-ATPase inhibition in human umbilical artery (Reprod Sci 2012) and protects endothelial barrier integrity in vitro. A small open-label series of 51 women reported reduced blood pressure variance and improved creatinine clearance (Am J Perinatol 2010). The Open Targets score for ATP1A1 against severe preeclampsia is 0.092, driven by clinical precedence rather than genetic association.

What actually happened

The DSMB called the study after roughly half of the planned enrollment. The posted results explain why. In the AMAG-423 arm, 4 of 28 neonates suffered stillbirth or infant death, 3 had severe IVH, and 2 had NEC. In the placebo arm, 3 of 29 neonates died, 2 had severe IVH, and 0 had NEC. Direction-of-effect favored placebo on each subcomponent, though confidence intervals were wide.

The maternal renal secondary endpoint moved in the wrong direction. Mean change in serum creatinine over 24 hours was +1.0 µmol/L on AMAG-423 (SD 8.51) and -2.8 µmol/L on placebo (SD 9.45). Maternal pulmonary edema occurred in 2 of 29 placebo patients and 0 of 28 AMAG-423 patients, a small numerical signal against placebo that did not redeem the primary composite.

Serious adverse events were balanced at 15 of 28 on AMAG-423 versus 8 of 29 on placebo, dominated by expected pregnancy events. The DSMB noted no safety signal of concern.

Failure mechanism, best guess

ECS elevation in preeclampsia is real, and so is its association with Na+/K+-ATPase inhibition. The translational gap was assuming that ECS sit on the causal path to neonatal outcomes at 23 to 32 weeks of gestation, rather than being downstream readouts of placental dysfunction. The dominant validated pathophysiology centers on antiangiogenic placental signaling, in particular sFlt-1 (the soluble form of FLT1) and its sequestration of VEGF and PlGF (JACC 2020). Neutralizing circulating ECS does not address the placental lesion generating sFlt-1.

A second contributor was endpoint geometry. The neonatal composite is dominated by gestational age at delivery, which is driven by maternal indication for delivery rather than the investigational drug. A 4-day infusion ending well before delivery has limited latitude to move events occurring weeks later.

How to prevent this next time

A target-disease association score below 0.10, driven entirely by clinical precedence with no genetic association, should anchor the Bayesian prior close to historical Phase 2 obstetric attrition. Industry-wide Phase 2 to Phase 3 transition rates for obstetric programs sit between 22 and 29 percent (BIO/Informa 2011-2020 analysis).

Two leverage points would have improved the predictive value of this investment. First, biomarker enrichment. A prospectively defined sFlt-1/PlGF ratio cutpoint above 85 (Verlohren et al., Am J Obstet Gynecol 2014) identifies a population whose placental disease is already advanced. A baseline marinobufagenin measurement above an ECS-driven subset threshold would have been a more direct mechanistic stratifier. Second, Bayesian futility with an explicit posterior probability of success:

PP(success)=∫θ​P(final success∣θ,interim data)⋅π(θ∣interim data)dθ

With a weakly skeptical prior (mean treatment effect of 5 percent absolute risk reduction, SD of 10 percent) and 57 patients at interim, the posterior probability of a true 10 percent absolute reduction or greater was below 0.05, consistent with the DSMB call.

The single highest leverage change would have been to require, before opening Phase 2b/3a, a prespecified human biomarker trial demonstrating that 4-day IV AMAG-423 actually lowers circulating marinobufagenin and restores Na+/K+-ATPase activity in maternal red cells in vivo.

What this means for similar programs

The ECS hypothesis remains biologically plausible. The lesson is not that ATP1A1 is the wrong place to look in preeclampsia, but that programs targeting circulating mediators downstream of placental injury should be powered against a delivery-side endpoint paired with an enrichment biomarker mechanistically proximal to the drug. The dose-response of this antibody was characterized for pulses of digoxin in poisoning, not for picomolar steady-state ECS load.

Open questions

How did the marinobufagenin pharmacodynamic biomarker move under AMAG-423? AMAG never published a PD subanalysis. Whether scavenging ECS measurably altered Na+/K+-ATPase activity in this cohort is the most actionable question left on the table.

Does the same Fab, given pre-symptomatically in a high-risk cohort defined by elevated sFlt-1/PlGF at 20 to 24 weeks, alter the trajectory of placental dysfunction? No one has tested it.

Sources

1. ClinicalTrials.gov NCT03008616, results section, accessed 2026-05-24. https://clinicaltrials.gov/study/NCT03008616.

2. Vu HV et al. Reprod Sci. 2012;19(12):1314.

3. Adair CD et al. Am J Perinatol. 2010;27(8):655.

4. Bagrov AY et al. Pharmacol Rev. 2009;61(1):9.

5. Rana S et al. Circ Res. 2019;124(7):1094.

6. Karumanchi SA, Granger JP. JACC. 2020;76(14):1690.

7. Pankiewicz K et al. Int J Mol Sci. 2023;24(16):12743.

8. Verlohren S et al. Am J Obstet Gynecol. 2014;210(2):157.e1.

9. Open Targets Platform, ATP1A1 vs severe preeclampsia (MONDO_0001641), score 0.0924, accessed 2026-05-24. https://platform.opentargets.org/target/ENSG00000163399/disease/MONDO_0001641.

10. OpenFDA FAERS, search term digifab, 197 reports, accessed 2026-05-24. https://api.fda.gov/drug/event.json?search=patient.drug.medicinalproduct:%22digifab%22.