← Back
NeuroTrials.ai
Neurology Clinical Trial Database

ARCADIA-MRI

Apixaban to Prevent Covert Infarcts After Cryptogenic Stroke in Patients With Atrial Cardiopathy — A Secondary Analysis of the ARCADIA Randomized Clinical Trial

Year of Publication: 2025

Authors: Lansberg MG, Wintermark M, Chen H, ..., Lazar RM

Journal: JAMA Neurology

Citation: JAMA Neurol. 2025;82(3):220-227. doi:10.1001/jamaneurol.2024.4838

Link: https://jamanetwork.com/journals/jamaneu...article/2829569


Clinical Question

In patients with recent cryptogenic stroke and atrial cardiopathy, does apixaban reduce incident nonlacunar covert infarcts on MRI compared to aspirin?

Bottom Line

Among patients with recent cryptogenic stroke and atrial cardiopathy enrolled in ARCADIA, apixaban compared with aspirin was associated with markedly fewer incident nonlacunar covert infarcts on MRI (5.1% vs 17.9%; weighted RR 0.29, 95% CI 0.10–0.83; P=0.02). The result contrasts with the neutral primary outcome of the parent ARCADIA trial and should be viewed as hypothesis-generating — the substudy enrolled a small, more drug-adherent subset and the CI is wide.

Major Points

  • ARCADIA-MRI was an ancillary MRI substudy of the ARCADIA trial (apixaban vs aspirin in cryptogenic stroke with atrial cardiopathy).
  • 310 of 1015 ARCADIA participants (31%) enrolled; 174 (56% of enrolled) had adequate baseline and follow-up MRIs and were included in the primary analysis (apixaban n=79, aspirin n=95).
  • Primary endpoint — incident nonlacunar covert infarct on follow-up MRI — occurred in 5.1% with apixaban vs 17.9% with aspirin (weighted RR 0.29, 95% CI 0.10–0.83; P=0.02).
  • Composite of nonlacunar covert infarct or nonlacunar symptomatic stroke: 9% vs 26%, RR 0.36 (0.17–0.79), P=0.01.
  • No benefit on covert lacunar infarcts (10% vs 13%, RR 0.80, 0.34–1.86) — consistent with cardioembolic mechanism for nonlacunar infarcts.
  • Annualized nonlacunar covert infarct rate: 2.0% (apixaban) vs 7.4% (aspirin) — lower than the prespecified 13.5% projection.
  • Result is discordant with the neutral parent ARCADIA trial (HR 1.00 for recurrent stroke); authors attribute discordance to higher drug adherence in the MRI substudy (15.5% premature discontinuation vs 50.8% in ARCADIA-MRI screen failures) and selection of a more compliant cohort.
  • Median follow-up 27 months; MRI scans interpreted by 2 raters blinded to treatment (interrater κ 0.81 for nonlacunar infarct).
  • No differential treatment effect across prespecified subgroups (NT-proBNP, ECG, age, race, diabetes, sex, baseline mRS).
  • Limitations: small sample (174 of 1015 ARCADIA participants), wide CI, baseline imbalances (more prior stroke/TIA and higher NT-proBNP in aspirin arm), high COVID-era attrition.
  • Authors conclude results are hypothesis-generating and may justify on-treatment analysis of the parent ARCADIA trial.

Design

Study Type: Prospective, multicenter, double-blind, parallel-arm RCT — ancillary MRI substudy of ARCADIA

Randomization: 1

Blinding: Double-blind; MRI raters masked to treatment assignment

Enrollment Period: November 14, 2019 to December 2, 2022 (parent ARCADIA stopped for futility Dec 14, 2022)

Follow-up Duration: Median 27 months (median 811 days between baseline and follow-up MRI; IQR 487–1288)

Centers: 75

Countries: USA

Sample Size: 174

Analysis: Intention-to-treat with inverse probability weighting (a priori primary). Poisson regression with observation time as covariate. Unweighted analysis confirmatory.


Inclusion Criteria

  • Enrolled in the parent ARCADIA trial (recent cryptogenic ischemic stroke + atrial cardiopathy biomarker: NT-proBNP >250 pg/mL OR P-wave terminal force V1 >5000 μV×ms OR LA diameter index ≥3 cm/m²)
  • Able to undergo brain MRI
  • Had not permanently discontinued ARCADIA study drug at time of consent
  • Able to provide informed consent

Exclusion Criteria

  • Diagnosis of dementia
  • Active illicit drug use
  • Psychiatric admission for depression within last 2 years
  • Less than 8 years of education
  • History of traumatic brain injury with loss of consciousness >30 minutes
  • Permanent discontinuation of ARCADIA study drug before consent
  • MRI contraindication

Baseline Characteristics

CharacteristicApixabanAspirin
N7995
Mean age (SD)66.3 (10.2) yr66.3 (11.0) yr
Male53.2%51.6%
Black16.7%21.3%
Hispanic3.8%2.1%
High school graduate38.0%52.6%
Hypertension72.2%74.7%
Prior/current tobacco41.8%32.6%
Diabetes20.3%29.5%
Prior stroke or TIA15.2%25.5%
Ischemic heart disease10.1%11.7%
Heart failure2.5%5.3%
CHA2DS2-VASc (mean SD)4.2 (1.2)4.5 (1.1)
mRS at baseline (median IQR)1 (0–2)1 (0–2)
NIHSS (median IQR)0 (0–2)1 (0–2)
NT-proBNP (median pg/mL)199 (43–355)271 (108–430)
PTFV1 (median μV×ms)5250 (2500–6300)5200 (2500–5625)
LA diameter index (median cm/m²)1.9 (1.6–2.1)1.8 (1.6–2.1)
Fazekas score (median IQR)2 (1–3)2 (1–3)
Time from ARCADIA randomization to MRI consent (median days)179 (48–364)93 (37–362)
Time baseline to follow-up MRI (median days)800 (479–1311)822 (487–1238)
Premature study drug discontinuation17.7%13.7%

Arms

FieldApixabanControl
InterventionApixaban 5 mg twice daily (2.5 mg twice daily if criteria met for dose reduction) per parent ARCADIA protocolAspirin 81 mg daily per parent ARCADIA protocol
DurationUntil parent trial end (Dec 21, 2022) or premature discontinuationUntil parent trial end (Dec 21, 2022) or premature discontinuation

Outcomes

OutcomeTypeControlInterventionHR / OR / RRP-value
Incident nonlacunar covert infarct on follow-up MRI (new ischemic lesion not present on baseline MRI, not lacunar, not from growth of index infarct, identified by 2 masked raters)Primary17.9% (17/95)5.1% (4/79)weighted RR 0.29P=0.02
Composite of nonlacunar covert infarct or nonlacunar symptomatic strokeSecondary26% (25/95)9% (7/79)RR 0.36P=0.01
Nonlacunar symptomatic strokeSecondary8% (8/95)4% (3/79)RR 0.52P=0.30
Lacunar covert infarctSecondary13% (12/95)10% (8/79)RR 0.80P=0.60
Composite lacunar covert infarct or lacunar symptomatic strokeSecondary15% (14/95)10% (8/79)RR 0.66P=0.33
Composite of any (lacunar or nonlacunar) covert infarctsSecondary26% (25/95)15% (12/79)RR 0.57P=0.08
Annualized rate of nonlacunar covert infarctSecondary7.4%/yr2.0%/yr
Detailed bleeding/safety outcomes are reported in the parent ARCADIA trial (JAMA 2024); ARCADIA-MRI focused on imaging outcomes. No new safety signals were reported in the MRI substudy.Adverse

Subgroup Analysis

No differential treatment effect on the primary radiological outcome across any of the prespecified subgroups (NT-proBNP criteria, ECG criteria, age ≤67 vs >67, Black vs other race, type 2 diabetes, sex, baseline mRS 0–2 vs 3–5). All p-values for interaction >0.19.


Criticisms

  • Small sample size (174 of 1015 ARCADIA participants, 17%) — relatively few events with wide confidence intervals (RR 0.29, 95% CI 0.10–0.83).
  • Discordant with the neutral primary outcome of the parent ARCADIA trial (HR 1.00 for recurrent stroke). Authors attribute discordance to higher drug adherence in the MRI substudy population (15.5% premature discontinuation vs 50.8% in screen failures).
  • Highly selected population: required no permanent discontinuation of study drug at consent + ability/willingness to undergo MRI; results may not generalize to less adherent patients.
  • Baseline imbalances: aspirin arm had numerically more prior stroke/TIA (25.5% vs 15.2%) and higher NT-proBNP (271 vs 199 pg/mL), which could have inflated the apparent treatment effect.
  • COVID-19 pandemic affected MRI completion rates — 56% adequate MRIs is lower than typical for an MRI substudy.
  • Aspirin-arm nonlacunar covert infarct rate (7.4%/yr) was substantially lower than the prespecified 13.5% projection, raising questions about external validity.
  • Hypothesis-generating only — authors explicitly note results should not change practice without a definitive on-treatment analysis or replication.

Funding

National Institute of Neurological Disorders and Stroke (5U01NS110728). Bristol Myers Squibb–Pfizer Alliance provided apixaban; Roche Diagnostics provided NT-proBNP assay supplies. ClinicalTrials.gov NCT03192215.

Based on: ARCADIA-MRI (JAMA Neurology, 2025)

Authors: Lansberg MG, Wintermark M, Chen H, ..., Lazar RM

Citation: JAMA Neurol. 2025;82(3):220-227. doi:10.1001/jamaneurol.2024.4838

Content summarized and formatted by NeuroTrials.ai.