Premature Atrial Contractions as a Marker of Atrial Cardiopathy: a Revised Analysis of the ARCADIA Randomized Trial

Abstract

Introduction

Atrial cardiopathy may be associated with an increased risk of stroke independent of atrial fibrillation (AF). In the ARCADIA trial, apixaban was not superior to aspirin in preventing recurrent stroke among patients with a cryptogenic stroke and atrial cardiopathy. We aimed to determine whether the presence of at least one premature atrial complex (PAC), a known harbinger of AF and stroke, would enhance the ability to identify individuals most likely to benefit from apixaban.

Methods

In ARCADIA, atrial cardiopathy was defined by NT-proBNP >250 pg/ml, a P-wave terminal force greater than 5000 μV × ms in lead V1, or a left atrial diameter index ≥ 3 cm/m² on echocardiogram. For the current analysis, the presence of any PAC on the baseline 12-lead ECG was substituted for the less atrial-specific NT-proBNP criterion. The presence of any PAC was also assessed as a sole atrial cardiopathy criterion.

Results

Of the 1,015 patients randomized in ARCADIA, 85 had at least one PAC. The revised atrial cardiopathy criteria were met by 593 patients; 301 were randomized to apixaban, and 292 to aspirin. The annualized recurrent stroke rates were 3.1% for apixaban versus 4.4% for aspirin (HR 0.71, 95% CI, 0.38–1.34, p=0.29). No differences in risk of recurrent stroke among participants with PACs, compared to those without PACs, were observed.

Conclusion

In patients enrolled in the ARCARDIA trial, utilizing the presence of PACs as a potential marker of atrial cardiopathy did not reveal definitive evidence of benefit of apixaban compared to aspirin.

Introduction

While it is commonly believed that atrial fibrillation (AF) leads to thromboemboli that arise due to AF-induced atrial stasis, recent and growing evidence suggests that thromboembolism may arise from an atrial cardiopathy, or atrial myopathy, with AF occurring as an epiphenomenon.^1,2

ARCADIA, the first major randomized trial to rigorously test the hypothesis that a direct oral anticoagulant might be superior to aspirin in preventing recurrent stroke among those with cryptogenic stroke and atrial cardiopathy, was negative.³ We hypothesized that the inclusion criteria may not have been sufficiently specific to the atria.⁴ Specifically, N-terminal pro-B-type natriuretic peptide (NT-proBNP) also may be elevated due to left ventricular wall stress^5,6 and therefore may be less indicative of pathology arising from the atria per se. In fact, in ARCADIA, when inclusion criteria were analyzed separately, apixaban was associated with more favorable (albeit not statistically significant) outcomes when examining markers of the more specific atrial pathologies (i.e., P wave terminal force in lead V1 and left atrial diameter).³

Premature atrial contractions (PACs) have been proposed as a marker of atrial cardiopathy,^7–9 and we have shown that they are one of the most potent predictors of incident AF.^10,11 Even independent of any recognized AF, PACs have been shown to predict stroke.^12–15 PACs may also be more than a passive marker given recent evidence from a swine model demonstrating that frequent induced PACs might cause adverse atrial structural remodeling and a heightened propensity to AF.¹⁶ While the frequency of PACs determined by continuous electrocardiographic monitoring is the gold standard to determine PAC burden, we have shown that even a single PAC from a single standard 10 second 12-lead ECG can distinguish those at higher risk of incident AF.¹⁷

Therefore, we aimed to evaluate whether inclusion of the presence of a PAC as a criterion for atrial cardiopathy might identify individuals likely to benefit from apixaban compared to aspirin.

Methods

ARCADIA was a prospective, multicenter, randomized (1:1) controlled trial to compare the efficacy of apixaban versus aspirin on the recurrence of cryptogenic stroke. The ARCADIA design has been previously thoroughly described.^3,18

All consented and randomized participants in the original ARCADIA trial were included in the current analyses. Pertinent inclusion criteria in the original study included age ≥ 45 years, a confirmed cryptogenic ischemic stroke occurring withing 180 days of randomization, a Modified Rankin Scale score ≤ 4 and evidence of atrial cardiopathy, defined as P-wave terminal force in V1 lead electrogram (PTFV1) greater than 5000 μV × ms, serum NT-proBNP level greater than 250 pg/mL, or left atrial diameter index of 3 cm/m² or greater on echocardiogram. Patients with evidence of lacunar infarcts, cervical or intracranial arterial stenosis ≥ 50% in relevant arteries to the site of infarction, history of atrial fibrillation, left ventricular ejection fraction <30%, chronic kidney disease, history of intracranial bleeding or contraindication to anti-thrombotic therapy were excluded.

The primary efficacy endpoint of the ARCADIA trial was recurrent ischemic, hemorrhagic, or undetermined type stroke. Secondary efficacy endpoints were recurrent ischemic stroke or systemic embolism and a composite of recurrent stroke of any type or death.

Two hypotheses were tested in the current revised analyses: first, that at least one PAC on the baseline 12-lead ECG would identify cryptogenic stroke patients who would benefit from apixaban compared to aspirin. Second, we hypothesized that replacing the presence of a PAC for the NT-proBNP inclusion criterion would provide a more specific definition of atrial cardiopathy patients, revealing those most likely to benefit from apixaban compared to aspirin.

The first technically adequate ECG done after stroke onset was used, and PACs were defined as present (at least one) or absent (none) by Minnesota code criteria (Minnesota codes 8.1.1, 8.1.3).¹⁹ All electrocardiograms were centrally read at the Epidemiological Cardiology Research Center (Wake Forest School of Medicine, Winston-Salem, NC) by readers who were blinded to treatment arm and study outcomes.

Health Canada, the StrokeNet Central Institutional Review Board, and institutional review boards or research ethics boards at participating sites approved the study protocol. All participants provided written, informed consent for trial participation.

Statistical analysis

Normally distributed continuous variables are presented as means and standard deviations and were compared using t-tests, while continuous variables with skewed distributions are presented as medians with interquartile ranges (IQR) and were compared using the Mann-Whitney U test. Categorical variables were compared using the Fisher’s exact or chi-squared test as appropriate.

The presence of at least one PAC was assessed as a predictor of the primary outcome. Differences in the primary outcome between randomization groups was assessed among those with PACs alone and in a group wherein the presence of a PAC was substituted for NT-proBNP as an inclusion criterion; atrial cardiopathy was thus redefined as PTFV1 ≥ 5000 μV × ms, left atrial diameter index ≥3 cm/m² or the presence of any PAC on 12-lead ECG. Curves of cumulative incidence of stroke recurrence among the study arms were constructed and compared using the log-rank test. Participants were followed from the day of randomization and censored from the analysis when they reached the primary end point, death, loss to follow-up, withdrawn consent, or the end of study follow-up. Cox proportional hazard regression models were used to generate hazard ratios (HR) and 95% confidence intervals (CI). Effect modification of randomization assignment by the presence or absence of a PAC on the primary outcome was evaluated by performing interaction analyses. Multivariable models included baseline characteristics for which differences between the study groups exhibited P values < 0.1.

Statistical analyses were performed using R Studio (R foundation for statistical computing, Vienna, Austria). A two-tailed p value < 0.05 was considered statistically significant.

Results

Between February 2018 and December 2022, 1,015 patients were included in the ARCADIA primary analysis. Of those, 85 patients exhibited at least one PAC on the baseline 12-lead ECG, and 593 patients met the modified atrial cardiopathy definition substituting a PAC for the NT-BNP criteria definition (Supplementary Figure S1). Of those who met the original NT-proBNP criterion, 422 did not exhibit any PACs, and 68 out of the 85 patients with PACs had elevated levels of NT-proBNP.

Analysis by the presence of PACs

Of those with any PAC, 38 patients were randomized to the apixaban group and 47 to the aspirin group. Baseline characteristics of both groups by the treatment arm is shown Supplementary Table S1.

Among those with PACs, the primary endpoint of recurrent stroke occurred in a total of 9 patients and did not differ between randomization groups (Table 1, Figure 1). Similarly, no other differences in the secondary endpoints among those with PACs were observed (Table 1). Moreover, the presence of any PACs on 12-lead ECG was not significantly associated with increased risk of recurrent stroke in the original ARCADIA cohort compared to those without any PACs (HR 1.49, 95% CI, 0.74–2.98, p=0.26), and the interaction between the presence of a baseline PAC and randomization to apixaban versus aspirin on the occurrence of the primary outcome was not statistically significant (P=0.74) (Supplementary Table S2). However, the number of outcomes in the group with PACs was small.

Table 1.

Efficacy Outcomes Restricted to 85 Patients with PACs

	Apixaban N=38			Aspirin N=47				P-Value
Outcome	Total events	Person-years	Annualized rate	Total events	Person-years	Annualized rate, %	Unadjusted hazard ratio (95% CI)
Primary Outcome	4	54.35	7.4	5	88.64	5.6	1.13 (0.30– 4.22)	0.86
Recurrent ischemic stroke or systemic embolisma	4	54.35	7.4	5	88.64	5.6	1.13 (0.30–4.22)	0.86
Recurrent stroke of any type or death from any cause	7	54.35	12.9	7	88.64	7.9	1.46 (0.51–4.16)	0.48

^aNone experienced systemic embolic events

Figure 1. Cumulative Rates of the Primary Efficacy Outcome of Recurrent Stroke in Patients with PACs, Stratified by Treatment Group.

Recurrent stroke included stroke of ischemic, hemorrhagic, or unknown type. Ticks on curves denote censoring.

Analysis by the revised atrial cardiopathy definition

The proportions of patients with each specific criterion are shown in Supplementary Figure S2. Using the new definition (substituting the presence of a PAC for the original NT-pro-BNP criterion), 301 (51%) patients were randomized to receive apixaban and 292 (49%) were randomized to aspirin. Baseline characteristics remained balanced, except for a higher prevalence of smoking in the apixaban group and significantly higher levels of NT-proBNP in the aspirin group (Table 2).

Table 2.

Baseline Characteristics of Patients Eligible According to the Revised Atrial Cardiopathy Criteria (Intention-to-Treat)

Characteristic	Apixaban (N=301)	Aspirin (N=292)	P-Value
Age, mean (SD), y	66 (10)	66 (11)	0.29
Female – no. (%)	128 (45)	141 (50)	0.21
Race– no. (%)
Asian	6 (2)	8 (3)
Black or African American	80 (27)	81 (28)
White	207 (70)	199 (69)
Other	4 (1)	1 (0.3)
Body mass index (kg/m2)	29.3 (7)	29.5 (7)	0.74
Systolic Blood Pressure (mmHg)	137 (20)	134 (18)	0.2
Hypertension– no. (%)	227 (76)	223 (77)	0.78
Prior or current tobacco use – no. (%)	150 (50)	116 (40)	0.013
Diabetes – no. (%)	76 (25)	84 (29)	0.31
Prior stroke or TIA – no. (%)	43 (14)	58 (20)	0.062
Ischemic heart disease – no. (%)	25 (8)	27 (9)	0.65
Heart failure – no. (%)	18 (6)	15 (5)	0.65
Peripheral arterial disease– no. (%)	4 (1)	8 (3)	0.26
NIH Stroke Scale score, median (IQR)	1 [0–3]	1 [0–3]	0.20
Atrial cardiopathy markers
Premature Atrial Complexes on ECG – no. (%)	38 (13)	47 (16)	0.227
PTFV1, median (IQR), μV × ms	6000 [5250–7000]	5950 [5250–7200]	0.85
LA diameter index, mean (SD), cm/m2	1.9 (0.4)	1.9 (0.3)	0.34
NT-proBNP, median (IQR), pg/mL*	121 [50–299]	164 [64–370]	0.011

^*Excluded from the criteria in the current analysis.

Over 1,067 person-years of follow-up, the primary efficacy endpoint of recurrent stroke occurred in 17 patients in the apixaban group (annualized rate, 3.1%) versus 23 in the aspirin group (annualized rate, 4.4%) (HR, 0.71, 95% CI, 0.38–1.34, P value for log-rank test=0.29) (Table 3 and Figure 2). After multivariable adjustment for a history of transient ischemic attack (TIA) or stroke, smoking, and NT-proBNP levels, the association remained similar (adjusted HR for those assigned to apixaban versus aspirin 0.68, 95% CI, 0.36–1.29, p=0.24).

Table 3.

Efficacy Outcomes Restricted to 593 Patients with Atrial Cardiopathy

	Apixaban N=301			Aspirin N=292				P-Value		P-Value
Outcome	Total events	Person-years	Annualized rate, %	Total events	Person-years	Annualized rate, %	Unadjusted hazard ratio (95% CI)		Adjusted hazard ratio* (95% CI)
Primary Outcome	17	542	3.13	23	523.22	4.39	0.71 (0.38–1.34)	0.29	0.68 (0.36–1.29)	0.24
Recurrent ischemic stroke or systemic embolism	16	542	2.95	25	523.22	4.7	0.62 (0.33–1.16)	0.13	0.59 (0.31–1.11)	0.09
Recurrent stroke of any type or death from any cause	27	542	4.9	35	523.22	6.7	0.75 (0.45–1.24)	0.26	0.70 (0.42–1.17)	0.17

^*Adjusted to smoking, history of TIA or stroke and NT-proBNP

Figure 2. Cumulative Rates of the Primary Efficacy Outcome of Recurrent Stroke in Patients with Atrial Cardiopathy Using the Modified Criteria.

The modified criteria included P-wave terminal force in V1 lead electrogram (PTFV1) greater than 5000 μV × ms (used in the main trial), left atrial diameter index of 3 cm/m² or greater on echocardiogram (used in the main trial), but replaced the presence of at least one premature contraction on the baseline 12-lead ECG for the original criterion of serum NT-proBNP level greater than 250 pg/mL. Recurrent stroke included stroke of ischemic, hemorrhagic, or unknown type. Ticks on curves denote censoring.

Using the same modified criteria, neither the secondary endpoint of recurrent ischemic stroke or systemic embolism occurrence nor the combined end point of recurrent ischemic stroke and all-cause mortality differed significantly between those randomized to apixaban or aspirin (Table 3).

Discussion

In patients with a recent cryptogenic stroke and evidence of atrial cardiopathy in the ARCADIA trial, neither the presence of PACs nor its interaction with apixaban versus aspirin was significantly associated with stroke recurrence. Although a trend toward benefit was observed with apixaban compared to aspirin using the modified definition of atrial cardiopathy that substituted the presence of at least one PAC for NT-proBNP, the difference did not reach statistical significance.

Given the failure of two previous trials (NAVIGATE ESUS (rivaroxaban) and RE-SPECT ESUS (dabigatran)) to demonstrate superiority of anticoagulation over antiplatelet therapy for patients with ESUS, and even generating safety concerns in an unselected population of patients with cryptogenic stroke,^20–22 it is reasonable to seek criteria that may select for patients most likely to benefit from anticoagulation. In fact, a secondary analysis of the NAVIGATE-ESUS suggested superiority of rivaroxaban over antiplatelet therapy in patients with LA diameter>4.6 cm, providing evidence that markers of atrial cardiopathy may be particularly promising.²³

However, the ARCADIA trial failed to show benefit of apixaban compared to aspirin in patients with cryptogenic stroke and evidence of atrial cardiopathy as defined by three easy to acquire biomarkers, and the association between LA diameter and recurrent stroke was not replicated in ARCADIA.³

The negative findings from ARCADIA can broadly be interpreted in two ways: first, perhaps anticoagulation is simply not of benefit in the absence of AF, even given optimal markers of atrial cardiopathy. Or, perhaps the cardiopathy markers used were not specific enough.

We hypothesized that presence of PACs, an established marker of atrial cardiopathy and a predictor of stroke,^{8–10,12,13,17,24} could refine the specificity of the atrial cardiopathy criteria, and identify patients likely to benefit from apixaban in the ARCADIA. But consistent with the primary ARCADIA analysis, our revised analysis using a revised atrial cardiopathy definition was not able to reveal superior efficacy of apixaban compared to aspirin in cryptogenic stroke population.³ Although the point estimates suggest a potential benefit with apixaban, they are not statistically significant. However, our findings may be hypothesis-generating and support the need for implementing atrial-specific criteria.

The non-significance of the current analysis aligns with a secondary analysis of NAVIGATE-ESUS,²³ which failed to show a benefit of rivaroxaban over aspirin in ESUS patients with a high PAC burden (≥720 PACs/24 hours). However, as with the current analysis, that former study was likely underpowered due to the limited number of participants with a high PAC burden. Moreover, in the recent ATTICUS trial which included an ESUS population with at least one predictive factor for atrial fibrillation, apixaban did not reduce new ischemic brain lesions on MRI.²⁵

Given the negative evidence from 4 randomized trials, have we concluded the search for a group that could benefit from anticoagulation?²⁶ Before we close the chapter on anticoagulation in patients with cryptogenic stroke, we could consider refining our definitions of atrial cardiopathy, using more accurate tools to assess LA dimensions and fibrosis, combined with more atrial specific biomarkers. This strategy could distill the more severe forms of atrial cardiopathy that might benefit from anticoagulation. For example, a recent study has shown that mid-regional pro-atrial natriuretic peptide (MR-proANP), an atrial specific biomarker, is strongly predictive of cardioembolic stroke,²⁷ and the MOSES trial (NCT03961334) is currently evaluating the benefit of direct oral anticoagulation among patients with elevated levels MR-proANP. Moreover, while LA late gadolinium enhancement was not associated with stroke etiology at baseline in the atrial cardiomyopathy in patients with stroke of undetected mechanism (COAST) study,²⁸ the related ongoing prospective study will investigate associations between cardiac magnetic resonance (CMR) findings and recurrent stroke (although it may have limited power to assess follow-up endpoints).²⁹ However, given that a cardioembolic source was ascertained in only 9% of the total recurrent strokes in the NAVIGATE-ESUS trial,³⁰ LA embolism may not be the primary cause of most recurrent strokes, making patient stratification using any definition of LA cardiopathy difficult.

Limitations

The primary limitation of the current analysis is that it is very likely insufficiently powered to definitively exclude the relevant research hypothesis because PACs were not an inclusion criterion of the main trial. Further exacerbating this limitation, early termination of the trial for futility resulted in truncated follow-up time. Additional limitations of the current analysis include our ascertainment of PACs from a 10 second 12-lead ECG and the fact that we did not examine the burden of PACs using continuous ECG monitoring.¹³ However, the presence of a PAC on a 12-lead ECG likely indicates a higher burden of PACs, and more PACs predict both an increased risk of incident atrial fibrillation (AF) and stroke.^10,13 Because the entire cohort was limited to those who met the original inclusion criteria, we were unable to include individuals with PACs who did not also fulfill at least one of the initial criteria. For example, the majority of patients with PACs had elevated levels of NT-proBNP, and elevated NT-proBNP levels could be less specific for atrial cardiopathy, potentially blunting a possible benefit of apixaban in atrial cardiopathy.

Therefore, we cannot rule out benefit of apixaban confidently, especially when point estimates suggest a possible benefit of apixaban treatment. Future studies with larger sample size using additional left atrial imaging combined with more atrial specific biomarkers may reveal interesting groups that might benefit from anticoagulation.^27,29

Conclusion

In patients with cryptogenic stroke enrolled in the ARCADIA trial, utilizing the presence of PACs on a single ECG as a potential marker of atrial cardiopathy did not reveal definite evidence of benefit of apixaban compared to aspirin. Future larger trials employing more atrial specific criteria, such as atrial specific biomarkers, precise LA size measurements, and PAC frequency assessments using prolonged cardiac rhythm monitoring may contribute to a more accurate assessment of risks and treatment efficacies.

Data Availability:

Deidentified participant data is available from the NINDS data repository.