Abstract
Background:
ARCADIA compared apixaban to aspirin for secondary stroke prevention in patients with cryptogenic stroke and atrial cardiopathy. One possible explanation for the neutral result is that biomarkers used did not optimally identify atrial cardiopathy. We examined the relationship between biomarker levels and subsequent detection of AF, the hallmark of atrial cardiopathy.
Methods:
Patients were randomized if they met criteria for atrial cardiopathy, defined as P-wave terminal force >5000 μV*ms in ECG lead V1 (PTFV1), NT-proBNP >250 pg/mL, or left atrial diameter index (LADI) ⩾3 cm/m2. For this analysis, the outcome was AF detected per routine care.
Results:
Of 3745 patients who consented to screening for atrial cardiopathy, 254 were subsequently diagnosed with AF; 96 before they could be randomized and 158 after randomization. In unadjusted analyses, ln(NT-proBNP) (RR per SD, 1.99; 95% CI, 1.85–2.13), PTFV1 (RR per SD, 1.15; 95% CI, 1.03–1.28) and LADI (RR per SD, 1.34; 95% CI, 1.20–1.50) were associated with AF. In a model containing all 3 biomarkers, demographics, and AF risk factors, age (RR per 10 years, 1.24; 95% CI, 1.09–1.41), ln(NT-proBNP) (RR per SD, 1.88; 95% CI, 1.67–2.11) and LADI (RR per SD, 1.25; 95% CI, 1.14–1.37) were associated with AF. These three variables together had a c-statistic of 0.82 (95% CI, 0.79–0.85) but only modest calibration. Discrimination was attenuated in sensitivity analyses of patients eligible for randomization who may have been more closely followed for AF.
Conclusions:
Biomarkers used to identify atrial cardiopathy in ARCADIA were moderately predictive of subsequent AF.
Keywords: Atrial fibrillation, atrial cardiomyopathy, atrial cardiopathy, atrial myopathy, stroke
Graphical abstract.
Introduction
Atrial fibrillation (AF) usually occurs in the setting of atrial cardiopathy – also referred to as atrial cardiomyopathy or atrial myopathy—which is defined as any complex of structural, architectural, contractile or electrophysiological changes affecting the atria with the potential to produce clinically relevant manifestations. 1 In patients with AF, the severity of the underlying atrial cardiopathy is associated with the risk of thromboembolism and stroke. 2 Observational studies have found associations between various biomarkers of atrial cardiopathy and the risk of ischemic stroke even in patients without clinically apparent AF. 3 The Atrial Cardiopathy and Antithrombotic Drugs in Prevention after Cryptogenic Stroke (ARCADIA) trial tested the hypothesis that anticoagulation would be superior to antiplatelet therapy for preventing stroke recurrence in patients with a recent cryptogenic stroke, evidence of atrial cardiopathy, and no apparent AF. One possible explanation for the trial’s neutral result is that the biomarkers or thresholds used to assess eligibility did not optimally identify atrial cardiopathy. We thus examined the relationship between baseline biomarker levels in ARCADIA patients and their likelihood of a subsequent diagnosis of AF, the most widely accepted hallmark of atrial cardiopathy.
Methods
Design
ARCADIA was a multicenter, randomized clinical trial comparing apixaban to aspirin for the prevention of recurrent stroke in patients with a recent cryptogenic stroke and evidence of atrial cardiopathy. The trial was funded by NIH and performed at 185 sites in NIH StrokeNet and the Canadian Stroke Consortium. The trial was approved by the StrokeNet Central Institutional Review Board and the institutional review boards at all participating sites, and all patients provided written, informed consent for trial participation and analysis of their data. Details of the trial rationale and methodology and the primary results have been published previously.4,5 This secondary analysis of the ARCADIA trial data followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines.
Patient population
Patients ⩾45 years old with a cryptogenic ischemic stroke within the prior 180 days were approached for written, informed consent to participate in ARCADIA and undergo biomarker screening for atrial cardiopathy. Cryptogenic stroke was defined as an ischemic stroke that lacked an apparent etiology after standard investigation, with a requirement for: (1) computed tomography or magnetic resonance imaging of the brain to exclude a lacunar infarct; (2) vascular imaging of the cervical and intracranial arteries to exclude large-artery atherosclerosis as the cause of stroke; and (3) transthoracic or transesophageal echocardiography, a 12-lead electrocardiogram (ECG), and ⩾24 h of continuous heart-rhythm monitoring to exclude a major-risk cardioembolic source including any AF. The full list of inclusion and exclusion criteria has been published previously.4,5 Among patients who consented to participate in ARCADIA and underwent biomarker screening, those who met biomarker criteria for atrial cardiopathy and continued to meet the other eligibility criteria were randomly assigned in double-blind fashion to apixaban or aspirin. Atrial cardiopathy was defined as the presence of at least one of the following biomarker criteria: P-wave terminal force in ECG lead V1 (PTFV1) >5000 μV*ms, serum N-terminal pro-B-type natriuretic peptide (NT-proBNP) >250 pg/mL, or left atrial diameter index (LADI) ⩾3 cm/m2 on echocardiogram. These biomarkers and thresholds were based on pilot data and previously reported associations of a twofold higher risk of stroke independent of AF. 3
Measurements
For this analysis, the predictor variables were PTFV1, NT-proBNP, and LADI. LADI was determined by the local echocardiography laboratory at each site. PTFV1 was centrally determined at the study ECG core using previously validated methods. 6 NT-proBNP was centrally measured in a CLIA-certified core laboratory using the Elecsys assay (Roche Diagnostics, Basel, Switzerland). In secondary analyses, we also used measurements of LADI and left atrial volume index (LAVI) taken at our echocardiography core laboratory from clinically obtained transthoracic echocardiograms sent in from enrolling sites.
Our outcome variable for this analysis was a new diagnosis of AF after participants consented to enroll in the trial. Additional heart-rhythm monitoring beyond the minimum 24 h of heart-rhythm monitoring required to diagnose cryptogenic stroke and establish eligibility for enrollment was not mandated but was allowed at the discretion of treating physicians and local investigators, both before and after randomization. In a subset of patients for whom we had data on heart-rhythm monitoring practices, we categorized monitoring as either an external ambulatory monitor or an implantable loop recorder. AF was ascertained by sites per their usual practice without a central definition or adjudication. The ARCADIA protocol specified that any duration of AF after enrollment and before randomization disqualified participants from moving forward with randomization, and sites were required to report any detection of AF as a newly developed exclusion criterion when terminating patients’ further trial participation instead of randomizing them. In patients who had already been randomized, newly detected AF was a prespecified endpoint that sites ascertained according to their own standard clinical practice. Once detected, AF triggered a protocol-mandated discontinuation of study drug with encouragement to start open-label anticoagulation.
Covariates used for this analysis were demographics and vascular comorbidities that have been widely reported as risk factors for AF: age, sex, race, ethnicity, hypertension, diabetes, coronary artery disease, heart failure, peripheral artery disease and tobacco use. 7
Statistical analysis
In patients who did not proceed to randomization, we were unable to use survival analysis because we lacked data on the exact date of AF diagnosis. Therefore, we used relative risk regression to examine the associations between our predictor variables and AF in the overall cohort of enrolled patients. We performed secondary analyses limited to (a) patients who were eligible for randomization and (b) patients who were ultimately randomized; for the latter, we had the exact date when AF was first detected after randomization and thus used survival analysis and Cox regression, with follow-up censored at the end of trial participation. We examined unadjusted associations between each atrial cardiopathy biomarker and AF, then included all three biomarkers in the same model, and then used stepwise backward elimination with a p-value threshold of 0.2 to develop a parsimonious model from among the three biomarkers, age, sex, race, ethnicity, hypertension, diabetes, coronary artery disease, heart failure, peripheral artery disease, and tobacco use. NT-proBNP was natural-log transformed given its skewed distribution. We calculated c-statistics and Harrell’s C to assess discrimination and plotted observed-versus-expected rates to assess model calibration. We calculated the proportion of AF cases that were detected by external ambulatory monitors or implantable loop recorders, and examined the association between atrial cardiopathy biomarkers and the likelihood of heart-rhythm monitoring.
Results
From February 1, 2018 through December 14, 2022, 3745 patients with a cryptogenic stroke consented to participate in the ARCADIA trial and undergo screening for atrial cardiopathy. Of these patients, 254 (6.8%) developed AF; 96 had AF before they could be randomized and were thus not randomized, whereas 158 had AF after randomization. The patients who went on to develop AF were older, more often female, had a higher prevalence of coronary artery disease and heart failure, and had higher levels of atrial cardiopathy biomarkers than patients without AF (Table 1). In a subset of 1633 patients for whom we had heart-rhythm monitoring data, 165 cases of AF were detected, of which 38 were detected on external ambulatory monitors, 82 on implantable loop recorders, and 2 on both. In a model adjusted for all three biomarkers, demographics, and AF risk factors, there was no association with the likelihood of heart-rhythm monitoring for ln(NT-proBNP) (RR per SD, 1.02; 95% CI, 0.97–1.07), PTFV1 (RR per SD, 1.00; 95% CI, 0.96–1.04), and LADI (RR per SD, 1.02; 95% CI, 0.98–1.06).
Table 1.
Characteristics of patients screened for atrial cardiopathy in ARCADIA, stratified by subsequent detection of AF.
| Characteristic a | Atrial fibrillation (N = 254) | No atrial fibrillation (N = 3491) |
|---|---|---|
| Age, mean (SD), years | 71.7 (9.7) | 65.8 (10.6) |
| Female, no. (%) | 134 (52.8%) | 1604 (45.9%) |
| Race, no. (%) [N = 3659] b | ||
| Asian | 2 (0.8%) | 68 (2.0%) |
| Black or African American | 43 (17.1%) | 649 (19.0%) |
| Other | 205 (81.7%) | 2645 (77.6%) |
| White | 1 (0.4%) | 46 (1.3%) |
| Ethnicity, no. (%) [N = 3724] b | ||
| Hispanic or Latino | 15 (6.0%) | 343 (9.9%) |
| Not Hispanic or Latino | 237 (94.0%) | 3129 (90.1%) |
| Medical comorbidities | ||
| Hypertension | 200 (78.7%) | 2512 (72.0%) |
| Prior or current tobacco use | 104 (40.9%) | 1426 (40.8%) |
| Diabetes | 76 (29.9%) | 1052 (30.1%) |
| Coronary artery disease | 26 (10.2%) | 265 (7.6%) |
| Heart failure | 18 (7.1%) | 136 (3.9%) |
| Peripheral artery disease | 5 (2.0%) | 67 (1.9%) |
| Atrial cardiopathy biomarkers | ||
| PTFV1, mean (SD), μV*ms [N = 3673] | 3798 (2,660) | 3426 (2,259) |
| NT-proBNP, median (IQR), pg/mL [N = 3580] | 416 (258–751) | 96 (45–231) |
| LA diameter index, mean (SD), cm/m2 [N = 3125] | 2.1 (0.4) | 1.8 (0.4) |
| Days from stroke to biomarker screening, mean (SD) | 1.9 (4.3) | 3.0 (25.3) |
AF: atrial fibrillation; IQR: interquartile range; LA: left atrial; NT-proBNP: amino terminal pro-B-type natriuretic peptide; PTFV1: P-wave terminal force in lead V1; SD: standard deviation.
Percentages may not total 100 because of rounding.
Other race was defined as American Indian or Alaska Native, Native Hawaiian or Other Pacific Islander, or more than one race. Site investigators and coordinators were instructed to directly ask participants to report their self-identified race and ethnicity, which were then categorized per NIH guidelines.
In unadjusted analyses, ln(NT-proBNP) (RR per SD, 1.99; 95% CI, 1.85–2.13), PTFV1 (RR per SD, 1.15; 95% CI, 1.03–1.28), and LADI (RR per SD, 1.34; 95% CI, 1.20–1.50) were all associated with an increased risk of AF. In a model containing all 3 biomarkers, demographics, and AF risk factors, age (RR per 10 years, 1.24; 95% CI, 1.09–1.41), ln(NT-proBNP) (RR per SD, 1.88; 95% CI, 1.67–2.11) and LADI (RR per SD, 1.25; 95% CI, 1.14–1.37) were associated with AF (Table 2).
Table 2.
Associations between baseline atrial cardiopathy biomarkers and subsequent detection of atrial fibrillation in ARCADIA.
| Biomarker | Model 1 | Model 2 | Model 3 |
|---|---|---|---|
| PTFV1 | 1.15 (1.03–1.28) | 1.03 (0.92–1.14) | - |
| NT-proBNP | 1.99 (1.85–2.13) | 1.83 (1.69–1.97) | 1.88 (1.67–2.11) |
| LADI | 1.34 (1.20–1.50) | 1.25 (1.14–1.38) | 1.25 (1.14–1.37) |
LADI: left atrial diameter index; NT-proBNP: amino terminal pro-B-type natriuretic peptide; PTFV1: P-wave terminal force in lead V1.
Data are presented as risk ratios (95% confidence intervals) per standard-deviation increase in the atrial cardiopathy biomarker variables. Model 1 was unadjusted. Model 2 included all three biomarkers together. Model 3 additionally adjusted for age, sex, race, ethnicity, hypertension, diabetes, coronary artery disease, heart failure, peripheral artery disease, and tobacco use, with variables reduced using stepwise reverse selection with a p-value threshold of 0.2.
A model comprised of the three atrial cardiopathy biomarkers had a c-statistic of 0.82 (95% CI, 0.79–0.85) and modest calibration (Figure 1). Discrimination was unchanged when we dropped PTFV1 from the model (c-statistic, 0.82; 95% CI, 0.79–0.85) and instead added age (c-statistic, 0.82; 95% CI, 0.79–0.85). Discrimination was also unchanged when instead of local LADI measurements we used LADI or LAVI as measured in our core echocardiography laboratory (Table 3). Calibration was also similar across these models (data not shown).
Figure 1.
Calibration of atrial cardiopathy biomarkers for predicting AF in ARCADIA.
Each open circle represents 1 of 20 groups of ARCADIA trial participants. Patients were grouped by their predicted probability of atrial fibrillation (AF) based on a relative risk regression model comprised of NT-proBNP, left atrial dimension index, and P-wave terminal force in ECG lead V1. The circle’s position on the x-axis represents the group’s predicted probability of atrial fibrillation. The circle’s position on the y-axis represents the actual proportion of patients in the group who developed atrial fibrillation. The dashed blue line represents perfect calibration
Table 3.
Discrimination of baseline atrial cardiopathy biomarkers for predicting subsequent AF in ARCADIA.
| Variables included in predictive model a | C-statistic (95% CI) b |
|---|---|
| PTFV1, NT-proBNP, LADI | 0.82 (0.79–0.85) |
| NT-proBNP, LADI | 0.82 (0.79–0.85) |
| NT-proBNP, LADIcentral | 0.80 (0.77–0.83) |
| NT-proBNP, LAVIcentral | 0.81 (0.78–0.84) |
| NT-proBNP, LADI, age | 0.82 (0.79–0.85) |
| NT-proBNP | 0.80 (0.77–0.83) |
| LADI | 0.67 (0.63–0.72) |
| LADIcentral | 0.67 (0.64–0.71) |
| LADcentral | 0.66 (0.63–0.70) |
| LAVIcentral | 0.71 (0.67–0.74) |
| PTFV1 | 0.54 (0.50–0.58) |
| Age | 0.66 (0.63–0.70) |
AF: atrial fibrillation; CI: confidence interval; LAD: left atrial dimension; LADI: left atrial diameter index; LAVI: left atrial volume index; NT-proBNP: amino terminal pro-B-type natriuretic peptide; PTFV1: P-wave terminal force in lead V1.
NT-proBNP (N = 3580) and PTFV1 (N = 3673) were centrally measured. LADI was determined by the local echocardiography laboratory at each site (N = 3008), whereas LADcentral [N = 3136], LADIcentral [N = 3125], and LAVIcentral [N = 2897] were measured at the echocardiography core laboratory from clinically obtained transthoracic echocardiograms sent in from enrolling sites.
When comparing the discrimination of individual atrial cardiopathy biomarkers, ln(NT-proBNP) had better discrimination than locally measured LADI or centrally measured LAD, LADI, or LAVI (p < 0.001 for all). Discrimination did not differ significantly between local versus central LADI measurements (p = 0.29), local LADI versus central LAVI measurements (p = 0.11), central LAD versus LADI measurements (p = 0.70), or central LADI versus LAVI measurements (p = 0.20).
In secondary analyses limited to patients who met at least one of the atrial cardiopathy biomarkers and were thus eligible for randomization, significant associations with AF remained for both ln(NT-proBNP) (RR per SD, 1.30; 95% CI, 1.15–1.47) and LADI (RR per SD, 1.18; 95% CI, 1.10–1.26), but the associations were attenuated compared to the full cohort and a model with the 3 biomarkers had relatively low discrimination (c-statistic, 0.67; 95% CI, 0.63–0.71) (Table 4). PTFV1 was inversely associated with AF (RR per SD, 0.84; 95% CI, 0.74–0.95), likely due to interdependencies induced by selection based on biomarker eligibility. Our findings were similar when analyzing only randomized patients who were followed longitudinally during study treatment, among whom we found associations with AF for both ln(NT-proBNP) (HR per SD, 1.38; 95% CI, 1.15–1.66) and LADI (HR per SD, 1.25; 95% CI, 1.11–1.41) with modest discrimination for the model with all 3 biomarkers (Harrell’s C, 0.69; Table 4). AF was diagnosed at a median 30 (IQR, 8–59) weeks after randomization.
Table 4.
Sensitivity analysis of associations between baseline atrial cardiopathy biomarkers and subsequent detection of atrial fibrillation in the ARCADIA trial.
| Patient population | PTFV1 a | NT-proBNP a | LADI a | Discrimination b |
|---|---|---|---|---|
| Eligible | 0.84 (0.74–0.95) | 1.30 (1.15–1.47) | 1.18 (1.10–1.26) | 0.67 |
| Randomized | 0.81 (0.68–0.95) | 1.38 (1.15–1.66) | 1.25 (1.11–1.41) | 0.69 |
LADI: left atrial diameter index; NT-proBNP: amino terminal pro-B-type natriuretic peptide; PTFV1: P-wave terminal force in lead V1.
Data are presented as risk ratios (95% confidence intervals) or hazard ratios (95% confidence intervals) per standard-deviation increase in the atrial cardiopathy biomarker variables in a model that included all three biomarkers together.
Data are presented as c-statistics or Harrell’s C.
Discussion
In this secondary analysis of the ARCADIA trial, we found that the biomarkers used to define atrial cardiopathy were associated with and moderately predictive of subsequent detection of AF. Quantitative levels of several of the atrial cardiopathy biomarkers continued to be associated with future AF detection even among all patients eligible for randomization or among those who were actually randomized and followed longitudinally, but with reduced predictive performance.
The biomarkers used to define atrial cardiopathy in ARCADIA have previously been associated with the risk of AF and ischemic stroke in community-based cohorts of mostly stroke-free individuals.8–13 Several of these biomarkers have also been associated with the likelihood of identifying AF after cryptogenic stroke, with inconsistent findings reported for LA size or volume14–16 as compared to NT-proBNP which has been consistently reported as a highly predictive biomarker of future AF after cryptogenic stroke.16–18 Based on these previous findings, we and others had hypothesized that these atrial cardiopathy biomarkers would identify a subset of cryptogenic stroke patients who would be at high risk of left atrial thromboembolism and would benefit from anticoagulation. 3 However, we found no benefit of apixaban over aspirin for preventing recurrent stroke in patients selected into the ARCADIA trial on the basis of these biomarkers. One possible reason for the trial’s neutral results is that these biomarkers failed to satisfactorily identify underlying atrial cardiopathy. In this context, the results of this secondary analysis shed additional light on the trial’s neutral findings. The chosen atrial cardiopathy biomarkers were associated with and predictive of subsequent AF, but imperfectly so, and the majority of enrolled patients did not demonstrate evidence of AF during the trial. Given that patients with even brief episodes of subclinical AF have been recently found to benefit from anticoagulation for primary stroke prevention,19–21 future efforts may be warranted to identify atrial cardiopathy markers that more reliably predict future AF. This may allow more focused efforts to optimally screen for AF, perhaps allowing other trials of empiric anticoagulation or left atrial appendage occlusion in patients at very high risk of AF.
Our results have implications for clinical practice and future research. The ARCADIA trial found no benefit of anticoagulation for secondary stroke prevention in patients enriched for atrial cardiopathy and future AF risk. On the other hand, anticoagulation has recently been shown to reduce stroke in patients with even brief episodes of subclinical AF.19–21 Combined, these results suggest that anticoagulating patients based on currently widely available biomarkers of underlying AF pathophysiology or a prediction of future AF risk is not beneficial and that AF itself remains the optimum biomarker for starting anticoagulation. The MOSES trial is evaluating anticoagulation for secondary stroke prevention in patients with atrial cardiopathy as defined by mid-regional pro-atrial natriuretic peptide. 22 The Find-AF2 trial is comparing usual care to two tiers of prolonged heart-rhythm monitoring guided by the degree of AF risk, assessed via baseline atrial ectopy. 23 It may be worthwhile to perform other randomized trials comparing various biomarker-guided strategies, 24 potentially combined with additional risk factors as proposed in previous clinical scores, 25 for prolonged heart-rhythm monitoring after stroke, for example excluding the lowest-risk patients from monitoring and empirically anticoagulating the highest-risk patients based on predictive biomarkers.
This analysis has several limitations. First, trial participants who met the atrial cardiopathy biomarker thresholds may have been followed longer and more closely by study sites as they prepared to randomly assign them to study treatments. Such ascertainment bias could have caused a spurious association between biomarker positivity and subsequent AF detection. We found significant associations between the biomarkers and AF even among those who met the eligibility criteria for randomization and among those who were actually randomized, but these latter associations were attenuated. Second, AF was determined by local sites per usual clinical practice without central definition or adjudication, which may have introduced heterogeneity and misclassification, and we lacked data on the clinical characteristics of AF episodes (such as paroxysmal, persistent, clinical and subclinical). Third, we lacked data on additional clinical factors that are associated with an increased risk of AF (such as body mass index in non-randomized patients, chronic kidney disease, obstructive sleep apnea and thyroid disease).
Conclusion
The biomarkers used to identify atrial cardiopathy in the ARCADIA trial were associated with and predictive of subsequent AF detection, suggesting that the neutral results of the trial were not entirely due to suboptimal biomarkers of atrial cardiopathy. However, the predictive performance of the biomarkers was modest, supporting further research to identify other measures that can identify a more severe form of atrial cardiopathy with a high risk of AF.
Acknowledgments
None.
Footnotes
The author(s) declared the following potential conflicts of interest with respect to the research, authorship, and/or publication of this article: Dr. Kamel reports serving as Deputy Editor for JAMA Neurology; serving on clinical trial steering/executive committees for the STROKE-AF (Medtronic), LIBREXIA-AF (Janssen), and LAAOS-4 (Boston Scientific) trials; serving as a consultant or endpoint adjudication committee member for AbbVie, AstraZeneca, Boehringer Ingelheim, and Novo Nordisk; and household ownership interests in TETMedical, Spectrum Plastics Group, and Ascential Technologies. Dr. Healey reports research grants and speaking fees from BMS/Pfizer, Servier, Boston Scientific, and Medtronic. Dr. Kasner reports grant funding from Bayer, Bristol-Myers Squibb, Daiichi Sankyo, Genentech, Medtronic, and Diamedica, consulting fees from AstraZeneca, Abbie, and Novo Nordisk, and royalties from UpToDate. Dr. Elkind reports employment by the American Heart Association, royalties from UpToDate for chapters on stroke, and honoraria for lectures from the Atria Academy for Science and Medicine. The other authors report no conflicts of interest.
Funding: The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: NIH funded the trial, the BMS-Pfizer Alliance provided in-kind study drug to the StrokeNet Central Pharmacy for distribution, and Roche Diagnostics provided ancillary funding for laboratory supplies for amino terminal pro-B-type natriuretic peptide (NT-proBNP) assays. The funders played no role in the design and conduct of the study or the collection, management, analysis, and interpretation of the data. An NIH representative (Janis) reviewed and approved the manuscript; the BMS-Pfizer and Roche representatives were sent the manuscript to review before submission for publication but their approval was not required. None of the funders had the right to veto publication or to control the decision regarding to which journal the paper was submitted.
Ethical approval: The trial was approved by the StrokeNet Central Institutional Review Board and the institutional review boards at all participating sites.
Informed consent: All patients provided written, informed consent for trial participation and analysis of their data.
Guarantor: Dr. Kamel
Contributorship: All authors contributed to the concept and design of the study, and to the acquisition, analysis, or interpretation of data. Dr. Kamel drafted the first version of the manuscript, and all other authors provided critical review of the manuscript for important intellectual content. Dr. Kamel performed the statistical analysis. Drs. Elkind, Kamel, Longstreth, and Tirschwell obtained funding.
ORCID iD: Hooman Kamel 
https://orcid.org/0000-0002-5745-0307
References
- 1. Goette A, Kalman JM, Aguinaga L, et al. EHRA/HRS/APHRS/SOLAECE expert consensus on atrial cardiomyopathies: definition, characterization, and clinical implication. Heart Rhythm 2017; 14: e3–e40. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2. Hirsh BJ, Copeland-Halperin RS, Halperin JL. Fibrotic atrial cardiomyopathy, atrial fibrillation, and thromboembolism: mechanistic links and clinical inferences. J Am Coll Cardiol 2015; 65: 2239–2251. [DOI] [PubMed] [Google Scholar]
- 3. Kamel H, Okin PM, Elkind MS, et al. Atrial fibrillation and mechanisms of stroke: time for a new model. Stroke 2016; 47: 895–900. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4. Kamel H, Longstreth Wt, Jr, Tirschwell DL, et al. The AtRial cardiopathy and antithrombotic drugs in prevention after cryptogenic stroke randomized trial: Rationale and methods. Int J Stroke 2019; 14: 207–214. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5. Kamel H, Longstreth WT, Jr, Tirschwell DL, et al. Apixaban to prevent recurrence after cryptogenic stroke in patients with atrial cardiopathy: the ARCADIA randomized clinical trial. JAMA 2024; 331(7): 573–581. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6. Kamel H, O’Neal WT, Okin PM, et al. Electrocardiographic left atrial abnormality and stroke subtype in the atherosclerosis risk in communities study. Ann Neurol 2015; 78: 670–678. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7. Elliott AD, Middeldorp ME, Van Gelder IC, et al. Epidemiology and modifiable risk factors for atrial fibrillation. Nat Rev Cardiol 2023; 20: 404–417. [DOI] [PubMed] [Google Scholar]
- 8. Patton KK, Ellinor PT, Heckbert SR, et al. N-terminal pro-B-type natriuretic peptide is a major predictor of the development of atrial fibrillation: the Cardiovascular Health Study. Circulation 2009; 120: 1768–1774. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9. Soliman EZ, Prineas RJ, Case LD, et al. Ethnic distribution of ECG predictors of atrial fibrillation and its impact on understanding the ethnic distribution of ischemic stroke in the atherosclerosis risk in communities (ARIC) study. Stroke 2009; 40: 1204–1211. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10. Vaziri SM, Larson MG, Benjamin EJ, et al. Echocardiographic predictors of nonrheumatic atrial fibrillation. The Framingham Heart Study. Circulation 1994; 89: 724–730. [DOI] [PubMed] [Google Scholar]
- 11. Folsom AR, Nambi V, Bell EJ, et al. Troponin T, N-terminal pro-B-type natriuretic peptide, and incidence of stroke: the atherosclerosis risk in communities study. Stroke 2013; 44: 961–967. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12. Kamel H, Soliman EZ, Heckbert SR, et al. P-wave morphology and the risk of incident ischemic stroke in the multi-ethnic study of atherosclerosis. Stroke 2014; 45: 2786–2788. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13. Benjamin EJ, D'Agostino RB, Belanger AJ, et al. Left atrial size and the risk of stroke and death. The Framingham Heart Study. Circulation 1995; 92: 835–841. [DOI] [PubMed] [Google Scholar]
- 14. Thijs VN, Brachmann J, Morillo CA, et al. Predictors for atrial fibrillation detection after cryptogenic stroke: results from CRYSTAL af. Neurology 2016; 86: 261–269. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15. Favilla CG, Ingala E, Jara J, et al. Predictors of finding occult atrial fibrillation after cryptogenic stroke. Stroke 2015; 46: 1210–1215. [DOI] [PubMed] [Google Scholar]
- 16. Kneihsl M, Bisping E, Scherr D, et al. Predicting atrial fibrillation after cryptogenic stroke via a clinical risk score—a prospective observational study. Eur J Neurol 2022; 29: 149–157. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17. Kneihsl M, Gattringer T, Bisping E, et al. Blood biomarkers of heart failure and hypercoagulation to identify atrial fibrillation-related stroke. Stroke 2019; 50: 2223–2226. [DOI] [PubMed] [Google Scholar]
- 18. Fonseca AC, Brito D, Pinho e, Melo T, et al. N-terminal pro-brain natriuretic peptide shows diagnostic accuracy for detecting atrial fibrillation in cryptogenic stroke patients. Int J Stroke 2014; 9: 419–425. [DOI] [PubMed] [Google Scholar]
- 19. Healey J, Lopes R, Granger C, et al. Apixaban for stroke prevention in subclinical atrial fibrillation. New Engl J Med 2024; 390: 107–117. [DOI] [PubMed] [Google Scholar]
- 20. Kirchhof P, Toennis T, Goette A. Anticoagulation with edoxaban in patients with atrial high-rate episodes. New Engl J Med 2023; 389: 2301–2303. [Google Scholar]
- 21. McIntyre WF, Benz AP, Becher N, et al. Direct oral anticoagulants for stroke prevention in patients with device-detected atrial fibrillation: a study-level meta-analysis of the NOAH-AFNET 6 and ARTESiA trials. Circulation 2024; 149: 981–988–Epub 2023 Nov 12. DOI: 10.1161/circulationaha.123.067512 [DOI] [PubMed] [Google Scholar]
- 22. Schweizer J, Arnold M, König IR, et al. Measurement of midregional pro-atrial natriuretic peptide to discover atrial fibrillation in patients with ischemic stroke. J Am Coll Cardiol 2022; 79: 1369–1381. [DOI] [PubMed] [Google Scholar]
- 23. Wachter R, Gröschel K, Gelbrich G, et al. Holter-electrocardiogram-monitoring in patients with acute ischaemic stroke (Find-AF RANDOMISED): an open-label randomised controlled trial. Lancet Neurol 2017; 16: 282–290. [DOI] [PubMed] [Google Scholar]
- 24. Cameron A, Cheng HK, Lee R-P, et al. Biomarkers for atrial fibrillation detection after stroke. Neurology 2021; 97: e1775–e1789. [DOI] [PubMed] [Google Scholar]
- 25. Palaiodimou L, Theodorou A, Triantafyllou S, et al. Performance of different risk scores for the detection of atrial fibrillation among patients with cryptogenic stroke. Stroke 2024; 55: 454–462. [DOI] [PubMed] [Google Scholar]