This was a retrospective cohort study conducted using EHR data. The study followed patients with AF, examining the association between PP and the risk of major bleeding events. OAC status was recorded as a baseline covariate.

The study was conducted using data from REACHnet, a regional PCORnet-affiliated clinical data network based in Louisiana. REACHnet integrates longitudinal clinical data from multiple health systems across the region, capturing information on demographics, diagnoses, medications, vital signs, procedures, and outcomes. The data network draws from a diverse patient population receiving care across outpatient, inpatient, and specialty settings within these affiliated health systems.

The study period spanned from 2010 to 2019. Patients were identified, and their index dates (the date of AF diagnosis) were established within this window. Data collection for baseline covariates occurred prior to the index date, and follow-up continued from the index date until the occurrence of a primary outcome event, loss to follow-up, death, or December 31, 2019, whichever came first.

Patient identification followed a nonprobability, consecutive sampling approach. All patients within the REACHnet who met the eligibility criteria during the study period (2010‐2019) were included in the analysis; no random or purposive sampling was applied. This census-type approach was chosen to maximize sample size and minimize selection bias within the available data source. The index date was defined as the date AF diagnosis had been recorded during the study window.

A total of 4935 patients with AF were identified and included after applying all eligibility criteria. No formal a priori sample size or power calculation was performed. The study enrolled all eligible patients identified within the REACHnet during the study period (2010‐2019). Statistical precision of the primary estimates is reflected in the 95% CIs reported in the Results section.

This study involved a secondary analysis of deidentified EHR data obtained through REACHnet and did not constitute human subjects research requiring institutional review board review under 45 CFR §46.101(b) [19]. Accordingly, institutional review board approval and informed consent were not required. Informed consent for primary data collection was obtained by the participating health systems as part of routine clinical care, and the applicable data use agreements permitted secondary research use without additional patient consent. All data were deidentified prior to analysis in accordance with the HIPAA (Health Insurance Portability and Accountability Act) privacy rule, and no direct patient identifiers were available to the research team at any stage. Data were accessed and stored in a secure, access-controlled environment, consistent with REACHnet governance requirements. No compensation was provided, as no direct participant contact occurred. This paper and all supplementary materials present results in aggregate form only, and no images or descriptions that could identify individual participants are included.

The participant flowchart is presented in Figure 1. Among 4935 patients with AF, 1560 (n=472, 30.3% female) were in the lowest tertile of PP (T1), 1669 (n=542, 32.5% female) in the middle tertile (T2), and 1706 (n=592, 34.7% female) in the highest tertile (T3).

Patients in the highest PP tertile were significantly older (mean age 66.6, SD 9.3 y) than those in T2 (mean age 63.0, SD 11.2 y) and T1 (mean age 61.3, SD 11.7 y; P<.001). Comorbidities, including hypertension, diabetes, CKD, congestive heart failure, peripheral artery disease, anemia, and prior stroke, were more prevalent in the highest PP tertile (all P<.05). The use of β-blockers and ACE inhibitors was also more common in higher PP tertiles (Table 1).

During a median follow-up of 5 years, among 4935 patients, 677 (13.7%) experienced a bleeding event, while 4258 (86.3%) remained event-free. Overall, intracranial bleeding occurred in 60 (1.2%) patients, gastrointestinal bleeding in 195 (4.0%), and other bleeding events in 149 (3.0%).

Among the specific bleeding subtypes, only gastrointestinal bleeding showed a statistically significant difference across PP tertiles (P=.007), with the highest incidence observed in the upper tertile group. In contrast, rates of intracranial and other bleeding events did not differ significantly between groups (P=.08 and P=.32, respectively). Full subgroup counts are presented in Table 2.

Kaplan-Meier survival analysis demonstrated a statistically significant association between higher PP tertiles and an increased risk of gastrointestinal bleeding (log-rank P=.004), with patients in the highest tertile (T3) experiencing the lowest bleeding-free survival over time. In contrast, no significant differences were observed in intracranial bleeding-free survival across tertiles (P=.08), nor in the risk of other bleeding events (P=.58). For the composite outcome of any bleeding, although some separation of curves was noted, the difference did not reach statistical significance (P=.22; Figures 2-5).

In Figures 2-5, Kaplan-Meier curves depicting bleeding-free survival by PP tertile (T1: low <46 mm Hg, T2: middle 46‐62 mm Hg, T3: high >62 mm Hg) in 4935 adults with AF (REACHnet, Louisiana, United States, 2010‐2019). Follow-up extended from the date of AF diagnosis to the first bleeding event, death, loss to follow-up, or December 31, 2019. Outcomes depicted are as follows: Figure 2, composite bleeding (any gastrointestinal, intracranial, or other clinically significant bleeding; log-rank P=.22); Figure 3, gastrointestinal bleeding (log-rank P=.004); Figure 4, ICH (log-rank P=.08); and Figure 5, other clinically significant bleeding (log-rank P=.58). Group differences were assessed using the log-rank test.

In multivariable Cox proportional hazards models, PP emerged as a significant independent predictor of gastrointestinal bleeding. Each 10 mm Hg increase in PP was independently associated with a 14.9% higher risk of gastrointestinal bleeding (hazard ratio [HR] 1.149, 95% CI 1.010‐1.318; P=.04), even after adjusting for SBP, OAC usage, and a comprehensive set of clinical covariates, including age, sex, race, hypertension, diabetes, CKD, heart failure, anemia, and medication use (HR per 1 mm Hg=1.014, 95% CI 1.001‐1.028; P=.04; Table 3). PP remained not significantly associated with intracranial bleeding (HR 1.008, 95% CI 0.986‐1.030; P=.50), other bleeding (HR 0.990, 95% CI 0.976‐1.005; P=.19), or any bleeding (HR 1.003, 95% CI 0.996‐1.011; P=.36).

Given the known differential association between the anticoagulant type and gastrointestinal bleeding risk, Kaplan-Meier analyses were performed separately for individuals treated with warfarin and DOAC. Among patients treated with warfarin (T1: n=134, T2: n=157, and T3: n=194), no significant differences in gastrointestinal bleeding-free survival were observed across PP tertiles (log-rank P=.96). Similarly, among patients treated with DOAC (T1: n=226, T2: n=215, and T3: n=242), gastrointestinal bleeding-free survival did not differ significantly across tertiles (log-rank P=.29; Figures 6 and 7).

In Figures 6 and 7, Kaplan-Meier curves depict gastrointestinal bleeding-free survival by PP tertiles (T1: low <46 mm Hg, T2: middle 46‐62 mm Hg, T3: high >62 mm Hg), stratified by the anticoagulant type in patients with AF (REACHnet, Louisiana, United States, 2010‐2019). The follow-up extended from the date of AF diagnosis to the first gastrointestinal bleeding event, death, loss to follow-up, or December 31, 2019. Figure 6 depicts patients treated with DOAC (T1: n=134, T2: n=157, and T3: n=194; log-rank P=.96). Figure 7 depicts patients treated with warfarin (T1: n=226, T2: n=215, and T3: n=242; log-rank P=.29). Group differences were assessed using the log-rank test.

This study aimed to evaluate whether elevated PP independently predicts major bleeding events in patients with AF. Consistent with our hypothesis, higher PP was independently associated with an increased risk of gastrointestinal bleeding after adjusting for established clinical risk factors, including SBP and the ORBIT bleeding score. This association was specific to gastrointestinal bleeding and was not observed for ICH, other bleeding subtypes, or the composite bleeding outcome, suggesting a potentially organ-specific hemodynamic mechanism.

Hypertension is a well-established contributor to bleeding risk in patients with AF. The HAS-BLED score assigns 1 point for SBP >160 mm Hg [20]. High SBP has been linked to intracerebral hemorrhage in patients on warfarin and to higher overall bleeding rates in older patients with AF cohorts [21]. In a recent analysis of Japanese octogenarians with AF, those with poorly controlled home SBP (≥145 mm Hg) had a significantly higher incidence of major bleeding and stroke than those with SBP <125 mm Hg [21]. However, these conventional approaches treat BP as a static or binary variable and do not account for the pulsatile hemodynamic forces that may independently contribute to vascular injury [22]. PP, defined as the difference between SBP and DBP, reflects arterial stiffness and the magnitude of pulsatile stress transmitted to end-organ vasculature, a dimension of hemodynamic risk that existing bleeding scores do not capture [23].

To our knowledge, no prior large-scale study has examined PP in relation to bleeding outcomes in patients with AF. Previous investigations of PP have focused predominantly on its role as a predictor of cardiovascular events such as myocardial infarction, stroke, and cardiovascular mortality, particularly in older adults and hypertensive populations [24]. The extension of PP as a risk marker to bleeding outcomes represents a novel application of this readily available hemodynamic parameter, and this study was designed to address this gap in the literature.

In this large cohort of 4935 patients with AF, higher PP was independently associated with an increased risk of gastrointestinal bleeding. Patients in the highest PP tertile were older and had a higher burden of vascular comorbidities. In multivariable Cox regression, a clinically meaningful and statistically robust increase in gastrointestinal bleeding hazard was observed per 10 mm Hg increment in PP, even after adjusting for SBP, OAC usage, and the ORBIT bleeding score in logistic regression. The consistent direction and statistical significance of this association across both Cox and logistic regression models, with CIs excluding the null in both analyses, strengthen confidence in PP as an independent predictor of gastrointestinal bleeding. The magnitude of effect is clinically plausible, given the hemodynamic burden that elevated PP imposes on the gastrointestinal vasculature. To our knowledge, this is the first study to demonstrate PP as an independent predictor of gastrointestinal bleeding in patients with AF.

Gastrointestinal bleeding remains one of the most common and clinically impactful bleeding complications among patients with AF [25,26]. Among older patients with AF, the incidence of gastrointestinal bleeding was 1.92 per 100 person-years, highlighting the real-world burden and recurrence potential of this outcome [27]. A recent large international study (INTERBLEED) identified age as the strongest predictor of gastrointestinal bleeding in patients with cardiovascular disease, with those aged ≥71 years having more than 4-fold higher risk than those ≤60 years (odds ratio 4.16, P<.001) [28]. While aging remains a robust risk factor, its effects are likely mediated through vascular stiffening and elevated PP, a hemodynamic consequence of reduced arterial compliance. In our analysis, PP remained independently associated with gastrointestinal bleeding even after adjusting for age, suggesting that PP may reflect cumulative subclinical vascular damage beyond chronological aging. Mechanistically, PP is a surrogate for arterial stiffness; as arteries stiffen with age and hypertension, they lose elasticity and the ability to dampen the pulsatile output of the heart [29]. This loss of arterial buffering capacity results in the greater transmission of pulsatile energy to downstream microvascular beds, including those of the gastrointestinal tract [23]. The splanchnic circulation, which receives a substantial proportion of cardiac output, may be particularly vulnerable to this augmented pulsatile stress [30].

Our finding that PP was not significantly related to ICH risk hints at possible organ-specific differences. Intracerebral microvessels, which are also sensitive to hypertension, might be more acutely affected by absolute BP spikes or long-term hypertensive remodeling rather than by pulsatile pressure per se [31]. The cerebral vasculature possesses unique autoregulatory mechanisms, including the myogenic response and neurovascular coupling, which actively modulate blood flow across a range of perfusion pressures [32]. These protective mechanisms may partially attenuate the impact of pulsatile stress on cerebral microvessels, whereas the gastrointestinal vasculature lacks comparably robust autoregulatory defenses [33]. In contrast, gastrointestinal bleeding sources (such as submucosal arterioles or angiodysplastic vessels) may be more directly influenced by the ongoing pulsatile stress that PP represents [34]. This pathophysiologic distinction warrants further investigation, but it aligns with our results showing PP’s impact on gastrointestinal bleeds and the relative lack of effect on ICH.

Our findings have several practical implications. Bleeding risk scores such as HAS-BLED and ORBIT could be enhanced by incorporating PP or related measures of arterial stiffness. Current scoring systems typically treat hypertension as a binary variable, without explicitly capturing the hemodynamic burden of vascular stiffness [21]. Given that PP is derived from standard BP measurements already obtained in routine clinical practice, its integration into existing risk models would impose no additional cost or procedural burden [24]. A refined scoring system that incorporates PP as a continuous variable, or as a categorical threshold, could improve the discrimination of patients at the highest risk for gastrointestinal bleeding, thereby enabling more targeted surveillance and intervention strategies.

In our analysis, PP remained an independent predictor of gastrointestinal bleeding even after accounting for SBP and overall bleeding risk as captured by the ORBIT score, whereas SBP alone did not demonstrate a significant association. This underscores that the pulsatile component of BP, rather than absolute systolic values, may better reflect vascular fragility. Since PP can be easily calculated from routine BP readings, its incorporation into clinical workflows may enhance bleeding-risk discrimination, particularly for gastrointestinal events. Clinically, patients with elevated PP might warrant closer surveillance and more stringent BP control. Specifically, clinicians managing patients with AF and elevated PP could consider more frequent hemoglobin monitoring, lower thresholds for endoscopic evaluation of gastrointestinal symptoms, and optimization of antihypertensive regimens with agents known to reduce arterial stiffness, such as ACE inhibitors, angiotensin receptor blockers, or calcium channel blockers, which may preferentially lower PP relative to other drug classes [35,36]. Whether pharmacologic reduction of PP translates into a measurable decrease in gastrointestinal bleeding events remains an important question for future interventional studies [37].

Another forward-looking implication involves the potential role of technology in monitoring PP and arterial health. The proliferation of wearable BP devices enables near-continuous tracking of both systolic SBP and DBP in real-world settings [38]. Frequent out-of-office BP measurements using smartwatch sensors can detect patterns such as episodic hypertension or increased PP with much greater resolution than periodic clinic visits [39]. Integrating these technologies with clinical risk models could facilitate real-time, individualized bleeding-risk assessment. For example, machine-learning algorithms applied to continuous PP data streams could identify patients whose PP trajectories are trending upward, triggering clinical alerts before a bleeding event occurs [40]. Collectively, these approaches highlight the promise of PP as an accessible biomarker bridging traditional risk stratification and digital precision medicine.

This study has several limitations that should be considered when interpreting the findings. PP was derived from a single peripheral measurement, which may not fully capture the dynamic or central hemodynamic burden. However, this proof-of-concept analysis supports the need for future studies employing continuous monitoring or central PP estimation. Second, the retrospective observational design inherently carries risks of residual confounding and bias, despite multivariable adjustment. Prospective validation in diverse cohorts is warranted. Third, the quality of anticoagulation in patients treated with warfarin, specifically prothrombin time–international normalized ratio and time in therapeutic range, was not available in the EHR data and could not be accounted for in the analysis. As anticoagulation intensity is a known determinant of bleeding risk, this represents a potential source of residual confounding that may have influenced the observed bleeding outcomes.

Our study identifies elevated PP as a significant and independent predictor of gastrointestinal bleeding among patients with AF, representing the first evidence of this association in this population. The organ-specific nature of this finding, with PP predicting gastrointestinal but not intracranial or overall bleeding, suggests that distinct hemodynamic pathways underlie different bleeding subtypes, and that subtype-specific risk modeling may be more informative than composite bleeding end points alone. Unlike existing risk scores that treat hypertension as a binary variable, PP captures the pulsatile hemodynamic burden of arterial stiffness and may therefore improve bleeding risk discrimination when incorporated into clinical prediction models such as HAS-BLED or ORBIT. In our sensitivity analysis, PP remained independently predictive of gastrointestinal bleeding after adjusting for SBP, which itself was not significant, underscoring that the pulsatile rather than the absolute pressure component drives this risk. Patients with AF and elevated PP may represent a high-risk subgroup warranting heightened gastrointestinal surveillance, proactive gastroprotection, and careful selection of anticoagulant agents with favorable gastrointestinal safety profiles. Future research should aim to validate these findings in broader settings and explore how best to incorporate PP into clinical practice, whether by refining risk scores, targeting arterial stiffness reduction pharmacologically, guiding BP targets, or leveraging wearable technology for continuous vascular monitoring.